CN116380321A - Method for Decoupling Decoupling Model of Tire-Road Contact Triaxial Force Sensor - Google Patents

Abstract

The invention provides a method for determining a decoupling model of a tire-road contact three-way force sensor, which comprises the steps of establishing an artificial neural network structure of the decoupling model of the tire-road contact three-way force sensor, wherein the artificial neural network structure comprises an input layer hidden layer and an output layer, and the activation function of the input layer and the hidden layer is ReLU; the optimized tire-road contact three-dimensional force sensor decoupling model optimization parameters obtained by adopting an Adam learning rate self-adaptive optimization algorithm to adjust the artificial neural network structure parameters based on the optimization parameters comprise at least one of the following: and the step moment estimation exponential decay rate is used for iteratively updating parameters of the decoupling model of the tire-road contact three-way force sensor based on training data by adopting a batch updating method until the MSE value of the loss function is continuously decreased to obtain training data of the decoupling model of the decoupling tire-road contact three-way force sensor in actual measurement, wherein the training data comprises three directions of the tire-road contact three-way force sensor, which respectively correspond to input load amounts and respectively correspond to respective expected output electric signals.

Description

Method for Decoupling Decoupling Model of Tire-Road Contact Triaxial Force Sensor

technical field

The invention relates to the field of decoupling technology, in particular to a method for determining a decoupling model of a tire-road contact three-way force sensor.

Background technique

Due to factors such as machining accuracy and structural design principles of the sensor of the tire-road contact three-way force measurement system, there may be inter-dimensional coupling between the sensor, that is, there is a strong coupling between the force signal and the output signal of the strain bridge. , so that the overall measurement accuracy of the sensor system is affected. To improve the measurement accuracy of the tire-road contact three-dimensional force measurement system, it is very important to reduce the coupling between dimensions.

At present, the methods for reducing inter-dimensional coupling mainly include static decoupling algorithms based on coupling error modeling, radial basis function (RBF, Radial basis function) neural network decoupling algorithms, and so on. Since the actual output of the sensor is not purely linear, although the static decoupling algorithm, that is, linear decoupling, can reduce the interference error of the sensor to a certain extent, the decoupling effect is not ideal; and the radial basis function (RBF) neural network When traversing all sample points, if there is noise in the sample point, it will reduce its generalization ability.

Therefore, how to improve the decoupling accuracy and stability, thereby significantly improving the measurement accuracy of the tire-road three-way force sensor measurement system, has become an urgent problem to be solved.

Contents of the invention

The invention provides a method for determining the decoupling model of a tire-road contact three-way force sensor, which is used to solve the problem of high dependence on the linearity of the detection device in the prior art, inaccurate decoupling results, and unstable effects, which may lead to solution problems. Defects such as gradient explosion and gradient disappearance of coupling results can be solved, and the accuracy and stability of decoupling can be improved, thereby significantly improving the measurement accuracy of the tire-road three-way force sensor measurement system.

The invention provides a tire-road contact three-way force sensor decoupling model determination method, including:

Establish the artificial neural network structure of tire-road contact three-way force sensor decoupling model, described artificial neural network structure comprises input layer, hidden layer and output layer, and the activation function of described input layer and described hidden layer is ReLU activation function ;

Adaptive optimization algorithm of Adam learning rate is used to adjust the parameters of the artificial neural network structure based on optimization parameters to obtain an optimized tire-road contact three-way force sensor decoupling model, wherein the optimization parameters include at least one of the following terms: step size, exponential decay rate for moment estimation, small constant for numerical stabilization, and descent gradient;

Using the batch update method, based on the training data, iteratively update the parameters of the tire-road contact three-way force sensor decoupling model until the value of the loss function MSE continues to decrease, and obtain the actual measurement for the tire-road contact three-way force The tire-road contact three-way force sensor decoupling model decoupled in , wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor, and the three directions correspond to The input loads respectively correspond to the respective expected output electrical signals.

According to a method for determining a tire-road contact three-way force sensor decoupling model provided by the present invention, the batch update method is used to iteratively update the parameters of the tire-road contact three-way force sensor decoupling model based on training data, Until the value of the loss function MSE continues to decrease, the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force is obtained, including:

Execute multiple parameter update processes to obtain the MSE values obtained in multiple parameter update processes respectively. When it is determined that the value of MSE continues to decrease, stop executing the parameter update process. Based on the tire- A road contact three-way force sensor decoupling model, determining the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force;

Wherein, the parameter update process includes: selecting n input loads from the training data in the three directions of the training set, and performing uniaxial loading in the three directions respectively, detecting the output electrical signal data, and according to the output electrical signal The expected output electrical signal corresponding to the data and the input load respectively, to obtain the MSE value.

According to a method for determining the tire-road contact three-way force sensor decoupling model provided by the present invention, the decoupling model of the tire-road contact three-way force sensor is determined based on the updated tire-road contact three-way force sensor decoupling model in the last executed parameter update process. The decoupling model of tire-road contact three-way force sensor decoupled in the actual measurement of tire-road contact three-way force, including:

Input the output electrical signal data of the three directions in the verification set into the decoupling model of the tire-road contact three-way force sensor updated during the last parameter update process, and obtain the updated tire-road contact force sensor data during the last parameter update process. - Input loads in three directions output by the road contact three-way force sensor decoupling model;

In the case of verifying that the input loads in the three directions output by the tire-road contact three-way force sensor decoupling model are consistent with the actual loads corresponding to the output electrical signal data in the three directions, the parameters of the last execution The tire-road contact three-way force sensor decoupling model updated during the update process is used as the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force.

According to a tire-road contact three-way force sensor decoupling model determination method provided by the present invention, the method further includes:

The output electrical signal collected by the tire-road contact three-way force measurement system is input into the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force, and the obtained The input load values in the three directions output by the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force are described.

According to a tire-road contact three-way force sensor decoupling model determination method provided by the present invention, the method further includes:

Static calibration is performed on the tire-road contact three-way force sensor to obtain a static calibration result; the static calibration result is used to characterize the calibration coefficient matrix of the tire-road contact three-way force sensor.

According to a method for determining the decoupling model of tire-road contact three-way force sensor provided by the present invention, the hidden layer is divided into 5 layers, and the number of neurons contained in each hidden layer is 40, 160, 640, 160, 40.

The present invention also provides a decoupling model determination device for tire-road contact three-way force sensor, including:

Building a module for building the artificial neural network structure of the tire-road contact three-way force sensor decoupling model, the artificial neural network structure includes an input layer, a hidden layer and an output layer, and the activation of the input layer and the hidden layer The function is the ReLU activation function;

The first acquisition module is used to adopt the Adam learning rate adaptive optimization algorithm, adjust the parameters of the artificial neural network structure based on the optimization parameters, and obtain the optimized tire-road contact three-way force sensor decoupling model, wherein the The optimization parameters include at least one of the following: step size, exponential decay rate of moment estimation, small constant for numerical stability, and descent gradient;

The second acquisition module is used to use the batch update method to iteratively update the parameters of the decoupling model of the tire-road contact three-way force sensor based on the training data until the value of the loss function MSE continues to decrease, so as to obtain the parameters for the tire-road contact The decoupled tire-road contact three-way force sensor decoupling model in the actual measurement of the contact three-way force, wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor respectively, so The input loads corresponding to the three directions respectively correspond to the respective expected output electrical signals.

The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, any one of the above-mentioned Decoupling model determination method of tire-road contact triaxial force sensor.

The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the decoupling model of any one of the tire-road contact three-way force sensors described above can be determined. method.

The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, any method for determining the decoupling model of the tire-road contact three-way force sensor described above can be implemented.

The method for determining the tire-road contact three-way force sensor decoupling model provided by the present invention uses the ReLU function as the activation function of the artificial neural network structure of the tire-road contact three-way force sensor decoupling model to obtain the tire-road contact three-way force The sensor decoupling model can make the tire-road contact three-way force sensor decoupling model have a strong generalization ability, making the model closer to the optimal solution; using the Adam learning rate adaptive optimization algorithm to adjust the artificial neural network For the parameters of the structure, the learning rate can be dynamically adjusted by calculating and correcting the first-order moment and second-order moment of each round of gradient; the batch update method is used to iteratively update the parameters of the tire-road contact three-way force sensor decoupling model until the loss function MSE The value of is continuously decreasing, and the optimal solution of the parameters of the tire-road contact three-way force sensor decoupling model can be obtained. Using the tire-road contact three-way force sensor decoupling model for decoupling can use the forward transmission of signals in the neural network , The characteristics of error backward transmission, by continuously adjusting the network weight value, the final output of the network is as close as possible to the expected output, and the inter-dimensional coupling is minimized, thereby significantly improving the measurement of the tire-road contact three-dimensional force measurement system precision.

Description of drawings

In order to more clearly illustrate the present invention or the technical solutions in the prior art, the accompanying drawings that need to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings in the following description are the present invention. For some embodiments of the invention, those skilled in the art can also obtain other drawings based on these drawings without creative effort.

Fig. 1 is a schematic flow chart of a tire-road contact three-way force sensor decoupling model determination method provided by the present invention;

Fig. 2 is a schematic diagram of the artificial neural network algorithm model of the tire-road contact three-way force sensor decoupling model provided by the present invention;

Fig. 3 is the image contrast chart of ReLU function and Sigmoid function provided by the present invention;

Fig. 4 is a schematic diagram of the error iteration curve of the decoupling method provided by the present invention;

Fig. 5 is a schematic diagram of the artificial neural network structure of the decoupling model of tire-road contact three-way force sensor provided by the present invention;

Fig. 6 is a schematic structural diagram of a decoupling model determination device for a tire-road contact three-way force sensor provided by the present invention;

Fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention , but not all examples. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

The tire-road contact three-dimensional force measurement system is the basic testing equipment for research on the contact stress between automobile tires and road surfaces. As a new multi-dimensional force sensor system, it can be used to monitor and analyze the stress on the road surface, and has high measurement accuracy. However, due to the influence of factors such as the machining accuracy of the sensor and the structural design principle, the inter-dimensional coupling phenomenon of the sensor may occur, that is, there is a strong coupling between the force signal and the output signal of the strain bridge, which affects the overall measurement accuracy of the sensor system. Influence.

To sum up, in order to improve the measurement accuracy of the tire-road contact three-dimensional force measurement system, it is very important to reduce the coupling between dimensions. The present invention uses a BP neural network based on the approximate biological activation function ReLU function to measure the tire-road contact three-dimensional force Decoupling, by adjusting the weight value of the network parameters, can minimize the inter-dimensional coupling, thereby significantly improving the measurement accuracy of the tire-road contact three-dimensional force measurement system.

Fig. 1 is a schematic flow chart of a tire-road contact three-way force sensor decoupling model determination method provided by the present invention, including the following steps:

Step 100, establishing the artificial neural network structure of the tire-road contact three-way force sensor decoupling model, the artificial neural network structure includes an input layer, a hidden layer and an output layer, and the activation functions of the input layer and the hidden layer are: ReLU activation function;

Optionally, the tire-road contact three-dimensional force sensor decoupling model can be used to reduce the inter-dimensional coupling phenomenon that may occur during the measurement process of the tire-road contact three-dimensional force measurement system, and improve the performance of the tire-road contact three-dimensional force measurement system. measurement accuracy.

Optionally, the hidden layer of the artificial neural network structure of the decoupling model of the tire-road contact three-way force sensor may have multiple layers.

Optionally, the ReLU activation function can be

Optionally, for the input vector x from the previous layer of neural network entering the neuron, the neuron using the ReLU function can output max(0,W ^T X+b) to the next layer of neurons or as the input vector of the entire neural network output.

Optionally, the output of the input layer or the hidden layer can be used as the input of the next layer of neural network.

Optionally, the output of the output layer can be the output of the entire neural network.

Optionally, the activation function of the input layer and the hidden layer may be a ReLU activation function in order to introduce nonlinear factors into the neurons, so that the neural network can arbitrarily approximate any nonlinear function.

Fig. 2 is a schematic diagram of the artificial neural network algorithm model of the tire-road contact three-way force sensor decoupling model provided by the present invention. As shown in Fig. 2, W1 and W2 are respectively the input layer to the hidden layer and the hidden layer to the output layer The weight matrix of , b1 and b2 are the bias matrix of the hidden layer and the bias matrix of the output layer respectively. The signal input to the artificial neural network structure of the tire-road contact three-way force sensor decoupling model is an electrical signal, and the signal output by the artificial neural network structure of the tire-road contact three-way force sensor decoupling model is a force signal.

Fig. 3 is an image comparison diagram of the ReLU function and the Sigmoid function provided by the present invention. As shown in Fig. 3, the general activation function Sigmoid function of the BP neural network has the problems of large amount of calculation in the derivation process and high time complexity of model training. The sparsity of the ReLU function can well solve the problem of gradient disappearance caused by the Sigmoid function, so the activation function of the BP neural network in the related art can be changed from the Sigmoid function to the ReLU function.

Step 110, using the Adam learning rate adaptive optimization algorithm, based on the optimization parameters, adjusting the parameters of the artificial neural network structure, and obtaining the optimized tire-road contact three-way force sensor decoupling model, wherein the optimization parameters include at least one of the following: step size, exponential decay rate for moment estimates, small constant for numerical stabilization, and descent gradient;

Optionally, Adam is a stochastic gradient descent algorithm for adaptively adjusting the learning rate, which can dynamically adjust the learning rate by calculating and correcting the first-order moment and second-order moment of each round of gradients by combining the idea of moment estimation.

Optionally, the surveyor can set the initial parameter of the artificial neural network structure to θ, and then use the Adam learning rate adaptive optimization algorithm to calculate the parameter change value Δθ based on the optimized parameters, and then adjust the parameter value Δθ based on the parameter change value Δθ. The parameters of the artificial neural network structure are θ ₁ =θ+Δθ.

Optionally, the measurement personnel can use the Adam learning rate adaptive optimization algorithm to adjust the parameters of the artificial neural network structure based on any of the following situations:

1. Step length;

2. Exponential decay rate of moment estimation;

3. Small constants for numerical stability;

4. Descent gradient;

5. Exponential decay rate of step size and moment estimation;

6. Step size and small constants for numerical stability;

7. Step size and descent gradient;

8. Exponential decay rate for moment estimation and small constants for numerical stabilization;

9. Exponential decay rate and descent gradient of moment estimation;

10. Small constants and descending gradients for numerical stability;

11. Step size, exponential decay rate for moment estimation and small constants for numerical stabilization;

12. Step size, exponential decay rate and descent gradient of moment estimation;

13. Step size, small constant for numerical stabilization and descent gradient;

14. Exponential decay rate for moment estimation, small constants and descent gradients for numerical stabilization;

15. Step size, exponential decay rate for moment estimates, small constants for numerical stabilization and descent gradients.

In one of the embodiments of the present invention, the process of parameter optimization is as follows:

The parameter optimizer adopts Adam, and first calculates the estimators m _t and u _t of the first-order moment and second-order moment of the gradient g _t in the t-th iteration:

m _t =β ₁ m _t-1 +(1-β ₁ )g _t ;

Wherein, β ₁ , β ₂ ∈[0,1) are attenuation functions, which can be set as β ₁ =0.9, β ₂ =0.999, and both m ₀ and u ₀ are initialized as d-dimensional zero vectors.

Correct the deviations of m _t and u _t respectively:

Adam's parameter update formula is:

Where ε is a small constant (generally 10 ^-8 ) used for numerical stability, and its function is to prevent the denominator from being 0.

Step 120, using the batch update method, based on the training data, iteratively updating the parameters of the decoupling model of the tire-road contact three-way force sensor until the value of the loss function MSE continues to decrease, and obtaining the tire-road contact three-way force sensor The decoupled tire-road contact three-way force sensor decoupling model in the actual measurement of , wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor, and the three directions The corresponding input loads respectively correspond to the respective expected output electrical signals.

Optionally, the batch update method may be to divide the entire training data into several individual training samples, and then import them one by one into the tire-road contact three-way force sensor decoupling model, and iteratively update the tire-road contact three-way force sensor decoupling model. The parameters of the model.

Optionally, in the tire-road contact three-way force sensor decoupling model, the force and the electrical signal can be in a linear relationship, and by measuring the output electrical signals corresponding to the three directions, the output force corresponding to the three directions can be obtained .

Optionally, the measurement personnel can import the training data into the tire-road contact three-way force sensor decoupling model, detect the output electrical signal data, and calculate the MSE value according to the actual output electrical signal and the expected output electrical signal. If the algorithm iterates, the MSE The value of the loss function MSE does not continue to decrease, so the measurement personnel need to update the parameters of the decoupling model of the tire-road contact three-way force sensor until the value of the loss function MSE continues to decrease.

Optionally, after establishing the artificial neural network structure of the tire-road contact three-way force sensor decoupling model, based on the BP neural network structure of the ReLU activation function, the tester can construct a data training set, a verification set and a test set, and then use The batch update method, based on the training set data, iteratively updates the parameters of the tire-road contact three-way force sensor decoupling model until the value of the loss function MSE continues to decrease, and obtains the actual measurement for the tire-road contact three-way force Decoupled tire-road contact three-way force sensor decoupling model.

Optionally, the three directions of the tire-road contacting three-way force sensor may be X-axis, Y-axis and Z-axis directions.

Optionally, the surveyor can select n input loads in each of the three directions of the X-axis, Y-axis and Z-axis in the data training set, respectively import the tire-road contact three-way force sensor decoupling model, and detect the output electrical signal data , according to the actual output electrical signal and the expected output electrical signal, the MSE value is calculated. If the MSE value does not continue to decrease during the algorithm iteration, the measurement personnel need to update the parameters of the decoupling model of the tire-road contact three-way force sensor until the loss function The value of MSE keeps decreasing, and the tire-road contact three-way force sensor decoupling model is obtained for decoupling in the actual measurement of the tire-road contact three-way force.

The method for determining the tire-road contact three-way force sensor decoupling model provided by the present invention uses the ReLU function as the activation function of the artificial neural network structure of the tire-road contact three-way force sensor decoupling model to obtain the tire-road contact three-way force The sensor decoupling model can make the tire-road contact three-way force sensor decoupling model have a strong generalization ability, making the model closer to the optimal solution; using the Adam learning rate adaptive optimization algorithm to adjust the artificial neural network For the parameters of the structure, the learning rate can be dynamically adjusted by calculating and correcting the first-order moment and second-order moment of each round of gradient; the batch update method is used to iteratively update the parameters of the tire-road contact three-way force sensor decoupling model until the loss function MSE The value of is continuously decreasing, and the optimal solution of the parameters of the tire-road contact three-way force sensor decoupling model can be obtained. Using the tire-road contact three-way force sensor decoupling model for decoupling can use the forward transmission of signals in the neural network , The characteristics of error backward transmission, by continuously adjusting the network weight value, the final output of the network is as close as possible to the expected output, and the inter-dimensional coupling is minimized, thereby significantly improving the measurement of the tire-road contact three-dimensional force measurement system precision.

Optionally, the batch update method is used to iteratively update the parameters of the decoupling model of the tire-road contact three-way force sensor based on the training data until the value of the loss function MSE continues to decrease, and the tire-road contact is obtained. Decoupled tire-road contact three-dimensional force sensor decoupling model in the actual measurement of three-dimensional force, including:

Execute multiple parameter update processes to obtain the MSE values obtained in multiple parameter update processes respectively. When it is determined that the value of MSE continues to decrease, stop executing the parameter update process. Based on the tire- A road contact three-way force sensor decoupling model, determining the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force;

Wherein, the parameter update process includes: selecting n input loads from the training data in the three directions of the training set, and performing uniaxial loading in the three directions respectively, detecting the output electrical signal data, and according to the output electrical signal The expected output electrical signal corresponding to the data and the input load respectively, to obtain the MSE value.

Optionally, the training set can be used as a data sample for model fitting to train parameters in the tire-road contact three-way force sensor decoupling model.

Optionally, the larger the samples in the training set, the better the decoupling effect of the trained tire-road contact three-way force sensor decoupling model.

Optionally, the process of training the parameters in the tire-road contact triaxial force sensor decoupling model can be supervised.

获得，式中n为输入载荷量的数量，y为输入载荷量分别对应的期望输出电信号，y_i′为实际输出电信号。Optionally, the value of the loss function MSE can be obtained by the formula

Obtained, where n is the number of input loads, y is the expected output electrical signal corresponding to the input loads, and y _i ' is the actual output electrical signal.

Optionally, after the surveyor obtains the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force, the initial weight and threshold of the BP neural network can be obtained.

Optionally, after performing multiple parameter update processes and obtaining the MSE values obtained in multiple parameter update processes, the measurement personnel can observe whether the value of MSE will decrease with the number of iterations. When the value of MSE is infinitely close to 0 or If the value of MSE does not change significantly with the number of iterations, it is determined that the parameters of the obtained decoupling model are the optimal solution, and the execution of the parameter update process is stopped. The force sensor decoupling model is to determine the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force.

Optionally, the measurement personnel can perform 10 parameter update processes, and the MSE values obtained in the 10 parameter update processes are 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, then the measurement personnel It can be determined that the value of MSE will decrease sequentially with the number of iterations, and the execution of the parameter update process will be stopped. Based on the tire-road contact three-way force sensor decoupling model updated during the last execution of the parameter update process, the tire-road contact force sensor decoupling model used to determine the Decoupled tire-road contact three-way force sensor decoupling model in actual measurement of road contact three-way force. Optionally, after performing multiple parameter update processes and obtaining the MSE values obtained in multiple parameter update processes, the measurement personnel can observe whether the value of MSE will decrease sequentially with the number of iterations, if the value of MSE shows an increasing trend or Non-linear change trend, the surveyors can adjust the parameters in the decoupling model algorithm, and re-execute the parameter update process for several times until the value of MSE maintains a stable decreasing trend.

Fig. 4 is a schematic diagram of the error iteration curve of the decoupling method provided by the present invention. As shown in Fig. 4, as the number of algorithm iterations increases, the value of MSE shows a decreasing trend, and the parameters of the obtained decoupling model are determined to be the optimal solution .

The method for determining the tire-road contact three-way force sensor decoupling model provided by the present invention calculates the MSE value updated each iteration during the iterative update of the parameters of the tire-road contact three-way force sensor decoupling model, and Observe whether the MSE value continues to decrease as the number of iterations increases. If it is determined that the MSE value continues to decrease, it can be considered that the tire-road contact three-way force sensor decoupling model updated during the last parameter update process is the optimal model.

Optionally, the tire-road contact three-way force sensor decoupling model updated in the last executed parameter update process is used to determine the tire-road contact three-way force sensor decoupling used in the actual measurement. - Road contact three-way force sensor decoupling model, including:

Input the output electrical signal data of the three directions in the verification set into the decoupling model of the tire-road contact three-way force sensor updated during the last parameter update process, and obtain the updated tire-road contact force sensor data during the last parameter update process. - Input loads in three directions output by the road contact three-way force sensor decoupling model;

In the case of verifying that the input loads in the three directions output by the tire-road contact three-way force sensor decoupling model are consistent with the actual loads corresponding to the output electrical signal data in the three directions, the parameters of the last execution The tire-road contact three-way force sensor decoupling model updated during the update process is used as the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force.

Optionally, in order to make the decoupling effect of the decoupling model of the tire-road contact three-way force sensor better, the measurement personnel can input the output electrical signal data of the three directions in the verification set into the model, and continuously update the tire-road contact force sensor decoupling effect. The parameters of the road contact three-way force sensor decoupling model, therefore, after the last parameter update process, the tire-road contact three-way force sensor decoupling model should have the best decoupling effect and can be used in the actual tire-road contact Decoupled decoupling model in three-way force measurement. In order to verify whether the model is optimal, the measurement personnel can input the output electrical signal data in the three directions in the verification set into the model, and verify the input in the three directions of the output. Whether the load is equal to the actual load corresponding to the output electrical signal data, if it is equal, it proves that the model is optimal and can be used for the decoupling of the tire-road contact three-way force sensor decoupling in the actual tire-road contact three-way force measurement model; if it is not equal, the measurement personnel need to continue to perform the parameter update process of the decoupling model until the optimal tire-road contact three-dimensional force that can be used for decoupling in the actual tire-road contact three-dimensional force measurement is obtained Sensor decoupling model.

Optionally, the measurement personnel can input the output electrical signal data V _x , V _y and V _z of the three directions in the verification set to the tire-road contact three-way force sensor decoupling updated during the last parameter update process In the model, the input loads F _x , F _y and F _z in the three directions output by the tire-road contact three-way force sensor decoupling model updated during the last execution of the parameter update process are obtained. If the output power in the three directions In the actual load corresponding to the signal data, the actual load corresponding to the output electrical signal data in the X-axis direction is equal to F _x ; the actual load corresponding to the output electrical signal data in the Y-axis direction is equal to F _y ; the output electrical signal data in the Z-axis direction is corresponding to If the actual load is equal to F _z , the measurer can use the tire-road contact three-way force sensor decoupling model updated during the last parameter update process as the decoupling model used in the actual measurement of the tire-road contact three-way force Decoupling model of three-way force sensor for tire-road contact.

The tire-road contact three-way force sensor decoupling model determination method provided by the present invention, after obtaining the tire-road contact three-way force sensor decoupling model updated in the last execution of the parameter update process, will also verify the concentrated three-way The output electrical signal data in three directions is input into the optimized decoupling model, and the input load in three directions is measured. By comparing whether the input load is consistent with the actual load, the model can be verified and the optimal model can be determined.

Optionally, the tire-road contact three-way force sensor decoupling model determination method further includes:

The output electrical signal collected by the tire-road contact three-way force measurement system is input into the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force, and the obtained The input load values in the three directions output by the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force are described.

Optionally, after the surveyor obtains the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force, the actual measurement data, such as the tire-road contact three-way The output electrical signal collected by the force measurement system is input into the decoupling model to check whether the model is running normally. If the input load values in three directions can be obtained, it means that the model is normal.

Optionally, the measurement personnel may base the input load values in the three directions output by the decoupling tire-road contact three-way force sensor decoupling model used in the actual measurement of the tire-road contact three-way force, to the model Validate and determine the optimal model.

Optionally, the surveyor can base on the input load values in the three directions output by the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force, from I Compare the type error, type II error and overall error with other commonly used decoupling methods, observe whether the type I error, type II error and overall error are smaller than other commonly used decoupling methods, and prove the feasibility and advantages of the ReLU function as an activation function sex.

The tire-road contact three-way force sensor decoupling model determination method provided by the present invention obtains the output of the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force The input load values in the three directions can verify the model and determine the optimal model, which proves the feasibility and superiority of the ReLU function as an activation function.

Optionally, the tire-road contact three-way force sensor decoupling model determination method further includes:

Static calibration is performed on the tire-road contact three-way force sensor to obtain a static calibration result; the static calibration result is used to characterize the calibration coefficient matrix of the tire-road contact three-way force sensor.

Optionally, the calibration coefficient matrix of the tire-road contact three-way force sensor can be used to determine the relationship between the output electrical signal and the force of the tire-road contact three-way force sensor. The output force is calculated according to the output electrical signal during the measurement.

Optionally, before the static calibration of the tire-road contact three-way force sensor, the measuring personnel can make a reliable connection between the loading system, the sensor and the measuring system.

Optionally, in order to verify the stability of the loading system and avoid various problems caused by the instability of the loading system during the static calibration of the tire-road contact three-dimensional force sensor, the measurement personnel can conduct pre-experiments, such as first Load 10% of the load and observe whether there is any abnormality in the loading process. If there is no abnormality, it can be considered that the loading system is stable, and the measurement personnel can start to statically calibrate the tire-road contact three-way force sensor.

Optionally, when performing static calibration on the tire-road contact three-way force sensor, in the X-axis, Y-axis and Z-axis directions, the load can be increased from 0 to full according to the measurement point within the range of the sensor. Range value (every 5N is a measuring point), then unload to 0; the loading operation needs to be carried out slowly, softly and steadily; repeat this process 10 times, or 15 times, or 20 times, the present invention does not make a limit to this, accurately record The electrical signal data generated by the load.

Optionally, after completing the static calibration of the tire-road contact three-way force sensor and obtaining the static calibration results, the measurement personnel can establish an artificial neural network structure for the decoupling model of the tire-road contact three-way force sensor.

In one of the embodiments of the present invention, the static calibration includes the following steps:

(1) Put the tire-road contact three-way force sensor on the loading platform of the automatic deadweight force standard machine, and check whether the loading device, sensor, control system and other parts are connected correctly and whether the connection is tight;

(2) After confirming the accurate and reliable connection of each part, conduct a pre-test and load according to 10% of the full scale value to verify the stability of the loading system;

(3) After the pre-experiment is completed, the formal loading is carried out. Taking the Z-axis direction as an example, within the range of the sensor, increase the load from 0 to the full-scale value according to the measurement point (every 5N is a measurement point), and then unload it to 0. This process is repeated 10 times, and the operation needs to be slow and steady during the loading process;

(4) The X-axis and Y-axis are tested according to step (3), and finally the static calibration of the three directions of the tire-road contact three-way force sensor is completed.

The method for determining the decoupling model of the tire-road contact three-way force sensor provided by the present invention can find the calibration coefficient matrix of the sensor through static calibration of the tire-road contact three-way force sensor, so that in actual measurement, it can be based on the output voltage The signal is calculated to prepare for the establishment and optimization of the subsequent tire-road contact three-way force sensor decoupling model.

Optionally, the hidden layer is divided into 5 layers, and the number of neurons included in each hidden layer is 40, 160, 640, 160, and 40 in sequence.

Fig. 5 is a schematic diagram of the artificial neural network structure of the tire-road contact three-way force sensor decoupling model provided by the present invention. As shown in Fig. 5, the ReLU-BP neural network structure adopts 5 hidden layers, and the number of hidden layer nodes is 40 respectively. , 160, 640, 160, 40. The input is an electrical signal, and the output is a force signal.

The tire-road contact three-way force sensor decoupling model determination method provided by the present invention adopts 5 hidden layers, the number of hidden layer nodes is 40, 160, 640, 160, 40, and the activation function of each layer is ReLU The activation function has strong generalization ability, and the more adaptable network structure and activation function make the model closer to the optimal solution.

In one embodiment of the present invention, in order to verify the feasibility of the three-way force decoupling method of the BP neural network based on the ReLU function, 150 sets of output electrical signal data were selected from each of the three directions for inspection. In order to verify the superiority of the ReLU activation function selected in the present invention, the type I error, type II error, and overall error were compared with other commonly used decoupling methods, and it has a good decoupling effect.

Table 1 Uniaxial loading error

Table 2 overall error

From the experimental results in Table 1 and Table 2 above, it can be seen that the tire road contact three-dimensional force decoupling algorithm based on the ReLU activation function of the BP neural network has both Type I and Type II errors compared with the least squares decoupling algorithm. It is significantly improved, and because the ReLU function is simple and efficient, it will show a faster decoupling speed, which further proves the feasibility and advantages of the ReLU function as an activation function.

The tire-road contact three-way force sensor decoupling model determination device provided by the present invention is described below. The tire-road contact three-way force sensor decoupling model determination device described below is the same as the tire-road contact three-way force sensor described above. Decoupling model determination methods can be referred to each other.

Fig. 6 is a schematic structural diagram of a tire-road contact three-way force sensor decoupling model determination device provided by the present invention, including an establishment module 610, a first acquisition module 620 and a second acquisition module 630, wherein:

Establishment module 610, used to establish the artificial neural network structure of tire-road contact three-way force sensor decoupling model, the artificial neural network structure includes an input layer, a hidden layer and an output layer, the input layer and the hidden layer The activation function is the ReLU activation function;

The first acquisition module 620 is configured to adopt the Adam learning rate self-adaptive optimization algorithm, adjust the parameters of the artificial neural network structure based on the optimization parameters, and obtain the optimized tire-road contact three-way force sensor decoupling model, wherein, The optimization parameters include at least one of the following: step size, exponential decay rate of moment estimation, small constant for numerical stabilization, and descent gradient;

The second acquisition module 630 is configured to use a batch update method to iteratively update the parameters of the decoupling model of the tire-road contact three-way force sensor based on the training data until the value of the loss function MSE continues to decrease, and obtain the tire-road contact force sensor used for tire-road contact. The tire-road contact three-way force sensor decoupling model decoupled in the actual measurement of the road contact three-way force, wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor respectively, The input load amounts corresponding to the three directions respectively correspond to respective expected output electrical signals.

The tire-road contact three-way force sensor decoupling model determination device provided by the present invention uses the ReLU function as the activation function of the artificial neural network structure of the tire-road contact three-way force sensor decoupling model to establish a module, which has strong generalization Ability, and a more adaptable network structure and activation function, make the model closer to the optimal solution; the first acquisition module uses the Adam learning rate adaptive optimization algorithm to adjust the parameters of the artificial neural network structure, which can be calculated and corrected for each The first-order moment and second-order moment of the wheel gradient are used to dynamically adjust the learning rate; the second acquisition module uses the batch update method to iteratively update the parameters of the tire-road contact three-way force sensor decoupling model until the value of the loss function MSE continues to decrease, which can Obtain the optimal solution of the parameters of the tire-road contact three-way force sensor decoupling model, and use the tire-road contact three-way force sensor decoupling model for decoupling, which can minimize the coupling between dimensions, thereby significantly improving the tire-road contact Measurement accuracy of a three-way force measuring system.

It can be understood that the device for determining the tire-road contact three-way force sensor decoupling model provided by the present invention corresponds to the method for determining the tire-road contact three-way force sensor decoupling model provided in the above-mentioned embodiments. For the relevant technical features of the device for determining the decoupling model of the road contact three-way force sensor, please refer to the relevant technical features of the method for determining the decoupling model of the tire-road contact three-way force sensor provided by the above-mentioned embodiments, and details will not be repeated here.

FIG. 7 illustrates a schematic diagram of the physical structure of an electronic device. As shown in FIG. 7, the electronic device may include: a processor (processor) 710, a communication interface (Communications Interface) 720, a memory (memory) 730, and a communication bus 740, Wherein, the processor 710 , the communication interface 720 , and the memory 730 communicate with each other through the communication bus 740 . The processor 710 can call the logic instructions in the memory 730 to execute the method for determining the tire-road contact three-way force sensor decoupling model. The method includes: establishing an artificial neural network structure for the tire-road contact three-way force sensor decoupling model, The artificial neural network structure includes an input layer, a hidden layer and an output layer, and the activation function of the input layer and the hidden layer is a ReLU activation function; the adaptive optimization algorithm of Adam learning rate is adopted, and based on the optimized parameters, the described The parameters of the artificial neural network structure are used to obtain an optimized tire-road contact three-way force sensor decoupling model, wherein the optimized parameters include at least one of the following: step size, exponential decay rate of moment estimation, and value for numerical stability. Small constants and descending gradients; using a batch update method to iteratively update the parameters of the tire-road contact three-way force sensor decoupling model based on the training data until the value of the loss function MSE continues to decrease, and the tire-road The decoupled tire-road contact three-way force sensor decoupling model in the actual measurement of the contact three-way force, wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor respectively, so The input loads corresponding to the three directions respectively correspond to the respective expected output electrical signals.

In addition, the above-mentioned logic instructions in the memory 730 may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes. .

On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can The method for determining the tire-road contact three-way force sensor decoupling model provided by performing the above-mentioned methods includes: establishing an artificial neural network structure for the tire-road contact three-way force sensor decoupling model, and the artificial neural network structure includes Input layer, hidden layer and output layer, the activation function of described input layer and described hidden layer is ReLU activation function; Adopt the adaptive optimization algorithm of Adam learning rate, based on optimization parameter, adjust the parameter of described artificial neural network structure, An optimized tire-road contact three-way force sensor decoupling model is obtained, wherein the optimized parameters include at least one of the following: step size, exponential decay rate of moment estimation, small constant for numerical stability, and descent gradient; Using the batch update method, based on the training data, iteratively update the parameters of the tire-road contact three-way force sensor decoupling model until the value of the loss function MSE continues to decrease, and obtain the actual measurement for the tire-road contact three-way force The tire-road contact three-way force sensor decoupling model decoupled in , wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor, and the three directions correspond to The input loads respectively correspond to respective expected output electrical signals.

In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it is implemented to implement the tire-road contact three-way force sensor provided by the above methods. A method for determining a decoupling model, the method comprising: establishing an artificial neural network structure for a tire-road contact three-way force sensor decoupling model, the artificial neural network structure comprising an input layer, a hidden layer and an output layer, the input layer and the The activation function of the hidden layer is a ReLU activation function; the Adam learning rate self-adaptive optimization algorithm is used to adjust the parameters of the artificial neural network structure based on the optimization parameters to obtain the optimized tire-road contact three-way force sensor decoupling model , wherein the optimization parameters include at least one of the following: step size, exponential decay rate of moment estimation, small constant for numerical stability, and descending gradient; adopt batch update method, based on training data, iteratively update the tire- The parameters of the road contact three-way force sensor decoupling model until the value of the loss function MSE continues to decrease, and the tire-road contact three-way force sensor decoupling model used for decoupling in the actual measurement of the tire-road contact three-way force is obtained , wherein the training data includes the input loads corresponding to the three directions of the tire-road contact three-way force sensor, and the input loads corresponding to the three directions respectively correspond to the respective expected output electrical signals.

The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without any creative efforts.

Through the above description of the implementations, those skilled in the art can clearly understand that each implementation can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic discs, optical discs, etc., including several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (10)

1. A method for determining a decoupling model of a tire-road contact three-way force sensor, comprising the steps of:

an artificial neural network structure of a tire-road contact three-way force sensor decoupling model is established, the artificial neural network structure comprises an input layer, a hidden layer and an output layer, and the activation functions of the input layer and the hidden layer are ReLU activation functions;

adopting an adaptive optimization algorithm of an Adam learning rate, and adjusting parameters of the artificial neural network structure based on optimization parameters to obtain an optimized tire-road contact three-way force sensor decoupling model, wherein the optimization parameters comprise at least one of the following: step size, exponential decay rate of moment estimation, small constants for numerical stabilization, and gradient of descent;

And iteratively updating parameters of the tire-road contact three-way force sensor decoupling model based on training data by adopting a batch updating method until the value of a loss function MSE is continuously reduced to obtain the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force, wherein the training data comprises input load amounts respectively corresponding to three directions of the tire-road contact three-way force sensor, and the input load amounts respectively corresponding to the three directions respectively correspond to respective expected output electric signals.

2. The method for determining a decoupling model of a tire-road contact three-way force sensor according to claim 1, wherein the step of iteratively updating parameters of the decoupling model of the tire-road contact three-way force sensor based on training data using a batch update method until a value of a loss function MSE is continuously decreased, to obtain the decoupling model of the tire-road contact three-way force sensor for decoupling in an actual measurement of the tire-road contact three-way force, comprises:

executing a plurality of parameter updating processes to obtain MSE values respectively obtained in the plurality of parameter updating processes, stopping executing the parameter updating process under the condition that the MSE values are continuously reduced, and determining the tire-road contact three-way force sensor decoupling model for decoupling in actual measurement of the tire-road contact three-way force based on the tire-road contact three-way force sensor decoupling model updated in the parameter updating process executed last time;

Wherein, the parameter updating process comprises: n input load amounts are selected from training data in three directions in a training set, single-axis loading is carried out in the three directions respectively, output electric signal data are detected, and MSE values are obtained according to the output electric signal data and expected output electric signals corresponding to the input load amounts respectively.

3. The method of determining a three-way contact force sensor decoupling model according to claim 2, wherein the determining the three-way contact force sensor decoupling model for decoupling in the actual measurement of three-way contact force based on the three-way contact force sensor decoupling model updated in the last-performed parameter updating process includes:

inputting output electric signal data in three directions in the verification set into a tire-road contact three-way force sensor decoupling model updated in the parameter updating process executed in the last time, and obtaining input loads in three directions output by the tire-road contact three-way force sensor decoupling model updated in the parameter updating process executed in the last time;

and under the condition that the input loads in three directions output by the tire-road contact three-way force sensor decoupling model are verified to be consistent with the actual loads corresponding to the output electric signal data in the three directions, taking the tire-road contact three-way force sensor decoupling model updated in the last execution parameter updating process as the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force.

4. A method of determining a decoupling model for a tire-road contact three-way force sensor as in claims 1-3, further comprising:

inputting the output electric signals acquired by the tire-road contact three-way force measuring system into a tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force, and obtaining input load values in the three directions output by the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force.

5. The method of determining a decoupling model for a tire-road contact three-way force sensor of claim 1, further comprising:

static calibration is carried out on the tire-road contact three-way force sensor, and a static calibration result is obtained; and the static calibration result is used for representing a calibration coefficient matrix of the tire-road contact three-way force sensor.

6. The method for determining the decoupling model of the tire-road contact three-dimensional force sensor according to claim 1, wherein the hidden layers are divided into 5 layers, and the number of neurons contained in each hidden layer is 40, 160, 640, 160 and 40 in sequence.

7. A tire-road contact three-way force sensor decoupling model determining device, characterized by comprising:

The device comprises a building module, a model-based module and a model-based module, wherein the building module is used for building an artificial neural network structure of a tire-road contact three-way force sensor decoupling model, the artificial neural network structure comprises an input layer, a hidden layer and an output layer, and the activation functions of the input layer and the hidden layer are ReLU activation functions;

the first acquisition module is used for adopting an adaptive optimization algorithm of an Adam learning rate, adjusting parameters of the artificial neural network structure based on optimization parameters to obtain an optimized tire-road contact three-way force sensor decoupling model, wherein the optimization parameters comprise at least one of the following: step size, exponential decay rate of moment estimation, small constants for numerical stabilization, and gradient of descent;

the second obtaining module is used for iteratively updating parameters of the tire-road contact three-way force sensor decoupling model based on training data by adopting a batch updating method until the value of a loss function MSE is continuously reduced to obtain the tire-road contact three-way force sensor decoupling model for decoupling in the actual measurement of the tire-road contact three-way force, wherein the training data comprises input load amounts respectively corresponding to three directions of the tire-road contact three-way force sensor, and the input load amounts respectively corresponding to the three directions respectively correspond to respective expected output electric signals.

8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the tire-road contact three-way force sensor decoupling model determination method of any one of claims 1 to 6 when the program is executed by the processor.

9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a tire-road contact three-way force sensor decoupling model determination method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements a method for determining a decoupling model of a tire-road contact three-way force sensor according to any one of claims 1 to 6.

Images