CN116739154A - Fault prediction method and related equipment thereof - Google Patents

Abstract

This application discloses a fault prediction method and related equipment. In the process of fault prediction for equipment, relatively comprehensive factors are considered, so that the final fault prediction result of the equipment can have high accuracy. The method of this application includes: obtaining historical fault information of the target device. The historical fault information is used to indicate N faults that have occurred in the target device from the first moment to the T-1 moment, T≥2, N≥2; Based on historical fault information, the causal relationship between N faults and the weight of N faults are obtained. The weight of N faults is used to indicate the importance of N faults; based on historical fault information, the causal relationship between N faults and The weight of N faults is used to obtain the probability of N faults occurring in the target device at the T-th moment.

Description

A fault prediction method and related equipment

Technical field

The embodiments of the present application relate to the technical field of artificial intelligence (AI), and in particular, to a fault prediction method and related equipment.

Background technique

Equipment failure prediction refers to the fact that the equipment failure has not yet occurred. Through the neural network model in AI technology, it is predicted whether a certain or certain failures may occur in the equipment based on certain information of the equipment, so as to inform the equipment engineers of the failure prediction results, thus making Equipment engineers can repair equipment in time based on fault prediction results.

In related technologies, historical fault information of the equipment can be obtained first. The historical fault information is used to indicate multiple faults that have occurred in the equipment in the past, and the historical fault information can be input into the neural network model. Then, the neural network model can perform a series of processing on the historical fault information to obtain the probability of multiple faults occurring in the equipment in the future, which is the fault prediction result of the equipment. At this point, the fault prediction for the equipment is completed.

In the above process, when the neural network model performs fault prediction on the equipment, that is, when processing the historical fault information of the equipment, it usually only considers the multiple faults that have occurred in the equipment in the past. The degree of contribution paid and the factors considered are relatively single, which will lead to the final failure prediction result of the equipment being inaccurate.

Contents of the invention

Embodiments of the present application provide a fault prediction method and related equipment. In the process of fault prediction for the equipment, relatively comprehensive factors are considered, so that the final fault prediction result of the equipment can have high accuracy.

The embodiment of the present application provides a fault prediction method, which is implemented through a target model. The method includes:

When it is necessary to perform fault prediction for the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate that the target device is in the first N faults that have occurred from time to time T-1, T≥2, N≥2.

After obtaining the historical fault information of the target equipment, the historical fault information of the target equipment can be input into the target model. The target model can process the historical fault information of the target equipment, thereby obtaining the target equipment's historical fault information from the 1st moment to the T-1th time. The causal relationship between N faults that have occurred in time and the weight of N faults. Among them, the weight of N faults is used to indicate the role played by N faults in the process of fault prediction of the target device, that is, the importance of N faults.

After obtaining the causal relationship between N faults and the weights of N faults, the target model can process the historical fault information of the target equipment, the causal relationships between N faults, and the weights of N faults, thereby obtaining the target equipment's fault information. The probability of N faults occurring at the T moment is the fault prediction result of the target device. At this point, the fault prediction for the target device is completed.

It can be seen from the above method: when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate the target. N faults that have occurred in the equipment from the 1st time to the T-1th time. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In one possible implementation method, based on historical fault information, obtaining the causal relationship between N faults and the weight of the N faults includes: performing first feature extraction on the historical fault information to obtain the causal relationship between the N faults. ; Extract sub-historical fault information from historical fault information. Sub-historical fault information is used to indicate P faults that have occurred in the target device from the T-wth moment to the T-1th moment. N faults include P faults, N ≥P≥1, T>w≥1; perform second feature extraction on the causal relationship between N faults and sub-historical fault information to obtain the weights of N faults. In the foregoing implementation, after receiving the historical fault information of the target device, the target model can perform a first feature extraction on the historical fault information of the target device, thereby obtaining the characteristics of the target device from the 1st moment to the T-1th moment. The causal relationship between N faults that have occurred. After obtaining the causal relationship between N faults, the target model can extract the sub-historical fault information of the target device from the historical fault information of the target device. The sub-historical fault information of the target device is used to indicate that the target device has arrived at the T-wth time. P faults occurred at time T-1. After obtaining the causal relationship between N faults and the sub-historical fault information of the target equipment, the target model can perform second feature extraction on the causal relationship between the N faults and the sub-historical fault information, thereby accurately obtaining the weights of the N faults. .

In one possible implementation method, based on historical fault information, the causal relationship between N faults, and the weight of the N faults, obtaining the probability of N faults occurring in the target device at the T-th moment includes: based on the N faults The causal relationship between N faults and the weight of N faults are used to obtain the causal relationship between M faults among the N faults, N≥M≥1; based on the historical fault information and the causal relationship between the M faults, the target device is obtained The probability that N faults occur at the T moment. In the aforementioned implementation method, the target model can first process the causal relationship between N faults and the weight of the N faults, thereby obtaining the causal relationship between M faults among the N faults, and the causal relationship between the M faults. The relationship can also be called the fault evidence chain of the target device, which is used to provide explanations for the fault prediction results of the target device. After obtaining the causal relationship between M faults, the target model can also process the historical fault information of the target device and the causal relationship between the M faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. , which is the fault prediction result of the target device. Then, the fault evidence chain of the target device and the fault prediction result of the target device can be used as two outputs of the target model, thereby providing users with a fault prediction service for the target device and a visual explanation of the fault prediction for the target device.

In a possible implementation manner, based on the causal relationship between N faults and the weight of the N faults, obtaining the causal relationship between M faults among the N faults includes: the causal relationship between the N faults In , N-M faults whose weights are less than the first weight threshold are eliminated, and the causal relationships between M faults among the N faults are obtained. In the aforementioned implementation, after obtaining the sub-historical fault information of the target device, the weights of N faults, and the causal relationships between the N faults, the target model can assign a weight smaller than the first in the causal relationship between the N faults. N-M faults with weight threshold are eliminated, thereby obtaining the causal relationship between M faults among the N faults. It is worth noting that at this time, the causal relationship between these M faults is accompanied by the weight of these M faults. , that is to say, the fault evidence chain of the target device not only contains the causal relationship between the M faults, but also includes the weights of the M faults, which can provide users with more detailed explanations of fault prediction.

In one possible implementation method, based on historical fault information and the causal relationship between the M faults, obtaining the probability of N faults occurring in the target device at the T-th moment includes: analyzing the causal relationships between the M faults and The third feature extraction is performed on the sub-historical fault information to obtain the probability of N faults occurring in the target equipment at the T-th moment. In the foregoing implementation, after obtaining the causal relationship between the M faults, the target model can perform third feature extraction on the causal relationship between the M faults and the sub-historical fault information of the target device, thereby accurately obtaining the target device in the first The probability of N faults occurring in T moments.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network. In the aforementioned implementation manner, the target model may include at least one of a recurrent neural network and a convolutional neural network. Therefore, the first feature extraction or the second feature extraction implemented by the target model includes feature extraction based on the recurrent neural network and convolution-based feature extraction. At least one of feature extraction by neural networks.

In a possible implementation manner, the third feature extraction includes at least one of the following: feature extraction based on a recurrent neural network, feature extraction based on a temporal convolutional network, and feature extraction based on a multi-layer perceptron. In the foregoing implementation, the target model may include at least one of a recurrent neural network, a temporal convolutional network, and a multi-layer perceptron, so the third feature extraction or the fourth feature extraction implemented by the target model includes features based on the recurrent neural network. At least one of extraction, feature extraction based on temporal convolutional network, and feature extraction based on multi-layer perceptron.

The second aspect of the embodiment of the present application provides a model training method. The method includes: obtaining historical fault information of the target device. The historical fault information is used to indicate that the target device has occurred between the 1st moment and the T-1th moment. There have been N faults, T≥2, N≥2; input the historical fault information into the model to be trained, and obtain the probability of N faults occurring in the target equipment at the T-th moment. The model to be trained is used to: Based on the historical fault information , obtain the causal relationship between N faults and the weight of N faults. The weight of N faults is used to indicate the importance of N faults; based on historical fault information, the causal relationship between N faults and the Weight, obtain the probability of N failures occurring in the target device at the T-th moment; train the model to be trained based on the probability to obtain the target model.

The target model trained by the above method has fault prediction function. Specifically, when it is necessary to perform fault prediction for the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate that the target device is in the first stage. N faults that have occurred from time to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In one possible implementation method, the model to be trained is used to: extract the first feature from historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information from the historical fault information, and sub-historical fault information. The information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N faults The second feature extraction is performed on the causal relationship between the faults and the sub-historical fault information to obtain the weights of N faults.

In a possible implementation method, the model to be trained is used to: obtain the causal relationship between M faults among the N faults based on the causal relationship between N faults and the weight of the N faults, N≥M ≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T moment.

In one possible implementation method, the model to be trained is used to: among the causal relationships between N faults, eliminate N-M faults whose weights are less than the first weight threshold, and obtain M faults among the N faults. causal relationship between.

In one possible implementation method, the model to be trained is used to: extract the third feature of the causal relationship between M faults and sub-historical fault information, and obtain the results of N faults occurring in the target equipment at the T-th moment. Probability.

In one possible implementation method, the model to be trained is also used to: among the causal relationships between M faults, eliminate M-K faults whose weights are less than the second weight threshold, and obtain K faults among the M faults. The causal relationship between K faults, M>K≥1; perform fourth feature extraction on the causal relationship between K faults and sub-historical fault information, and obtain the new probability of N faults occurring in the target equipment at the T-th moment; based on Training the model to be trained with probability to obtain the target model includes: training the model to be trained based on probability and new probability to obtain the target model.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction or the fourth feature extraction includes at least one of the following: feature extraction based on recurrent neural networks, feature extraction based on temporal convolutional networks, and feature extraction based on multi-layer perceptrons.

The third aspect of the embodiment of the present application provides a fault prediction device, which includes a target model. The device includes: a first acquisition module for acquiring historical fault information of the target equipment, and the historical fault information is used to indicate that the target equipment is in use. N faults that have occurred from the first moment to the T-1 moment, T≥2, N≥2; the second acquisition module is used to obtain the causal relationship between the N faults and N based on historical fault information. The weight of N faults is used to indicate the importance of N faults; the third acquisition module is used to obtain the target device based on historical fault information, the causal relationship between N faults and the weight of N faults. The probability that N faults occur at the T moment.

It can be seen from the above device that when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate the target. N faults that have occurred in the equipment from the 1st time to the T-1th time. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In a possible implementation manner, the second acquisition module is used to: perform first feature extraction on historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information, sub-history, from historical fault information Fault information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N The causal relationship between faults and sub-historical fault information are used for second feature extraction to obtain the weights of N faults.

In a possible implementation manner, the third acquisition module is used to: obtain the causal relationship between M faults among the N faults based on the causal relationship between the N faults and the weight of the N faults, N≥ M≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T moment.

In a possible implementation manner, the third acquisition module is used to eliminate N-M faults whose weights are less than the first weight threshold among the causal relationships between N faults, and obtain M faults among the N faults. causal relationship between.

In a possible implementation manner, the third acquisition module is used to extract the third feature of the causal relationship between the M faults and the sub-historical fault information, and obtain the N faults of the target equipment at the T-th moment. Probability.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction includes at least one of the following: feature extraction based on a recurrent neural network, feature extraction based on a temporal convolutional network, and feature extraction based on a multi-layer perceptron.

The fourth aspect of the embodiment of the present application provides a model training device. The device includes: an acquisition module for acquiring historical fault information of the target device, and the historical fault information is used to indicate that the target device is at the first moment to the T-th moment. N faults that have occurred in one moment, T ≥ 2, N ≥ 2; the processing module is used to input historical fault information into the model to be trained, and obtain the probability of N faults occurring in the target equipment at the T moment, The model to be trained is used to: based on historical fault information, obtain the causal relationship between N faults and the weights of N faults. The weight of N faults is used to indicate the importance of N faults; based on historical fault information, N faults The causal relationship between the faults and the weight of N faults are used to obtain the probability of N faults occurring in the target device at the T-th moment; the training module is used to train the to-be-trained model based on probability to obtain the target model.

The target model trained by the above device has fault prediction function. Specifically, when it is necessary to perform fault prediction for the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate that the target device is in the first stage. N faults that have occurred from time to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In one possible implementation method, the model to be trained is used to: extract the first feature from historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information from the historical fault information, and sub-historical fault information. The information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N faults The second feature extraction is performed on the causal relationship between the faults and the sub-historical fault information to obtain the weights of N faults.

In a possible implementation method, the model to be trained is used to: obtain the causal relationship between M faults among the N faults based on the causal relationship between N faults and the weight of the N faults, N≥M ≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T moment.

In one possible implementation method, the model to be trained is used to: among the causal relationships between N faults, eliminate N-M faults whose weights are less than the first weight threshold, and obtain M faults among the N faults. causal relationship between.

In one possible implementation method, the model to be trained is used to: extract the third feature of the causal relationship between M faults and sub-historical fault information, and obtain the results of N faults occurring in the target equipment at the T-th moment. Probability.

In one possible implementation method, the model to be trained is also used to: among the causal relationships between M faults, eliminate M-K faults whose weights are less than the second weight threshold, and obtain K faults among the M faults. The causal relationship between them, M>K≥1; perform fourth feature extraction on the causal relationship between K faults and sub-historical fault information, and obtain the new probability of N faults occurring in the target equipment at the T-th moment; training The module is used to train the model to be trained based on probability and new probability to obtain the target model.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction or the fourth feature extraction includes at least one of the following: feature extraction based on recurrent neural networks, feature extraction based on temporal convolutional networks, and feature extraction based on multi-layer perceptrons.

A third aspect of the embodiment of the present application provides a fault prediction device, which includes a memory and a processor; the memory stores codes, and the processor is configured to execute the code. When the code is executed, the fault prediction device performs the first step The method described in any possible implementation manner of the aspect or the first aspect.

The fourth aspect of the embodiment of the present application provides a model training device, which includes a memory and a processor; the memory stores code, and the processor is configured to execute the code. When the code is executed, the model training device executes the second step The method described in any possible implementation manner of the aspect or the second aspect.

A fifth aspect of the embodiments of the present application provides a circuit system. The circuit system includes a processing circuit. The processing circuit is configured to perform the first aspect, any one of the possible implementations of the first aspect, the second aspect, or the third aspect. The method described in any one of the possible implementation methods in the two aspects.

A sixth aspect of the embodiments of the present application provides a chip system. The chip system includes a processor for calling a computer program or computer instructions stored in a memory, so that the processor executes the steps described in the first aspect and the first aspect. Any possible implementation manner, the second aspect, or the method described in any possible implementation manner in the second aspect.

In one possible implementation, the processor is coupled to the memory through an interface.

In a possible implementation, the chip system further includes a memory, and computer programs or computer instructions are stored in the memory.

A seventh aspect of the embodiments of the present application provides a computer storage medium. The computer storage medium stores a computer program. When the program is executed by a computer, the computer implements any one of the possible methods in the first aspect and the first aspect. The method described in the implementation, the second aspect, or any possible implementation of the second aspect.

An eighth aspect of the embodiments of the present application provides a computer program product. The computer program product stores instructions. When the instructions are executed by a computer, the computer implements any one of the possible implementations of the first aspect and the first aspect. method, the second aspect, or any possible implementation manner of the second aspect.

In the embodiment of the present application, when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate the target device. N faults that have occurred from time 1 to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

Description of drawings

Figure 1 is a structural schematic diagram of the main framework of artificial intelligence;

Figure 2a is a schematic structural diagram of the fault prediction system provided by the embodiment of the present application;

Figure 2b is another structural schematic diagram of the fault prediction system provided by the embodiment of the present application;

Figure 2c is a schematic diagram of related equipment for fault prediction provided by the embodiment of the present application;

Figure 3 is a schematic diagram of the architecture of the system 100 provided by the embodiment of the present application;

Figure 4 is a schematic structural diagram of the target model provided by the embodiment of the present application;

Figure 5 is a schematic flow chart of the fault prediction method provided by the embodiment of the present application;

Figure 6 is a schematic diagram of the decision interpretation module provided by the embodiment of the present application;

Figure 7 is a schematic diagram of the decision chain optimization module provided by the embodiment of the present application;

Figure 8 is a schematic diagram of the cause-and-effect diagram and evidence chain provided by the embodiment of the present application;

Figure 9 is a schematic flow chart of the model training method provided by the embodiment of the present application;

Figure 10 is a schematic structural diagram of a fault prediction device provided by an embodiment of the present application;

Figure 11 is a schematic structural diagram of a model training device provided by an embodiment of the present application;

Figure 12 is a schematic structural diagram of an execution device provided by an embodiment of the present application;

Figure 13 is a schematic structural diagram of the training equipment provided by the embodiment of the present application;

Figure 14 is a schematic structural diagram of a chip provided by an embodiment of the present application.

Detailed ways

Embodiments of the present application provide a fault prediction method and related equipment. In the process of fault prediction for the equipment, relatively comprehensive factors are considered, so that the final fault prediction result of the equipment can have high accuracy.

The terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the terms so used are interchangeable under appropriate circumstances, and are merely a way of distinguishing objects with the same attributes in describing the embodiments of the present application. Furthermore, the terms "include" and "having" and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, product or apparatus comprising a series of elements need not be limited to those elements, but may include not explicitly other elements specifically listed or inherent to such processes, methods, products or equipment.

Equipment failure prediction refers to the fact that the equipment failure has not yet occurred. Through the neural network model in AI technology, it is predicted whether a certain or certain failures may occur in the equipment based on certain information of the equipment, so as to inform the equipment engineers of the failure prediction results, thus making Equipment engineers can repair equipment in time based on fault prediction results.

In related technologies, historical fault information of the equipment can be obtained first. The historical fault information is used to indicate multiple faults that have occurred in the equipment in the past, and the historical fault information can be input into the neural network model. Then, the neural network model can perform a series of processing on the historical fault information to obtain the probability of multiple faults occurring in the equipment in the future, which is the fault prediction result of the equipment. At this point, the fault prediction for the equipment is completed.

In the above process, the neural network model will perform complex processing on the historical fault information used to indicate multiple faults that have occurred in the equipment, thereby obtaining the contribution of these multiple faults, and then make further predictions based on the contribution of these multiple faults. Processed accordingly, the probability of multiple failures occurring in the equipment in the future can be obtained. It can be seen that in the process of fault prediction for equipment, the neural network model only considers the contribution of these multiple faults in the fault prediction process, and the factors considered are relatively single, which will lead to the failure of the final equipment. The forecast results are not accurate enough.

Furthermore, the neural network model provided by the related technology is a black box, which can only approximate the fault prediction results of the equipment through a linear model in the mathematical field. Then, the parameters of the linear model can be used as multiple faults in fault prediction. The contribution made in the process, in this way, although the relationship between the historical fault information and the fault prediction results of the equipment can be roughly explained (that is, it is predicted that one or some faults will occur in the equipment, and roughly explained Why these failures occur in equipment), but the relationship between historical failure information and equipment failure prediction results cannot be explained in detail.

In order to solve the above problem, embodiments of the present application provide a fault prediction method, which can be implemented in conjunction with artificial intelligence (artificial intelligence, AI) technology. AI technology is a technical discipline that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence. AI technology obtains the best results by perceiving the environment, acquiring knowledge and using knowledge. In other words, artificial intelligence technology is a branch of computer science that attempts to understand the nature of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Using artificial intelligence for data processing is a common application method of artificial intelligence.

First, the overall workflow of the artificial intelligence system is described. Please refer to Figure 1. Figure 1 is a structural schematic diagram of the main framework of artificial intelligence. The following is from the "intelligent information chain" (horizontal axis) and "IT value chain" (vertical axis) The above artificial intelligence theme framework is elaborated on in two dimensions. Among them, the "intelligent information chain" reflects a series of processes from data acquisition to processing. For example, it can be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, intelligent execution and output. In this process, the data has gone through the condensation process of "data-information-knowledge-wisdom". The "IT value chain" reflects the value that artificial intelligence brings to the information technology industry, from the underlying infrastructure of human intelligence and information (providing and processing technology implementation) to the systematic industrial ecological process.

(1)Infrastructure

Infrastructure provides computing power support for artificial intelligence systems, enables communication with the external world, and supports it through basic platforms. Communicate with the outside through sensors; computing power is provided by smart chips (hardware acceleration chips such as CPU, NPU, GPU, ASIC, FPGA, etc.); the basic platform includes distributed computing framework and network and other related platform guarantees and support, which can include cloud storage and Computing, interconnection networks, etc. For example, sensors communicate with the outside world to obtain data, which are provided to smart chips in the distributed computing system provided by the basic platform for calculation.

(2)Data

Data from the upper layer of the infrastructure is used to represent data sources in the field of artificial intelligence. The data involves graphics, images, voice, and text, as well as IoT data of traditional devices, including business data of existing systems and sensory data such as force, displacement, liquid level, temperature, and humidity.

(3)Data processing

Data processing usually includes data training, machine learning, deep learning, search, reasoning, decision-making and other methods.

Among them, machine learning and deep learning can perform symbolic and formal intelligent information modeling, extraction, preprocessing, training, etc. on data.

Reasoning refers to the process of simulating human intelligent reasoning in computers or intelligent systems, using formal information to perform machine thinking and problem solving based on reasoning control strategies. Typical functions are search and matching.

Decision-making refers to the process of decision-making after intelligent information is reasoned, and usually provides functions such as classification, sorting, and prediction.

(4) General ability

After the data is processed as mentioned above, some general capabilities can be formed based on the results of further data processing, such as algorithms or a general system, such as translation, text analysis, computer vision processing, speech recognition, and image processing. identification, etc.

(5) Intelligent products and industry applications

Intelligent products and industry applications refer to the products and applications of artificial intelligence systems in various fields. They are the encapsulation of overall artificial intelligence solutions, productizing intelligent information decision-making and realizing practical applications. Its application fields mainly include: intelligent terminals, intelligent transportation, Smart healthcare, autonomous driving, smart cities, etc.

Next, several application scenarios of this application will be introduced.

Figure 2a is a schematic structural diagram of a fault prediction system provided by an embodiment of the present application. The fault prediction system includes user equipment and data processing equipment. Among them, user equipment includes smart terminals such as mobile phones, personal computers, or information processing centers. The user equipment is the initiator of fault prediction. As the initiator of the fault prediction request, the user usually initiates the request through the user equipment.

The above-mentioned data processing equipment may be a cloud server, a network server, an application server, a management server, and other equipment or servers with data processing functions. The data processing equipment receives the fault prediction request from the intelligent terminal through the interactive interface, and then performs fault prediction through machine learning, deep learning, search, reasoning, decision-making and other methods through the memory for storing data and the processor for data processing. The memory in the data processing device can be a general term, including local storage and a database that stores historical data. The database can be on the data processing device or on other network servers.

In the fault prediction system shown in Figure 2a, the user device can receive the user's instructions. For example, the user device can obtain the historical fault information of the target device input/selected by the user, and then initiate a request to the data processing device, so that the data processing device can respond to the request from the user. The historical fault information of the target device of the user equipment performs fault prediction processing, thereby obtaining a fault prediction result for the target device. For example, the user equipment can obtain the historical fault information of the target device input by the user (this information is used to indicate multiple faults that have occurred in the target device in the past), and then the user device can initiate a fault prediction request to the data processing device, so that the data Based on the fault prediction request, the processing device performs a series of processing on the historical fault information of the target device to obtain the fault prediction result of the target device, that is, the probability of multiple failures of the target device in the future.

In Figure 2a, the data processing device can execute the fault prediction method according to the embodiment of the present application.

Figure 2b is another schematic structural diagram of the fault prediction system provided by the embodiment of the present application. In Figure 2b, the user equipment directly serves as a data processing equipment. The user equipment can directly obtain input from the user and directly perform processing by the hardware of the user equipment itself. Processing, the specific process is similar to Figure 2a, please refer to the above description, and will not be repeated here.

In the fault prediction system shown in Figure 2b, the user equipment can receive the user's instructions. For example, the user equipment can obtain the historical fault information of the target device input by the user (this information is used to indicate multiple faults that have occurred in the target device in the past). , and then the user device can perform a series of processing on the historical fault information of the target device to obtain the fault prediction result of the target device, that is, the probability of multiple failures of the target device in the future.

In Figure 2b, the user equipment itself can execute the fault prediction method according to the embodiment of the present application.

Figure 2c is a schematic diagram of related equipment for fault prediction provided by the embodiment of the present application.

The user equipment in Figure 2a and Figure 2b can be the local device 301 or the local device 302 in Figure 2c, and the data processing device in Figure 2a can be the execution device 210 in Figure 2c, where the data storage system 250 can To store the data to be processed by the execution device 210, the data storage system 250 can be integrated on the execution device 210, or can be set up on the cloud or other network servers.

The processors in Figure 2a and Figure 2b can perform data training/machine learning/deep learning through neural network models or other models (for example, models based on support vector machines), and use the final trained or learned model to execute on the image using the data Fault prediction application to obtain corresponding processing results.

Figure 3 is a schematic diagram of the architecture of the system 100 provided by the embodiment of the present application. In Figure 3, the execution device 110 is configured with an input/output (I/O) interface 112 for data interaction with external devices. The user Data can be input to the I/O interface 112 through the client device 140. In this embodiment of the present application, the input data may include: various to-be-scheduled tasks, callable resources, and other parameters.

When the execution device 110 preprocesses the input data, or when the calculation module 111 of the execution device 110 performs calculation and other related processing (such as implementing the function of the neural network in this application), the execution device 110 can call the data storage system 150 The data, codes, etc. in the system can be used for corresponding processing, and the data, instructions, etc. obtained by corresponding processing can also be stored in the data storage system 150 .

Finally, the I/O interface 112 returns the processing results to the client device 140, thereby providing them to the user.

It is worth mentioning that the training device 120 can generate corresponding target models/rules based on different training data for different goals or different tasks, and the corresponding target models/rules can be used to achieve the above goals or complete the above tasks. , thereby providing users with the desired results. The training data may be stored in the database 130 and come from training samples collected by the data collection device 160 .

In the case shown in FIG. 3 , the user can manually enter the input data, and the manual input can be operated through the interface provided by the I/O interface 112 . In another case, the client device 140 can automatically send input data to the I/O interface 112. If requiring the client device 140 to automatically send input data requires the user's authorization, the user can set corresponding permissions in the client device 140. The user can view the results output by the execution device 110 on the client device 140, and the specific presentation form may be display, sound, action, etc. The client device 140 can also be used as a data collection end to collect the input data of the input I/O interface 112 and the output results of the output I/O interface 112 as new sample data, and store them in the database 130 . Of course, it is also possible to collect without going through the client device 140. Instead, the I/O interface 112 directly uses the input data input to the I/O interface 112 and the output result of the output I/O interface 112 as a new sample as shown in the figure. The data is stored in database 130.

It is worth noting that Figure 3 is only a schematic diagram of a system architecture provided by an embodiment of the present application. The positional relationship between the devices, devices, modules, etc. shown in the figure does not constitute any limitation. For example, in Figure 3, the data The storage system 150 is an external memory relative to the execution device 110. In other cases, the data storage system 150 can also be placed in the execution device 110. As shown in Figure 3, the neural network can be trained according to the training device 120.

An embodiment of the present application also provides a chip, which includes a neural network processor NPU. The chip can be disposed in the execution device 110 as shown in FIG. 3 to complete the calculation work of the calculation module 111. The chip can also be installed in the training device 120 as shown in Figure 3 to complete the training work of the training device 120 and output the target model/rules.

Neural network processor NPU, NPU is mounted on the main central processing unit (CPU) (host CPU) as a co-processor, and the main CPU allocates tasks. The core part of the NPU is the arithmetic circuit. The controller controls the arithmetic circuit to extract the data in the memory (weight memory or input memory) and perform operations.

In some implementations, the computing circuit includes multiple processing units (PE). In some implementations, the arithmetic circuit is a two-dimensional systolic array. The arithmetic circuit may also be a one-dimensional systolic array or other electronic circuit capable of performing mathematical operations such as multiplication and addition. In some implementations, the arithmetic circuit is a general-purpose matrix processor.

For example, assume there is an input matrix A, a weight matrix B, and an output matrix C. The arithmetic circuit fetches the corresponding data of matrix B from the weight memory and caches it on each PE in the arithmetic circuit. The operation circuit takes matrix A data and matrix B from the input memory to perform matrix operations, and the partial result or final result of the obtained matrix is stored in the accumulator (accumulator).

The vector calculation unit can further process the output of the arithmetic circuit, such as vector multiplication, vector addition, exponential operation, logarithmic operation, size comparison, etc. For example, the vector computing unit can be used for network calculations in non-convolutional/non-FC layers in neural networks, such as pooling, batch normalization, local response normalization, etc.

In some implementations, the vector computation unit can store the processed output vector into a unified buffer. For example, the vector calculation unit may apply a nonlinear function to the output of the arithmetic circuit, such as a vector of accumulated values, to generate activation values. In some implementations, the vector computation unit generates normalized values, merged values, or both. In some implementations, the processed output vector can be used as an activation input to an arithmetic circuit, such as for use in a subsequent layer in a neural network.

Unified memory is used to store input data and output data.

The weight data directly transfers the input data in the external memory to the input memory and/or the unified memory through the storage unit access controller (direct memory access controller, DMAC), stores the weight data in the external memory into the weight memory, and stores the weight data in the unified memory. The data is stored in external memory.

The bus interface unit (BIU) is used to realize the interaction between the main CPU, DMAC and instruction memory through the bus.

The instruction fetch buffer connected to the controller is used to store instructions used by the controller;

The controller is used to call instructions cached in the memory to control the working process of the computing accelerator.

Generally, the unified memory, input memory, weight memory and instruction memory are all on-chip memories, and the external memory is a memory outside the NPU. The external memory can be a double data rate synchronous dynamic random access memory (double data rate). rate synchronous dynamic random access memory (DDRSDRAM), high bandwidth memory (high bandwidth memory, HBM) or other readable and writable memory.

Since the embodiments of the present application involve the application of a large number of neural networks, in order to facilitate understanding, the relevant terms involved in the embodiments of the present application and related concepts such as neural networks are first introduced below.

(1)Neural network

The neural network can be composed of neural units. The neural unit can refer to an arithmetic unit that takes xs and intercept 1 as input. The output of the arithmetic unit can be:

Among them, s=1, 2,...n, n is a natural number greater than 1, Ws is the weight of xs, and b is the bias of the neural unit. f is the activation function of the neural unit, which is used to introduce nonlinear characteristics into the neural network to convert the input signal in the neural unit into an output signal. The output signal of this activation function can be used as the input of the next convolutional layer. The activation function can be a sigmoid function. A neural network is a network formed by connecting many of the above-mentioned single neural units together, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field. The local receptive field can be an area composed of several neural units.

The work of each layer in the neural network can be described by the mathematical expression y=a(Wx+b): From the physical level, the work of each layer in the neural network can be understood as five pairs of input spaces (input vectors) Set) operations to complete the transformation from input space to output space (i.e., row space to column space of matrix). These five operations include: 1. Dimension raising/reducing; 2. Enlarging/reducing; 3. Rotation; 4. Translation; 5. "Bend". Among them, the operations of 1, 2, and 3 are completed by Wx, the operation of 4 is completed by +b, and the operation of 5 is implemented by a(). The reason why the word "space" is used here is because the object to be classified is not a single thing, but a class of things. Space refers to the collection of all individuals of this type of thing. Among them, W is a weight vector, and each value in the vector represents the weight value of a neuron in the neural network of this layer. This vector W determines the spatial transformation from the input space to the output space described above, that is, the weight W of each layer controls how to transform the space. The purpose of training a neural network is to finally obtain the weight matrix of all layers of the trained neural network (a weight matrix formed by the vector W of many layers). Therefore, the training process of neural network is essentially to learn how to control spatial transformation, and more specifically, to learn the weight matrix.

Because you want the output of the neural network to be as close as possible to the value you really want to predict, you can compare the predicted value of the current network with the really desired target value, and then update each layer of the neural network based on the difference between the two. weight vector (of course, there is usually an initialization process before the first update, that is, pre-configuring parameters for each layer in the neural network). For example, if the predicted value of the network is high, adjust the weight vector to make it predict lower Some, constant adjustments are made until the neural network can predict the truly desired target value. Therefore, it is necessary to define in advance "how to compare the difference between the predicted value and the target value". This is the loss function (loss function) or objective function (objective function), which is used to measure the difference between the predicted value and the target value. Important equations. Among them, taking the loss function as an example, the higher the output value (loss) of the loss function, the greater the difference. Then the training of the neural network becomes a process of reducing this loss as much as possible.

(2)Back propagation algorithm

The neural network can use the error back propagation (BP) algorithm to modify the size of the parameters in the initial neural network model during the training process, so that the reconstruction error loss of the neural network model becomes smaller and smaller. Specifically, forward propagation of the input signal until the output will produce an error loss, and the parameters in the initial neural network model are updated by backpropagating the error loss information, so that the error loss converges. The backpropagation algorithm is a backpropagation movement dominated by error loss, aiming to obtain the optimal parameters of the neural network model, such as the weight matrix.

(3) Cause-effect graph

A cause-and-effect diagram is a directed graph that describes the influence relationship between different variables. It can include multiple nodes and the edges between the nodes, where the nodes can refer to variables (for example, the faults involved in the embodiments of this application). The edges between nodes may refer to relationships between variables (for example, relationships between faults involved in the embodiments of this application).

(4)Evidence chain

The evidence chain is a one-way linked list that can be used to indicate the causal relationship between different variables (for example, the faults involved in the embodiments of this application).

The method provided by this application is described below from the training side of the neural network and the application side of the neural network.

The model training method provided by the embodiment of the present application involves the processing of data sequences, and can be specifically applied to methods such as data training, machine learning, and deep learning. The training data (for example, the target device in the model training method provided by the embodiment of the present application) historical fault information) to perform symbolic and formalized intelligent information modeling, extraction, preprocessing, training, etc., and finally obtain a trained neural network (for example, the target model in the model training method provided by the embodiment of the present application); Moreover, the fault prediction method provided by the embodiment of the present application can use the above-trained neural network to input input data (for example, the historical fault information of the target device in the fault prediction method provided by the embodiment of the present application) into the trained neural network. In the neural network, output data (for example, the fault prediction result of the target device in the fault prediction method provided by the embodiment of the present application) is obtained. It should be noted that the model training method and fault prediction method provided in the embodiments of this application are inventions based on the same concept, and can also be understood as two parts in a system, or two stages of an overall process: such as model training phase and model application phase.

The fault prediction method provided by the embodiment of the present application can be implemented through a target model. Figure 4 is a schematic structural diagram of the target model provided by the embodiment of the present application. As shown in Figure 4, the target model includes: a decision interpretation module and a decision chain optimization module. Among them, the input end of the decision explanation module serves as the input end of the entire target model, the output end of the decision interpretation module is connected to the input end of the decision chain optimization module, and the output end of the decision chain optimization module serves as the output end of the entire target model. In order to understand the workflow of the target model, the workflow of the target model is introduced below in conjunction with Figure 5. Figure 5 is a schematic flow chart of the fault prediction method provided by the embodiment of the present application. As shown in Figure 5, the method includes:

501. Obtain historical fault information of the target device. The historical fault information is used to indicate N faults that have occurred in the target device from the first moment to the T-1 moment, T≥2, N≥2.

In this embodiment, when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first. The historical fault information of the target device can include at least one fault that occurred in the target device at the first moment and the occurrence of these faults. Times, at least one fault occurred in the target equipment at the 2nd moment and the number of occurrences of these faults,..., at least one fault occurred in the target device at the T-1th moment and the number of occurrences of these faults (T is greater than or equal to A positive integer of 2), it can be understood that the faults that occur in the target equipment at different times can be completely the same, partially the same, or completely different. In this way, the historical fault information of the target device can be used to indicate N faults that have occurred in the target device from the 1st moment to the T-1th moment (N is a positive integer greater than or equal to 2. These N faults are mutually exclusive. are not the same, so these N faults can be understood as all types of faults that have occurred in the target device from the 1st moment to the T-1th moment).

For example, as shown in Figure 6 (Figure 6 is a schematic diagram of the decision interpretation module provided by the embodiment of the present application), in the historical fault information X _1:T-1 of the target device, X _1:T-1 contains the following data : The target device experienced fault v2 at the first moment, and it occurred three times. The target device experienced fault v2 and fault v3 at the second moment, and they occurred once and twice respectively. The target device did not fail at time 3,..., the target device experienced failure v2 and failure v3 at time T-2, and they occurred 2 times and 1 time respectively. The target device experienced fault v1 and fault v2 at time T-1, and they occurred once and three times respectively. It can be seen that X _1:T-1 can be used to indicate the fault v1, fault v2 and fault v3 that have occurred in the target device from the first moment to the T-1 moment.

It should be understood that in this embodiment, the faults that have occurred in the target device can be presented as alarm events that have occurred in the target device. Correspondingly, the historical fault information of the target device can be presented as a sequence of historical alarm events (usually a time sequence) of the target device. . Of course, the faults that have occurred in the target device and the historical fault information of the target device can also be presented in other ways, without limitation here.

It should also be understood that the time involved in this embodiment can be understood as a day, an hour or a moment, etc., with a certain length of time segment. For example, the first time can be understood as the first day, and the second time can be understood as Day 2,..., the T-1th moment can be understood as the T-1th day, the T-th moment can be understood as the T-th day, and so on.

502. Based on the historical fault information, obtain the causal relationship between the N faults and the weights of the N faults. The weights of the N faults are used to indicate the importance of the N faults.

After obtaining the historical fault information of the target equipment, the historical fault information of the target equipment can be input into the target model. The target model can process the historical fault information of the target equipment, thereby obtaining the target equipment's historical fault information from the 1st moment to the T-1th time. The causal relationship between N faults that have occurred at a time (the causal relationship between N faults is usually presented in the form of a cause-and-effect diagram) and the weight of the N faults. Among them, the weight of N faults is used to indicate the role played by N faults in the process of fault prediction for the target device, that is, the importance of N faults. Generally speaking, the greater the weight of the fault, the more important the fault is. , the smaller the weight of a fault, the less important the fault is.

Specifically, the target model can obtain the causal relationship between N faults and the weight of N faults in the following way:

(1) After receiving the historical fault information of the target device, the decision interpretation module of the target model can perform the first feature extraction on the historical fault information of the target device, thereby obtaining the target device's time from the 1st moment to the T-1th moment. The causal relationship between N faults that have occurred in . Further, since the decision interpretation module may include at least one of a recurrent neural network and a convolutional neural network, the first feature extraction implemented by the decision interpretation module may include feature extraction based on a recurrent neural network and a convolutional neural network-based feature extraction. At least one of feature extraction.

Still as in the above example, after X _1:T-1 is input to the decision explanation module, the decision explanation module can extract features of X _1:T-1 to obtain a cause-and-effect diagram It includes the causal relationship between fault v1 and fault v1 (the relationship points from fault v1 to fault v1, and the value is 1, indicating that fault V1 caused its own generation), the causal relationship between fault v1 and fault v2 (the relationship From fault v1 to fault v2, and the value is 0, indicating that fault V1 did not lead to the generation of fault V2),..., the causal relationship between fault v3 and fault v2 (the relationship is from fault v3 to fault v2, and takes The value is 0, indicating that fault V3 did not cause the generation of fault V2), the causal relationship between fault v3 and fault v3 (the relationship points from fault v3 to fault v3, and the value is 1, indicating that fault V3 caused its own generation) .

(2) After obtaining the causal relationship between N faults, the decision interpretation module can extract the sub-historical fault information of the target device from the historical fault information of the target device. The sub-historical fault information of the target device is used to indicate that the target device is in the P faults that have occurred from T-w time to T-1 time. It should be noted that w is a positive integer less than T and greater than or equal to 1, so the T-wth time is usually a time after the first time, and N faults include P faults. P is usually a positive integer less than or equal to N and greater than or equal to 1. Therefore, the P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment are mutually different faults, and these P faults The fault may be part or all of the N faults that have occurred in the target device from the first time to the T-1 time.

Still as the above example, in X _1:T-1 , the decision interpretation module can extract the sub-historical fault information X _Tw:T-1 of the target device. X _Tw:T-1 contains the following data: The target device is in the Fault v3 occurred at Tw times, and occurred twice,..., the target device experienced fault v2 and fault v3 at T-2 time, and occurred twice and once respectively. The target device experienced fault v1 and fault v2 at time T-1, and they occurred once and three times respectively. It can be seen that X _Tw:T-1 can be used to indicate the fault v1, fault v2 and fault v3 that have occurred in the target device from the Twth moment to the T-1th moment.

(3) After obtaining the causal relationship between N faults and the sub-historical fault information of the target equipment, the decision interpretation module can perform second feature extraction on the causal relationship between the N faults and the sub-historical fault information, thereby obtaining N The weight of the fault. After obtaining the weights of the N faults, the decision interpretation module can send the sub-historical fault information of the target equipment, the weights of the N faults, and the causal relationships between the N faults to the decision chain optimization module. Further, since the decision interpretation module may include at least one of a recurrent neural network and a convolutional neural network, the second feature extraction implemented by the decision interpretation module may include feature extraction based on a recurrent neural network and a convolutional neural network-based feature extraction. At least one of feature extraction.

Still as in the above example, we get and X _Tw:T-1 , the decision interpretation module can _{and _ _} The weight of is 0.1, and the weight of fault v3 is 0.5. Thereafter, the decision interpretation module can X _Tw:T-1 and μ _T-1 are sent to the decision chain optimization module.

503. Based on historical fault information, the causal relationship between N faults, and the weights of N faults, obtain the probability that N faults occur in the target device at the T-th moment.

After obtaining the causal relationship between N faults and the weights of N faults, the target model can process the historical fault information of the target equipment, the causal relationships between N faults, and the weights of N faults, thereby obtaining the target equipment's fault information. The probability of N faults occurring at the T moment is the fault prediction result of the target device. At this point, the fault prediction for the target device is completed.

Specifically, the target model can obtain the probability of N failures occurring in the target device at the T-th moment in the following way:

(1) The target model can first process the causal relationship between N faults and the weight of N faults, thereby obtaining the causal relationship between M faults among the N faults. It should be noted that M is a positive integer greater than or equal to 1 and less than or equal to N, so these M faults are different faults, and these M faults can be caused by the target device from the 1st moment to the Tth moment. - Some or all of the N faults that have occurred in one moment. It is worth noting that the causal relationship between these M faults can also be called the fault evidence chain of the target device. This evidence chain is used to provide a visual explanation for the fault prediction results of the target device as one of the outputs of the target model. .

(2) After obtaining the causal relationship between the M faults, the target model can also process the historical fault information of the target device and the causal relationship between the M faults, thereby obtaining the occurrence of the target device at the T-th moment. The probability of N faults is the fault prediction result of the target device. The fault prediction result of the target device is used as another output of the target model.

More specifically, the target model can obtain the causal relationship between M faults in the following way:

After obtaining the sub-historical fault information of the target device, the weights of the N faults, and the causal relationships between the N faults, the decision chain optimization module can set the weight to be less than the first weight threshold ( The size of this threshold can be set according to actual needs, and there are no restrictions here.) N-M faults are eliminated, so as to obtain the causal relationship between M faults among the N faults. It is worth noting that at this time, these M faults The causal relationship between faults is accompanied by the weight of these M faults. That is to say, the fault evidence chain of the target device not only contains the causal relationship between these M faults, but also includes the weight of these M faults. Therefore, the target The device's fault evidence chain can not only explain the most likely fault of the target device at the T-th moment (the fault is usually one of the M faults, and is located at the end of the causal relationship between the M faults ), it can also explain the other faults that caused this fault (that is, the rest of the M faults except this fault) and the contribution of the other faults in the process (that is, the weight).

Still as the above example, as shown in Figure 7 (Figure 7 is a schematic diagram of the decision chain optimization module provided by the embodiment of the present application), we get After X _Tw:T-1 and μ _T-1 , the decision chain optimization module can be found at/> , assign faults with weights less than 0.15

Eliminate, that is, eliminate fault v2. Then, we can get a new causal diagram It includes the causal relationship between fault v1 and fault v1 (the value of this relationship is 0.3, because the weight of fault v1 is superimposed), the causal relationship between fault v1 and fault v3 (the value of this relationship is 0, because the relationship is originally takes a value of 0), the causal relationship between fault v3 and fault v1 (the relationship takes a value of 0.5, because the weight of fault v3 is superimposed), the causal relationship between fault v3 and fault v3 (the

The relationship takes a value of 0.5 because the weight of fault v3 is superimposed). visible, An optimal fault evidence chain for the target device is provided, from fault v3 to fault v1, indicating that fault v3 leads to the generation of fault v1, and the contribution of fault v3 in the process is 0.5.

More specifically, the target model can obtain the probability of N failures occurring in the target device at the T-th moment in the following way:

After obtaining the causal relationship between the M faults, the decision chain optimization module can perform third feature extraction on the causal relationship between the M faults and the sub-historical fault information of the target device, thereby obtaining the target device's performance at the T-th moment. As for the probability of N faults, it can be understood that the M faults in the fault evidence chain also include the fault with the highest probability among the N faults, that is, the most likely fault of the target device at the T-th moment. Further, since the decision chain optimization module may include at least one of a recurrent neural network, a temporal convolutional network, and a multi-layer perceptron, the third feature extraction implemented by the decision interpretation module may include feature extraction based on the recurrent neural network, At least one of feature extraction based on temporal convolutional network and feature extraction based on multi-layer perceptron.

Still as in the above example, we get Finally, the decision chain optimization module can optimize X _Tw:T-1 and/> Perform feature extraction to obtain the fault prediction results of the target equipment/> It includes the probability that the target device fails v1 at the T-th time, the probability that the target device fails v2 at the T-th time, and the probability that the target device fails v3 at the T-th time.

In addition, the target model provided by the embodiment of the present application can also be compared with the neural network model provided by related technologies. The comparison results are shown in Table 1:

Table 1

Indicator one Indicator 2 Related technologies 0.68606 0.22222 Examples of this application 0.705548 0.47827

Based on Table 1, it can be seen that when tested on the same data set, compared with the models provided by related technologies, the target model provided by the embodiment of the present application has better performance in various indicators. In other words, the target model provided by the embodiment of the present application has better performance in various indicators. The provided target model has better performance and can provide users with better fault prediction services.

In order to further understand the cause-and-effect diagram and fault evidence chain involved in the embodiment of this application, the two are further introduced below in conjunction with Figure 8. As shown in Figure 8 (Figure 8 is a schematic diagram of the cause and effect diagram and the evidence chain provided by the embodiment of the present application), in the process of predicting alarm events based on the historical alarm event sequence, the target model can obtain a cause and effect diagram, which contains Causal relationships between multiple alarm events. Then, based on the different needs of users, the target model can predict the evidence chain of the alarm event on May 30, 2019, and can also predict the evidence chain of the alarm event on November 18, 2018.

In the embodiment of the present application, when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate the target device. N faults that have occurred from time 1 to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

Further, the target device can not only output the fault prediction result of the target device (i.e., the probability of N faults occurring in the target device at the T-th moment), but also output the fault evidence chain of the target device (i.e., M of the N faults) Causal relationship between faults), because the fault evidence chain of the target device contains the causal relationship between the most likely fault among the N faults and the remaining faults that caused the fault, therefore, the fault evidence chain of the target device can Fully and carefully explain the relationship between the target device's fault prediction results and the target device's historical fault information.

The above is a detailed description of the fault prediction method provided by the embodiment of the present application. The model training method provided by the embodiment of the present application will be introduced below. Figure 9 is a schematic flow chart of the model training method provided by the embodiment of the present application, as shown in Fig. As shown in 9, the method includes:

901. Obtain historical fault information of the target device. The historical fault information is used to indicate N faults that have occurred in the target device from the first moment to the T-1 moment, T≥2, N≥2.

In this embodiment, when it is necessary to train the model to be trained, a batch of training data can be obtained first. This batch of training data contains historical fault information of the target device. The historical fault information of the target device is used to indicate that the target device is at the first moment. To the N faults that have occurred at time T-1, T≥2, N≥2. It should be noted that the true probability of N failures occurring in the target device at the T-th moment is known.

902. Input the historical fault information into the model to be trained, and obtain the probability of N faults occurring in the target equipment at the T-th moment. The model to be trained is used to: based on the historical fault information, obtain the causal relationship between the N faults and N The weight of N faults, the weight of N faults is used to indicate the importance of N faults; based on historical fault information, the causal relationship between N faults and the weight of N faults, obtain the occurrence of the target device at the T-th moment The probability of N failures.

After obtaining the historical fault information of the target device, the historical fault information can be input into the model to be trained. Then, the model to be trained can obtain the causal relationship between the N faults and the weights of the N faults based on historical fault information. The weights of the N faults are used to indicate the importance of the N faults. Then, the model to be trained can obtain the (predicted) probability of N faults occurring in the target device at the T-th moment based on historical fault information, the causal relationship between the N faults, and the weights of the N faults.

In one possible implementation method, the model to be trained is used to: extract the first feature from historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information from the historical fault information, and sub-historical fault information. The information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N faults The second feature extraction is performed on the causal relationship between the faults and the sub-historical fault information to obtain the weights of N faults.

In a possible implementation method, the model to be trained is used to: obtain the causal relationship between M faults among the N faults based on the causal relationship between N faults and the weight of the N faults, N≥M ≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T moment.

In one possible implementation method, the model to be trained is used to: among the causal relationships between N faults, eliminate N-M faults whose weights are less than the first weight threshold, and obtain M faults among the N faults. causal relationship between.

In one possible implementation method, the model to be trained is used to: extract the third feature of the causal relationship between M faults and sub-historical fault information, and obtain the results of N faults occurring in the target equipment at the T-th moment. Probability.

In one possible implementation method, the model to be trained is also used to: among the causal relationships between M faults, eliminate M-K faults whose weights are less than the second weight threshold, and obtain K faults among the M faults. The fourth feature extraction is performed on the causal relationship between K faults and sub-historical fault information to obtain the new probability of N faults occurring in the target equipment at the T-th moment. It should be noted that after obtaining the causal relationship between the M faults, the model to be trained can set the weight in the causal relationship between the M faults to be less than the second weight threshold (this threshold is usually greater than the first weight threshold, the The size of the threshold can be set according to actual needs, and there are no restrictions here.) M-K faults are eliminated (K is a positive integer greater than or equal to 1 and less than M), so as to obtain the difference between K faults among M faults. Causal relationship. It is worth noting that the causal relationship between these K faults is accompanied by the weight of these K faults, which is used as the new fault evidence chain of the target device. That is, the new fault evidence chain of the target device not only includes these The causal relationship between K faults also includes the weight of these K faults. These K faults are usually part of the M faults. Then, the model to be trained can perform the fourth feature extraction on the causal relationship between K faults and the sub-historical fault information of the target device, thereby obtaining the new (predicted) probability of N faults occurring in the target device at the T-th moment, That is, the new fault prediction result of the target device.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction or the fourth feature extraction includes at least one of the following: feature extraction based on recurrent neural networks, feature extraction based on temporal convolutional networks, and feature extraction based on multi-layer perceptrons.

For the introduction of step 902, please refer to the relevant descriptions of steps 502 to 503 in the embodiment shown in FIG. 5, which will not be described again here.

903. Train the to-be-trained model based on probability to obtain the target model.

After obtaining the probability of N faults occurring in the target equipment at the T-th time, the probability of N failures occurring in the target equipment at the T-th time can be used to train the model to be trained until the model training conditions are met, thus obtaining the results shown in Figure 5. The target model in the example is shown.

Specifically, the model to be trained can be trained in the following ways:

After obtaining the probability of N failures occurring in the target equipment at the T-th time and the new probability of N failures occurring in the target equipment in the T-th time, since the true probability of N failures occurring in the target equipment in the T-th time has been It is known that the probability of N faults occurring on the target equipment at the T-th moment and the true probability of N faults occurring on the target equipment at the T-th moment can be calculated through the preset first loss function, thereby obtaining the first loss , the first loss is used to indicate the difference between the probability of N faults occurring in the target device at the T-th moment and the true probability of N faults occurring in the target device at the T-th moment, and through the preset second loss The function calculates the probability that the target device has N faults at the T-th moment and the new probability that the target device has N faults at the T-th moment, thereby obtaining the second loss. The second loss is used to indicate that the target device is in The difference between the probability of N failures occurring at the T-th time and the new probability of N failures occurring on the target device at the T-th time. Then, the first loss and the second loss can be used to construct the target loss.

After obtaining the target loss, the target loss can be used to update the parameters of the model to be trained to obtain the model to be trained with updated parameters, and use the next batch of training data to continue training the model to be trained with updated parameters until the model training conditions are met ( For example, the target loss satisfies the optimization purpose, wherein the optimization purpose of the first loss is to make the first loss as small as possible, the optimization purpose of the second loss is to make the second loss as large as possible, etc.), thereby obtaining the target model.

The target model trained in the embodiment of this application has a fault prediction function. Specifically, when it is necessary to perform fault prediction for the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate that the target device is in the first stage. N faults that have occurred from time to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

The above is a detailed description of the fault prediction method and model training method provided by the embodiment of the present application. The fault prediction device and the model training device provided by the embodiment of the present application will be introduced below. Figure 10 is a schematic structural diagram of a fault prediction device provided by an embodiment of the present application. As shown in Figure 10, the device includes:

The first acquisition module 1001 is used to obtain historical fault information of the target device. The historical fault information is used to indicate N faults that have occurred in the target device from the 1st moment to the T-1th moment, T≥2, N≥ 2;

The second acquisition module 1002 is used to obtain the causal relationship between N faults and the weights of N faults based on historical fault information. The weights of N faults are used to indicate the importance of N faults;

The third acquisition module 1003 is used to obtain the probability of N faults occurring in the target device at the T-th time based on historical fault information, the causal relationship between the N faults, and the weights of the N faults.

In the embodiment of the present application, when it is necessary to perform fault prediction on the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate the target device. N faults that have occurred from time 1 to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In a possible implementation manner, the second acquisition module 1002 is used to: perform first feature extraction on historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information from the historical fault information, and Historical fault information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N The second feature extraction is performed on the causal relationship between faults and sub-historical fault information to obtain the weights of N faults.

In a possible implementation manner, the third acquisition module 1003 is configured to: based on the causal relationship between N faults and the weight of the N faults, obtain the causal relationship between M faults among the N faults, N ≥M≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T-th moment.

In a possible implementation manner, the third acquisition module 1003 is used to eliminate N-M faults whose weights are less than the first weight threshold among the causal relationships between N faults, and obtain M faults among the N faults. causal relationship between.

In a possible implementation manner, the third acquisition module 1003 is used to extract the third feature of the causal relationship between the M faults and the sub-historical fault information to obtain N faults occurring in the target equipment at the T-th moment. The probability.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction includes at least one of the following: feature extraction based on a recurrent neural network, feature extraction based on a temporal convolutional network, and feature extraction based on a multi-layer perceptron.

Figure 11 is a schematic structural diagram of a model training device provided by an embodiment of the present application. As shown in Figure 11, the device includes:

The acquisition module 1101 is used to obtain historical fault information of the target device. The historical fault information is used to indicate N faults that have occurred in the target device from the 1st moment to the T-1th moment, T≥2, N≥2;

The processing module 1102 is used to input historical fault information into the model to be trained to obtain the probability of N faults occurring in the target device at the T-th moment. The model to be trained is used to: based on the historical fault information, obtain the probability of N faults between the N faults. Causal relationship and the weight of N faults. The weight of N faults is used to indicate the importance of N faults; based on historical fault information, the causal relationship between N faults and the weight of N faults, obtain the target device at T The probability of N faults occurring at a moment;

The training module 1103 is used to train the model to be trained based on probability to obtain the target model.

The target model trained in the embodiment of this application has a fault prediction function. Specifically, when it is necessary to perform fault prediction for the target device, the historical fault information of the target device can be obtained first and input into the target model. The historical fault information of the target device is used to indicate that the target device is in the first stage. N faults that have occurred from time to time T-1. Then, the target model can process the historical fault information of the target equipment to obtain the causal relationship between N faults and the weight of N faults. Finally, the target model can process the historical fault information of the target device, the causal relationship between N faults, and the weight of the N faults, thereby obtaining the probability of N faults occurring in the target device at the T-th moment. In the foregoing process, the target model not only considers the causal relationship between the N faults that have occurred in the target device from the first moment to the T-1 moment, but also considers the N fault prediction process for the target device. The weight of each fault (that is, the importance of N faults in the fault prediction process for the target device), the factors considered are relatively comprehensive, so that the final fault prediction result of the target device (that is, the target device at the T-th moment The probability of N faults occurring in ) can have high accuracy.

In one possible implementation method, the model to be trained is used to: extract the first feature from historical fault information to obtain the causal relationship between N faults; extract sub-historical fault information from the historical fault information, and sub-historical fault information. The information is used to indicate P faults that have occurred in the target equipment from the T-wth moment to the T-1th moment. N faults include P faults, N≥P≥1, T>w≥1; for N faults The second feature extraction is performed on the causal relationship between the faults and the sub-historical fault information to obtain the weights of N faults.

In a possible implementation method, the model to be trained is used to: obtain the causal relationship between M faults among the N faults based on the causal relationship between N faults and the weight of the N faults, N≥M ≥1; Based on historical fault information and the causal relationship between M faults, obtain the probability of N faults occurring in the target equipment at the T moment.

In one possible implementation method, the model to be trained is used to: among the causal relationships between N faults, eliminate N-M faults whose weights are less than the first weight threshold, and obtain M faults among the N faults. causal relationship between.

In one possible implementation method, the model to be trained is used to: extract the third feature of the causal relationship between M faults and sub-historical fault information, and obtain the results of N faults occurring in the target equipment at the T-th moment. Probability.

In one possible implementation method, the model to be trained is also used to: among the causal relationships between M faults, eliminate M-K faults whose weights are less than the second weight threshold, and obtain K faults among the M faults. The causal relationship between them, M>K≥1; perform fourth feature extraction on the causal relationship between K faults and sub-historical fault information, and obtain the new probability of N faults occurring in the target equipment at the T-th moment; training Module 1103 is used to train the model to be trained based on the probability and the new probability to obtain the target model.

In a possible implementation manner, the first feature extraction or the second feature extraction includes at least one of the following: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

In a possible implementation manner, the third feature extraction or the fourth feature extraction includes at least one of the following: feature extraction based on recurrent neural networks, feature extraction based on temporal convolutional networks, and feature extraction based on multi-layer perceptrons.

It should be noted that the information interaction, execution process, etc. between the modules/units of the above-mentioned device are based on the same concept as the method embodiments of the present application, and the technical effects they bring are the same as those of the method embodiments of the present application. The specific content can be Refer to the description in the method embodiments shown above in the embodiments of the present application, which will not be described again here.

The embodiment of the present application also relates to an execution device. Figure 12 is a schematic structural diagram of the execution device provided by the embodiment of the present application. As shown in Figure 12, the execution device 1200 can be embodied as a mobile phone, a tablet, a laptop, a smart wearable device, a server, etc., and is not limited here. The fault prediction device described in the corresponding embodiment of FIG. 10 may be deployed on the execution device 1200 to implement the function of fault prediction in the corresponding embodiment of FIG. 5 . Specifically, the execution device 1200 includes: a receiver 1201, a transmitter 1202, a processor 1203 and a memory 1204 (the number of processors 1203 in the execution device 1200 may be one or more, one processor is taken as an example in Figure 12) , wherein the processor 1203 may include an application processor 12031 and a communication processor 12032. In some embodiments of the present application, the receiver 1201, the transmitter 1202, the processor 1203, and the memory 1204 may be connected through a bus or other means.

Memory 1204 may include read-only memory and random access memory and provides instructions and data to processor 1203 . A portion of memory 1204 may also include non-volatile random access memory (NVRAM). The memory 1204 stores processor and operating instructions, executable modules or data structures, or a subset thereof, or an extended set thereof, where the operating instructions may include various operating instructions for implementing various operations.

The processor 1203 controls the execution of operations of the device. In specific applications, various components of the execution device are coupled together through a bus system. In addition to the data bus, the bus system may also include a power bus, a control bus, a status signal bus, etc. However, for the sake of clarity, various buses are called bus systems in the figure.

The methods disclosed in the above embodiments of the present application can be applied to the processor 1203 or implemented by the processor 1203. The processor 1203 may be an integrated circuit chip with signal processing capabilities. During the implementation process, each step of the above method can be completed by instructions in the form of hardware integrated logic circuits or software in the processor 1203 . The above-mentioned processor 1203 may be a general-purpose processor, a digital signal processor (DSP), a microprocessor or a microcontroller, and may further include an application specific integrated circuit (ASIC) or a field programmable gate. Array (field-programmable gate array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The processor 1203 can implement or execute the various methods, steps and logical block diagrams disclosed in the embodiments of this application. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly implemented by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other mature storage media in this field. The storage medium is located in the memory 1204. The processor 1203 reads the information in the memory 1204 and completes the steps of the above method in combination with its hardware.

The receiver 1201 may be configured to receive input numeric or character information and generate signal inputs related to performing relevant settings and functional controls of the device. The transmitter 1202 can be used to output numeric or character information through the first interface; the transmitter 1202 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; the transmitter 1202 can also include a display device such as a display screen .

In the embodiment of the present application, in one case, the processor 1203 is configured to obtain the fault prediction result of the target device through the target model in the corresponding embodiment of FIG. 5 .

The embodiment of the present application also relates to a training device. Figure 13 is a schematic structural diagram of the training device provided by the embodiment of the present application. As shown in Figure 13, the training device 1300 is implemented by one or more servers. The training device 1300 may vary greatly due to different configurations or performance, and may include one or more central processing units (CPUs) 1313 ( For example, one or more processors) and memory 1332, one or more storage media 1330 (eg, one or more mass storage devices) storing applications 1342 or data 1344. Among them, the memory 1332 and the storage medium 1330 may be short-term storage or persistent storage. The program stored in the storage medium 1330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the training device. Furthermore, the central processor 1313 may be configured to communicate with the storage medium 1330 and execute a series of instruction operations in the storage medium 1330 on the training device 1300 .

The training device 1300 may also include one or more power supplies 1326, one or more wired or wireless network interfaces 1350, one or more input and output interfaces 1358; or, one or more operating systems 1341, such as Windows ServerTM, Mac OS XTM , UnixTM, LinuxTM, FreeBSDTM and so on.

Specifically, the training device can execute the model training method in the corresponding embodiment of Figure 9 to obtain the target model.

Embodiments of the present application also relate to a computer storage medium. The computer-readable storage medium stores a program for performing signal processing. When the program is run on a computer, it causes the computer to perform the steps performed by the aforementioned execution device, or, The computer is caused to perform the steps performed by the aforementioned training device.

Embodiments of the present application also relate to a computer program product that stores instructions that, when executed by a computer, cause the computer to perform the steps performed by the foregoing execution device, or cause the computer to perform the steps performed by the foregoing training device. A step of.

The execution device, training device or terminal device provided by the embodiment of the present application may specifically be a chip. The chip includes: a processing unit and a communication unit. The processing unit may be, for example, a processor. The communication unit may be, for example, an input/output interface. Pins or circuits, etc. The processing unit can execute the computer execution instructions stored in the storage unit, so that the chip in the execution device executes the data processing method described in the above embodiment, or so that the chip in the training device executes the data processing method described in the above embodiment. Optionally, the storage unit is a storage unit within the chip, such as a register, cache, etc. The storage unit may also be a storage unit located outside the chip in the wireless access device, such as Read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (random access memory, RAM), etc.

Specifically, please refer to Figure 14. Figure 14 is a schematic structural diagram of a chip provided by an embodiment of the present application. The chip can be represented as a neural network processor NPU 1400. The NPU 1400 serves as a co-processor and is mounted to the Host CPU. ), the Host CPU allocates tasks. The core part of the NPU is the arithmetic circuit 1403. The arithmetic circuit 1403 is controlled by the controller 1404 to extract the matrix data in the memory and perform multiplication operations.

In some implementations, the computing circuit 1403 internally includes multiple processing units (Process Engine, PE). In some implementations, arithmetic circuit 1403 is a two-dimensional systolic array. The arithmetic circuit 1403 may also be a one-dimensional systolic array or other electronic circuit capable of performing mathematical operations such as multiplication and addition. In some implementations, arithmetic circuit 1403 is a general-purpose matrix processor.

For example, assume there is an input matrix A, a weight matrix B, and an output matrix C. The arithmetic circuit obtains the corresponding data of matrix B from the weight memory 1402 and caches it on each PE in the arithmetic circuit. The operation circuit takes matrix A data and matrix B from the input memory 1401 to perform matrix operations, and the partial result or final result of the matrix is stored in an accumulator (accumulator) 1408 .

The unified memory 1406 is used to store input data and output data. The weight data directly passes through the storage unit access controller (Direct Memory Access Controller, DMAC) 1405, and the DMAC is transferred to the weight memory 1402. Input data is also transferred to unified memory 1406 via DMAC.

BIU is the Bus Interface Unit, that is, the bus interface unit 1413, which is used for the interaction between the AXI bus and the DMAC and the Instruction Fetch Buffer (IFB) 1409.

The Bus Interface Unit 1413 (BIU for short) is used to fetch the memory 1409 to obtain instructions from the external memory, and is also used for the storage unit access controller 1405 to obtain the original data of the input matrix A or the weight matrix B from the external memory.

DMAC is mainly used to transfer the input data in the external memory DDR to the unified memory 1406 or the weight data to the weight memory 1402 or the input data to the input memory 1401 .

The vector calculation unit 1407 includes multiple arithmetic processing units, and if necessary, further processes the output of the arithmetic circuit 1403, such as vector multiplication, vector addition, exponential operation, logarithmic operation, size comparison, etc. It is mainly used for non-convolutional/fully connected layer network calculations in neural networks, such as Batch Normalization, pixel-level summation, upsampling of the predicted label plane, etc.

In some implementations, vector calculation unit 1407 can store the processed output vectors to unified memory 1406 . For example, the vector calculation unit 1407 can apply a linear function; or a nonlinear function to the output of the operation circuit 1403, such as linear interpolation on the prediction label plane extracted by the convolution layer, or a vector of accumulated values, to generate an activation value. . In some implementations, vector calculation unit 1407 generates normalized values, pixel-wise summed values, or both. In some implementations, the processed output vector can be used as an activation input to the arithmetic circuit 1403, such as for use in a subsequent layer in a neural network.

An instruction fetch buffer 1409 connected to the controller 1404 is used to store instructions used by the controller 1404;

The unified memory 1406, input memory 1401, weight memory 1402 and instruction fetch memory 1409 are all On-Chip memories. External memory is private to the NPU hardware architecture.

The processor mentioned in any of the above places can be a general central processing unit, a microprocessor, an ASIC, or one or more integrated circuits used to control the execution of the above programs.

In addition, it should be noted that the device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physically separate. The physical unit can be located in one place, or it can be distributed across multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between modules indicates that there are communication connections between them, which can be specifically implemented as one or more communication buses or signal lines.

Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by software plus necessary general hardware. Of course, it can also be implemented by dedicated hardware including dedicated integrated circuits, dedicated CPUs, dedicated memories, Special components, etc. to achieve. In general, all functions performed by computer programs can be easily implemented with corresponding hardware. Moreover, the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special-purpose circuits. circuit etc. However, for this application, software program implementation is a better implementation in most cases. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or that contributes to the existing technology. The computer software product is stored in a readable storage medium, such as a computer floppy disk. , U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., including several instructions to cause a computer device (which can be a personal computer, training device, or network device, etc.) to execute the steps described in various embodiments of this application. method.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it may be implemented in whole or in part in the form of a computer program product.

The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, the computer instructions may be transferred from a website, computer, training device, or data The center transmits to another website site, computer, training equipment or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store, or a data storage device such as a training device or a data center integrated with one or more available media. The available media may be magnetic media (eg, floppy disk, hard disk, magnetic tape), optical media (eg, DVD), or semiconductor media (eg, solid state disk (Solid State Disk, SSD)), etc.

Claims (20)

1. A method of fault prediction, the method implemented by a target model, the method comprising:

acquiring historical fault information of target equipment, wherein the historical fault information is used for indicating N faults of the target equipment in the 1 st time to the T-1 st time, T is more than or equal to 2, and N is more than or equal to 2;

based on the historical fault information, acquiring causal relationships among the N faults and weights of the N faults, wherein the weights of the N faults are used for indicating importance degrees of the N faults;

and acquiring the probability of the N faults of the target equipment in the T moment based on the historical fault information, the causal relationship among the N faults and the weights of the N faults.

2. The method of claim 1, wherein the obtaining causal relationships between the N faults and weights of the N faults based on the historical fault information comprises:

extracting first characteristics of the historical fault information to obtain causal relations among the N faults;

sub-historical fault information is extracted from the historical fault information, and is used for indicating P faults of target equipment in the T-w time to the T-1 time, wherein the N faults comprise the P faults, N is more than or equal to P is more than or equal to 1, and T is more than or equal to 1;

And carrying out second feature extraction on the causal relationship among the N faults and the sub-historical fault information to obtain weights of the N faults.

3. The method according to claim 1 or 2, wherein the obtaining the probability of the N faults occurring in the target device at the T-th moment based on the historical fault information, the causal relationship between the N faults, and the weights of the N faults comprises:

based on the causal relationship among the N faults and the weights of the N faults, acquiring the causal relationship among M faults in the N faults, wherein N is more than or equal to M is more than or equal to 1;

and acquiring the probability of the N faults of the target equipment in the T moment based on the historical fault information and the causal relation among the M faults.

4. A method according to claim 3, wherein said obtaining causal relationships between M of the N faults based on causal relationships between the N faults and weights of the N faults comprises:

and in the causal relationship among the N faults, eliminating the N-M faults with the weight smaller than a first weight threshold value, and obtaining the causal relationship among M faults in the N faults.

5. The method of claim 4, wherein the obtaining the probability that the target device has the N faults in a T-th time based on the historical fault information and causal relationships between the M faults comprises:

and carrying out third feature extraction on the causal relationship among the M faults and the sub-historical fault information to obtain the probability of the N faults of the target equipment in the T moment.

6. The method of claim 2, wherein the first feature extraction or the second feature extraction comprises at least one of: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

7. The method of claim 5, wherein the third feature extraction comprises at least one of: feature extraction based on a cyclic neural network, feature extraction based on a time convolution network and feature extraction based on a multi-layer perceptron.

8. A method of model training, the method comprising:

acquiring historical fault information of target equipment, wherein the historical fault information is used for indicating N faults of the target equipment in the 1 st time to the T-1 st time, T is more than or equal to 2, and N is more than or equal to 2;

Inputting the historical fault information into a model to be trained to obtain the probability of the N faults of the target equipment in the T moment, wherein the model to be trained is used for: based on the historical fault information, acquiring causal relationships among the N faults and weights of the N faults, wherein the weights of the N faults are used for indicating importance degrees of the N faults; acquiring the probability of the N faults of the target equipment in the T moment based on the historical fault information, the causal relationship among the N faults and the weights of the N faults;

and training the model to be trained based on the probability, so as to obtain a target model.

9. The method of claim 8, wherein the model to be trained is for:

extracting first characteristics of the historical fault information to obtain causal relations among the N faults;

sub-historical fault information is extracted from the historical fault information, and is used for indicating P faults of target equipment in the T-w time to the T-1 time, wherein the N faults comprise the P faults, N is more than or equal to P is more than or equal to 1, and T is more than or equal to 1;

And carrying out second feature extraction on the causal relationship among the N faults and the sub-historical fault information to obtain weights of the N faults.

10. The method according to claim 8 or 9, wherein the model to be trained is for:

based on the causal relationship among the N faults and the weights of the N faults, acquiring the causal relationship among M faults in the N faults, wherein N is more than or equal to M is more than or equal to 1;

and acquiring the probability of the N faults of the target equipment in the T moment based on the historical fault information and the causal relation among the M faults.

11. The method of claim 10, wherein the model to be trained is for:

and in the causal relationship among the N faults, eliminating the N-M faults with the weight smaller than a first weight threshold value, and obtaining the causal relationship among M faults in the N faults.

12. The method of claim 11, wherein the model to be trained is for:

and carrying out third feature extraction on the causal relationship among the M faults and the sub-historical fault information to obtain the probability of the N faults of the target equipment in the T moment.

13. The method of claim 12, wherein the model to be trained is further configured to:

M-K faults with the weight smaller than a second weight threshold are removed from the causal relation among the M faults, so that the causal relation among K faults in the M faults is obtained, and M is larger than K and is larger than or equal to 1;

performing fourth feature extraction on the causal relationship among the K faults and the sub-historical fault information to obtain new probabilities of the N faults of the target equipment in the T moment;

training the model to be trained based on the probability, thereby obtaining a target model comprises the following steps:

and training the model to be trained based on the probability and the new probability, thereby obtaining a target model.

14. The method of claim 9, wherein the first feature extraction or the second feature extraction comprises at least one of: feature extraction based on a recurrent neural network and feature extraction based on a convolutional neural network.

15. The method of claim 13, wherein the third feature extraction or fourth feature extraction comprises at least one of: feature extraction based on a cyclic neural network, feature extraction based on a time convolution network and feature extraction based on a multi-layer perceptron.

16. A fault prediction apparatus, the apparatus comprising a target model, the apparatus comprising:

the first acquisition module is used for acquiring historical fault information of target equipment, wherein the historical fault information is used for indicating N faults of the target equipment in the 1 st time to the T-1 st time, T is more than or equal to 2, and N is more than or equal to 2;

the second acquisition module is used for acquiring causal relations among the N faults and weights of the N faults based on the historical fault information, wherein the weights of the N faults are used for indicating importance degrees of the N faults;

and a third obtaining module, configured to obtain a probability that the N faults occur in the target device at a T-th moment based on the historical fault information, the causal relationship between the N faults, and the weights of the N faults.

17. A model training apparatus, the apparatus comprising:

the acquisition module is used for acquiring historical fault information of target equipment, wherein the historical fault information is used for indicating N faults of the target equipment from the 1 st moment to the T-1 st moment, T is more than or equal to 2, and N is more than or equal to 2;

the processing module is used for inputting the historical fault information into a model to be trained to obtain the probability of the N faults of the target equipment in the T moment, and the model to be trained is used for: based on the historical fault information, acquiring causal relationships among the N faults and weights of the N faults, wherein the weights of the N faults are used for indicating importance degrees of the N faults; acquiring the probability of the N faults of the target equipment in the T moment based on the historical fault information, the causal relationship among the N faults and the weights of the N faults;

And the training module is used for training the model to be trained based on the probability so as to obtain a target model.

18. A fault prediction device, the device comprising a memory and a processor; the memory stores code, the processor being configured to execute the code, the fault prediction device performing the method of any of claims 1 to 15 when the code is executed.

19. A computer storage medium storing one or more instructions which, when executed by one or more computers, cause the one or more computers to implement the method of any one of claims 1 to 15.

20. A computer program product, characterized in that it stores instructions that, when executed by a computer, cause the computer to implement the method of any one of claims 1 to 15.