WO2018078769A1

WO2018078769A1 - Control device

Info

Publication number: WO2018078769A1
Application number: PCT/JP2016/081917
Authority: WO
Inventors: 中川　慎二
Original assignee: 株式会社日立製作所
Priority date: 2016-10-27
Filing date: 2016-10-27
Publication date: 2018-05-03

Abstract

A control unit 21 calculates an output signal used for control, on the basis of an input signal. A diagnostic unit 1 (22) updates first criteria at a first point in time on the basis of target data representing data handled by the control unit 21, and diagnoses the target data at a second point in time subsequent to the first point in time in accordance with the updated first criteria. A diagnostic unit 2 (23), meanwhile, diagnoses the target data at the second point in time in accordance with predetermined second criteria. A control changing unit 24 performs a process matching the combination of diagnostic results 1 from the diagnostic unit 1 (22) and diagnostic results 2 from the diagnostic unit 2 (23).

Description

Control device

The present invention relates to a control device.

There is a special table 2008-503012 (Patent Document 1) as background technology in this technical field. This patent document 1 describes a “method for monitoring the operation of a control loop in a process plant, in a first operation region of the control loop in the process plant related to the operation of a unit in the process plant. Collecting relevant process gain data; determining a process gain prediction behavior in the first operating region based on the collected process gain data; and during operation of the control loop, the first operation Monitoring the process gain in a region; determining when the process gain being monitored is substantially deviated from the process gain prediction behavior in the first operating region; and Based on at least a substantial deviation from the process gain prediction behavior in the operating region, the control loop and the unit And a determining the abnormal situation associated with at least one of the operations have been described as methods "([Claim 1] see).

In addition, this Patent Document 1 discloses a “method for monitoring the operation of a control loop in a process plant, and collects process gain data related to the first operation region of the control loop in the process plant. Determining a predicted behavior of the process gain of the first operating region based on the collected process gain data; and for an expert engine, the process gain of the first operating region. Providing data indicative of predicted behavior; providing data of process gain associated with the control loop to the expert engine during operation of the control loop; and predicting behavior of the process gain. Based on the data shown and the process gain data associated with the control loop during operation of the control loop, And a the use of the expert engine to detect an abnormal situation associated with the control loop, is described as a method "(see [Claim 16]).

Special table 2008-503012 gazette

However, the above-mentioned prior art (Patent Document 1) does not change the contents of the control function using the result of detecting the abnormality of the control function. In addition, an abnormality detection function (diagnostic function) having different functions (a function capable of detecting an unknown abnormality / a function of detecting a known abnormality) is not operated simultaneously (in parallel).

An object of the present invention is to provide a control device capable of changing the control according to the diagnosis result while improving the reliability of the diagnosis.

In order to achieve the above object, the present invention provides a control unit that calculates an output signal used for control based on an input signal, and a first determination at a first timing based on target data indicating data handled by the control unit. A first diagnosing unit for diagnosing the target data at a second timing after the first timing according to the updated first determination criterion; and a second timing according to a predetermined second determination criterion. A second diagnosis unit that diagnoses the target data; and a control change unit that executes a process corresponding to a combination of the diagnosis result of the first diagnosis unit and the diagnosis result of the second diagnosis unit.

According to the present invention, the control can be changed according to the diagnosis result while improving the reliability of the diagnosis. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments.

FIG. 9 is an overall system diagram in which the control device according to the first to eighth embodiments operates. FIG. 6 is a diagram showing processing contents implemented in a ROM in the first to fourth embodiments. FIG. 9 is a diagram showing the entire diagnosis unit 1 in Embodiments 1 and 5 to 8. FIG. 10 is a diagram illustrating processing of a determination criterion update unit A in the first, fifth to eighth embodiments. It is a figure which shows the example of a process result of the judgment reference update part A in Embodiment 1. FIG. FIG. 10 is a diagram showing processing of a determination criterion update unit B in Embodiments 1, 5 to 8. It is a figure which shows the example of a process result of the judgment reference update part B in Embodiment 1. FIG. FIG. 10 is a diagram illustrating processing of a determination unit in the first, fifth, and eighth embodiments. It is the figure which showed the process of the diagnostic part 2 in Embodiment 1, 5, 6. FIG. FIG. 6 is a diagram showing processing of a control change unit in the first to fourth embodiments. It is a figure showing the whole diagnostic part 1 in Embodiment 2. FIG. FIG. 10 is a diagram illustrating a process of a determination criterion update unit A in the second embodiment. FIG. 10 is a diagram illustrating processing of a determination criterion update unit B in the second embodiment. FIG. 10 is a diagram illustrating processing of a determination unit in the second embodiment. FIG. 10 is a diagram showing processing of a diagnosis unit 2 in Embodiments 2 to 4, 7, and 8. It is a figure showing the whole diagnostic part 1 in Embodiment 3. FIG. It is the figure which showed the process of the judgment reference update part A in Embodiment 3. FIG. It is the figure which showed the process of the judgment reference update part B in Embodiment 3. FIG. It is a figure showing processing of a judgment part in Embodiment 3. It is a figure showing the whole diagnostic part 1 in Embodiment 4. FIG. It is the figure which showed the process of the judgment reference update part A in Embodiment 4. FIG. It is the figure which showed the process of the judgment reference update part B in Embodiment 4. FIG. It is the figure which showed the process of the judgment part in Embodiment 4. FIG. 10 is a diagram illustrating processing contents mounted on a ROM according to a fifth embodiment. It is the figure which showed the process of the control change part in Embodiment 5, 7, and 8. FIG. It is the figure which showed the process of the control part in Embodiment 5, 7, and 8. FIG. It is the figure which showed the process of emergency control 1 in Embodiment 5, 7, and 8. FIG. It is the figure which showed the process of emergency control 2 in Embodiment 5, 7, and 8. FIG. It is the figure which showed the process of emergency control 3 in Embodiment 5, 7, and 8. FIG. FIG. 18 is a diagram illustrating processing contents implemented in a ROM according to a sixth embodiment. It is the figure which showed the process of the control part in Embodiment 6. FIG. 20 is a diagram illustrating processing contents implemented in a ROM according to the seventh embodiment. FIG. 20 is a diagram illustrating processing contents implemented in a ROM according to an eighth embodiment. It is a figure which shows the control table of Embodiment 1. FIG. 3 is a time chart for explaining the control of the first embodiment. It is a figure which shows the control table of Embodiment 5, 7, and 8. FIG. 10 is a time chart for illustrating control of Embodiments 5, 7, and 8. It is a figure which shows the control table of Embodiment 6. 10 is a time chart for explaining the control of the sixth embodiment.

Hereinafter, the configuration and operation of the control device according to the first to seventh embodiments of the present invention will be described. The control device according to the first to eighth embodiments includes a diagnosis unit 1 that diagnoses a control function according to a determination criterion that is updated based on an output value or the like of the control function, and a diagnosis unit 1 that is determined according to a predetermined determination criterion. Independently (in parallel processing), the diagnosis unit 2 for diagnosing the control function, and depending on the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2, predetermined processing (control content of the control function is set) And a control change unit that executes (including processing to change).

Embodiment 1
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses machine learning (clustering). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method. When the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, a notification is given.

Hereinafter, the configuration and operation of the control device according to Embodiment 1 of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing the entire control device. The control device 11 is provided with an input circuit 16 for processing an external signal. Examples of the signal from the outside here include a sensor signal, an information signal from a server, an information signal from another control device or a data processing device, and the like. Signals from the outside are transmitted via a data bus, a dedicated line, wireless, or the like. A signal from the outside passes through the input circuit 16 and becomes an input signal and is sent to the input / output port 17. The input information sent to the input / output port 17 is written into the RAM 14 through the data bus 15.

The input circuit 16 and the input / output port 17 constitute an input device for inputting an input signal.

Various processes (programs) are written in the ROM 13 and executed by the CPU 12. At that time, the value written in the RAM 14 is used as appropriate for calculation. Of the calculation results, information (value) to be sent to the outside is sent to the input / output port 17 through the data bus 15 and sent to the output circuit 18 as an output signal. The signal is output from the output circuit 18 to the outside of the control device 11. The signal to the outside here may be a drive signal to an actuator, an information signal to a server, an information signal to another control device or a data processing device, and the like.

The input / output port 17 and the output circuit 18 constitute an output device that outputs an output signal.

FIG. 2 shows the contents of the process written in the ROM 13. As described above, using the input signal written in the RAM 14 through the input circuit 16 and the input / output port 17, the control unit 21 calculates an output signal such as a control signal necessary for control.

In other words, the control unit 21 calculates an output signal used for control based on the input signal. The CPU 12 (arithmetic unit) functions as the control unit 21, the diagnostic unit 1 (22), the diagnostic unit 2 (23), and the control change unit 24 by executing the processing contents (program) stored in the ROM 13. To do. That is, the CPU 12 (arithmetic unit) includes a control unit, a diagnosis unit 1, a diagnosis unit 2, and a control change unit (hereinafter, the same applies to other embodiments).

As described above, the output signal is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Each of the diagnosis unit 1 (22) and the diagnosis unit 2 (23) detects an abnormality of the control unit 21 using parameters calculated by the control unit 21. The parameter calculated by the control unit 21 may be an intermediate calculation result, or may be an input signal or an output signal to the control unit 21. The control change unit 24 performs control change (in this embodiment, abnormality notification) using the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23).

In other words, the diagnosis unit 1 (22) updates the first determination criterion at the first timing based on the target data indicating the data handled by the control unit 21, and according to the updated first determination criterion (first rule). The target data at the second timing after the first timing is diagnosed. On the other hand, the diagnosis unit 2 (23) diagnoses the target data at the second timing according to a predetermined second determination criterion (second rule). The control change unit 24 executes processing (abnormality notification) corresponding to the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23).

Parameter transmission / reception among the control unit 21, the diagnosis unit 1 (22), the diagnosis unit 2 (23), and the control change unit 24 is generally sent via the data bus via the RAM 14. The abnormality notification is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Details of each process will be described below.

<Diagnostic section 1 (FIG. 3)>
FIG. 3 is a diagram showing the entire diagnosis unit 1 (22), and includes the following calculation units.
・ Criteria update part A (100)
・ Criteria update part B (110)
・ Decision part (120)

Hereinafter, details of each part of the diagnosis unit 1 will be described.

<Judgment Criteria Update Unit A (FIG. 4)>
In this process, clustering information is calculated using data. Specifically, it is shown in FIG.

Clustering vectorized learning data / parameter values (vectors) using machine learning k-means. The obtained clustering result is output by the k-means method. The clustering result here is
・ Cluster number to which data divided by k-means method belongs ・ Average value of data belonging to each cluster (center vector)
Point to. In other words, the determination criterion update unit A (diagnosis unit 1) extracts the characteristics of the target data.

The details of the k-means method are described in many documents and books, and will not be described in detail here. FIG. 5 shows an example of the result of clustering a data group consisting of a certain two-dimensional vector with 4 clusters by the k-means method.

<Judgment Criteria Update Unit B (FIG. 6)>
In this process, the data range is set for each clustered data set using the above-described clustering information calculated by the determination criterion update unit A, and the result is output as range information. Specifically, it is shown in FIG.

・ The minimum value of each dimension of the data belonging to each cluster is set as the lower limit of each dimension in the range corresponding to each cluster.

・ The maximum value of each dimension of data belonging to each cluster is set as the upper limit of each dimension in the range corresponding to each cluster.

The criterion 1 here is:
-Refers to the lower and upper limits of each dimension that define the range corresponding to each cluster (center vector). In other words, the determination criterion update unit B (diagnosis unit 1) updates the determination criterion 1 based on the characteristics extracted by the determination criterion update unit A.

FIG. 7 shows an example of the result of setting the range by the setting means based on the division result shown in FIG.

<Decision part (FIG. 8)>
In this process, it is determined whether newly obtained data is normal (abnormal) based on the above-described range information. Specifically, it is shown in FIG.

∙ Specify the closest center vector in terms of L2 distance to the data / parameter value (vector) at the time of diagnosis.

・ If the data / parameter value (vector) at the time of diagnosis exists within the range corresponding to the specified center vector, it is determined as normal and f_ano_1 = 0. If it does not exist, it is determined as abnormal and f_ano_1 = 1.

<Diagnostic section 2 (FIG. 9)>
The processing content of the diagnosis part 2 (23) is shown. Specifically, it is shown in FIG.

The following rule-based processing is performed on x1 and x2 (two dimensions here) that are data / parameter values (vectors) at the time of diagnosis.

if
x1 ≦ k_x1_L
or
x1 ≧ k_x1_H
or
x2 ≦ k_x2_L
or
x2 ≧ k_x2_H
then
f_ano_2 = 1
else
f_ano_2 = 0
Here, k_x1_L, k_x1_H, k_x2_L, and k_x2_H are values that are determined in advance, and are desirably set as values corresponding to levels at which the control function is abnormal.

<Control change part (FIG. 10)>
The processing content of the control change part 24 is shown. Specifically, it is shown in FIG.

・ F_ano_1 = 1 and f_ano_2 = 0
In this case, f_ano_F = 1 is set and an abnormality is notified.

・ Other than that,
f_ano_F = 0 is set, and no abnormality is reported.

Note that the ROM 13 (storage device) of the control device 11 stores a control table (database) as shown in FIG. 34A. The control table of the present embodiment is a combination of the diagnosis result of the diagnosis unit 1 being abnormal (f_ano_1 = 1) and the diagnosis result of the diagnosis unit 2 being normal (f_ano_2 = 0), and an emergency process as a process corresponding to this combination. Time control 1 (f_ano_F = 1) is associated and stored. The control change unit 24 reads a process corresponding to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 from the control table, and executes the read process (emergency control 1: notification).

FIG. 34B is an example of a time chart of f_ano_1, f_ano_2, and f_ano_F. For example, when f_ano_F = 1, the CPU 12 (arithmetic unit) may turn on a warning lamp.

In this embodiment, an example of diagnosis using two-dimensional data is shown. However, both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) can be expanded in dimension, and are applied only to two-dimensional data. It is added that it is not a thing.

This configuration is also applied to a system composed of a central processing unit represented by a cloud and a plurality of terminals. In this case, the determination reference update unit A (100) and the determination reference update unit B (110) may perform processing in the cloud and use the calculation results of the control units in a plurality of terminals as data used at that time. Further, the determination unit (120) may be implemented at each terminal and detect an abnormality of the control unit of each terminal.

As described in the text, the data used in the judgment standard update unit A (100) and the judgment standard update unit B (110) are different from the data used in the judgment unit (120). For example, the determination reference update unit A (100) and the determination reference update unit B (110) perform processing using data when there is no problem in the processing of the control unit 21 (when there is a track record). The determination unit (120) is implemented to detect whether there is a problem in the function (software) when there is a possibility that there is a problem in the processing contents of the control unit 21 (when there is no record). As an example when there is a possibility that there is a problem (when there is no record), a case where the processing content of the control unit 21 is updated is considered.

The difference in function / action / effect between the diagnosis unit 1 (22) and the diagnosis unit 2 (23) will be described below.

-Function-Diagnosis unit 1: Machine learning Using data, for example, the data is divided into each set by clustering on the basis of a short distance. Diagnosis results vary according to the characteristics of the data used when generating the reference.

・ Diagnosis unit 2: Rule base Based on "known information (information that people know)", (generally) people specify criteria. The diagnosis result is unchanged.

-Action-Diagnosis unit 1: Machine learning Identify normal / abnormal (danger) based on diagnostic criteria generated using data.
It is possible to detect normal / abnormal (dangerous) unknown (hidden by humans) hidden in the data.

・ Diagnosis unit 2: Rule base Based on “known information (information that people know)”, (generally) people specify criteria, so they identify known (invariant) normal / abnormal (danger) Effect ・ Diagnostic part 1: Machine learning An unknown abnormality (danger) may be avoided.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

As described above, according to the configuration shown in the present embodiment, the function of the control unit 21 is diagnosed using both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) having different functions / actions / effects. Highly reliable diagnosis is possible.

Therefore, the control can be changed according to the diagnosis result while improving the reliability of the diagnosis.

[Embodiment 2]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses dimension reduction processing (sparse modeling). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method. When the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, a notification is given.

That is, the present embodiment is different from the first embodiment in that the judgment criterion of the diagnosis unit 1 is updated using the dimension reduction process. Hereinafter, the configuration and operation of the control apparatus according to the second embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing the entire control apparatus, but since it is the same as that of the first embodiment, it will not be described in detail. FIG. 2 shows the contents of the process written in the ROM 13, but it is the same as that of the first embodiment and will not be described in detail. Details of each process will be described below.

<Diagnostic section 1 (FIG. 11)>
FIG. 11 is a diagram showing the entire diagnosis unit 1 (22), and includes the following calculation units.

・ Criteria update part A (130)
・ Criteria update part B (140)
・ Decision part (150)
Hereinafter, the detail of each part of the diagnostic part 1 is demonstrated.

<Judgment Criteria Update Unit A (FIG. 12)>
In this processing, dimension reduction processing is performed using data. Specifically, it is shown in FIG.

∙ Reduce dimensionality of data / parameter values (vectors) during vectorization using sparse modeling (such as lasso). The output is the criterion 1a (linear format) and Phi (evaluation function value). The criterion 1a is a result of dimension reduction and line approximation by sparse modeling. Phi is the value of the evaluation function used for dimension reduction and linear approximation, and indicates the accuracy of linear approximation. Details of sparse modeling such as lasso regression have been described in many documents and books, and will not be described in detail here.

<Judgment Criteria Updater B (FIG. 13)>
In this process, the criterion 1b is calculated using the calculation result in the criterion update unit A. Specifically, as shown in FIG. 13, Div1 (judgment criterion 1b) is calculated as follows.
Div1 = k1 × Phi

The allowable range is defined using Phi, which is the accuracy of linear approximation. In this example, Phi is multiplied by k1, but another function related to Phi may be used.

<Decision part (FIG. 14)>
In this process, it is determined whether the newly obtained data is normal (abnormal) based on the determination criterion 1b obtained by the determination criterion update unit B. Specifically, it is shown in FIG.

・ From the data (vector) to be diagnosed, extract only the dimension elements that make up the linear format that is the criterion 1a.

・ A distance (normal length) between a new vector composed only of extracted elements and a criterion 1a (line format) is obtained.

・ When (normal length) ≧ Div1, set f_ano_1 = 1.

Otherwise, f_ano_1 = 0.

<Diagnostic section 2 (FIG. 15)>
The processing content of the diagnosis part 2 (23) is shown. Specifically, it is shown in FIG.

The following rule-based processing is performed on x1, x2, ... xn (n dimensions) that are data / parameter values (vectors) at the time of diagnosis.

if
x1 ≦ k_x1_L
or
x1 ≧ k_x1_H
or
x2 ≦ k_x2_L
or
x2 ≧ k_x2_H
...
or
xn ≦ k_xn_L
or
xn ≧ k_xn_H
then
f_ano_2 = 1
else
f_ano_2 = 0
Here, k_x1_L, k_x1_H, k_x2_L, k_x2_H,..., K_xn_L, k_xn_H are values that are determined in advance, and are preferably set as values corresponding to levels at which the control function is abnormal.

<Control change part (FIG. 10)>
The processing content of the control change part 24 is shown. Specifically, although it is shown in FIG. 10, it is the same as that of the first embodiment, and therefore will not be described in detail.

This configuration is also applied to a system composed of a central processing unit represented by a cloud and a plurality of terminals. In this case, the determination criterion update unit A (130) and the determination criterion update unit B (140) may perform processing in the cloud and use the calculation results of the control units in a plurality of terminals as data used at that time. Further, the determination unit (150) may be implemented in each terminal and detect an abnormality in the control unit of each terminal.

As described in the text, the data used in the judgment standard update unit A (130) and the judgment standard update unit B (140) are different from the data used in the judgment unit (150). For example, the determination reference update unit A (130) and the determination reference update unit B (140) perform processing using data when there is no problem in the processing of the control unit 21 (when there is a track record). The determination unit (150) is implemented to detect whether there is a problem in the function (software) when there is a possibility that there is a problem in the processing contents of the control unit 21 (when there is no record). As an example when there is a possibility that there is a problem (when there is no record), a case where the processing content of the control unit 21 is updated is considered.

・ Function ・ Diagnosis unit 1: Dimension reduction Uses data to generate diagnostic criteria (formulas, models, etc.) represented by only significant dimensions, leaving only significant dimensions for data with a very large number of dimensions. To do. Diagnosis results vary according to the characteristics of the data used when generating the reference.

-Action-Diagnosis unit 1: Dimension reduction Identifies normal / abnormal (danger) based on diagnostic criteria generated using data.
It may be possible to detect unknown normality / abnormality (danger) hidden in data. Alternatively, an abnormality (danger) can be predicted.

・ Diagnosis unit 2: Rule base Based on “known information (information that people know)”, (generally) people specify criteria, so they identify known (invariant) normal / abnormal (danger) Effect ・ Diagnosis unit 1: Dimension reduction Unknown abnormality (danger) may be avoided. Or you can expect.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

[Embodiment 3]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses compression processing (frequency analysis). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method. When the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, a notification is given.

That is, the present embodiment is different from the first embodiment in that the determination standard of the diagnosis unit 1 is updated using the compression process. Hereinafter, the configuration and operation of a control device according to Embodiment 3 of the present invention will be described with reference to the drawings.

<Diagnostic section 1 (FIG. 16)>
FIG. 16 is a diagram showing the entire diagnosis unit 1 (22), and includes the following calculation units.

・ Criteria update part A (160)
・ Criteria update part B (170)
・ Judgment part (180)
Hereinafter, the detail of each part of the diagnostic part 1 is demonstrated.

<Judgment Criteria Update Unit A (FIG. 17)>
In this processing, frequency analysis processing is performed using data. Specifically, it is shown in FIG.

∙ Use frequency analysis (Fourier transform etc.) for each component of vectorized learning data / parameter value (vector) to obtain the result of frequency analysis. Note that details of frequency analysis such as Fourier transform have been described in many documents and books, and will not be described in detail here.

<Judgment Criteria Updater B (FIG. 18)>
In this process, F_1_k (determination criterion 1) is calculated using the calculation result in the determination criterion update unit A. Specifically, as shown in FIG. 18, F_1_k (judgment criterion 1b) is obtained by the following processing.

・ F_i_k is a frequency whose power spectrum is greater than or equal to the predetermined value k_F1_i.

i: 1, 2, 3, ..., number of vector elements (number of components),
k: 1, 2, 3,..., number of relevant frequencies k_F1_i is preferably set for each element (each component).

<Decision part (FIG. 19)>
In this process, based on the determination criterion 1 obtained by the determination criterion update unit B, it is determined whether the newly obtained data is normal (abnormal). Specifically, it is shown in FIG.

・ Frequency analysis of each element of diagnosis target data (vector).

However, the target frequency is F_i_k.

・ When the power spectrum value of each frequency F_i_k is greater than or equal to k_F2_i
Let f_ano_1 = 1.

・ F_ano_1 = 0 otherwise.

It should be noted that power other than F_i_k is also targeted, and it may be determined as abnormal when the power exceeds a predetermined value.

<Diagnostic section 2 (FIG. 15)>
The processing contents of the diagnosis unit 2 (23) are specifically shown in FIG. 15, but are not described in detail because they are the same as those in the second embodiment.

<Control change part (FIG. 10)>
The processing contents of the control change unit 24 are specifically shown in FIG. 10, but are not described in detail because they are the same as those in the first embodiment.

This configuration is also applied to a system composed of a central processing unit represented by a cloud and a plurality of terminals. In this case, the determination criterion update unit A (160) and the determination criterion update unit B (170) may perform processing in the cloud and use the calculation results of the control units in a plurality of terminals as data used at that time. Further, the determination unit (180) may be implemented in each terminal and detect an abnormality in the control unit of each terminal.

As described in the text, the data used in the judgment standard update unit A (160) and the judgment standard update unit B (170) are different from the data used in the judgment unit (180). For example, the determination reference update unit A (160) and the determination reference update unit B (170) perform processing using data when there is no problem in the processing of the control unit 21 (when there is a track record). The determination unit (180) is implemented to detect whether there is a problem in the function (software) when there is a possibility that there is a problem in the processing contents of the control unit 21 (when there is no record). As an example when there is a possibility that there is a problem (when there is no record), a case where the processing content of the control unit 21 is updated is considered.

・ Function ・ Diagnostic unit 1: Compression (frequency analysis)
Using the data, only a significant frequency is left for data with a large number of data points in the time direction, and a diagnostic criterion represented only by the significant frequency is generated. Diagnosis results vary according to the characteristics of the data used when generating the reference.

・ Operation ・ Diagnostic unit 1: Compression (frequency analysis)
Identify normal / abnormal (dangerous) based on diagnostic criteria generated using data.

It may be possible to detect unknown normal / abnormality (danger) hidden in the data. Alternatively, an abnormality (danger) can be predicted.

・ Diagnosis unit 2: Rule base Based on “known information (information that people know)”, (generally) people specify criteria, so they identify known (invariant) normal / abnormal (danger) Effect ・ Diagnostic unit 1: Compression (frequency analysis)
An unknown abnormality (danger) may be avoided. Or you can expect.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

[Embodiment 4]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses statistical processing (calculating an average value and a variance value). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method. When the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, a notification is given.

That is, this embodiment is different from the first embodiment in that the determination criteria of the diagnosis unit 1 are updated using statistical processing. Hereinafter, the configuration and operation of a control device according to Embodiment 4 of the present invention will be described with reference to the drawings.

<Diagnostic section 1 (FIG. 20)>
FIG. 20 is a diagram showing the entire diagnosis unit 1 (22), and includes the following calculation units.

・ Criteria update part A (190)
・ Criteria update part B (200)
・ Decision part (210)
Hereinafter, the detail of each part of the diagnostic part 1 is demonstrated.

<Judgment Criteria Update Unit A (FIG. 21)>
In this processing, statistical processing is performed using data. Specifically, it is shown in FIG.

Statistical processing is performed for each component of vectorized learning data / parameter values (vectors), and results of average values (e_i) and variance values (sigma_i) are obtained (i: 1, 2, 3,... ,Element count).

<Judgment Criteria Update Unit B (FIG. 22)>
In this process, e_Li, e_Hi, and w_sigma_i (determination criterion 1) are calculated using the calculation result in the determination criterion update unit A. Specifically, as shown in FIG. 22, e_Li, e_Hi, and w_sigma_i (determination criterion 1) are obtained by the following processing.

・ E_L_i = e_i-k_e_L_i
・ E_H_i = e_i + k_e_H_i
・ W_sigma_i = k_i_sigma × sigma_i
i: 1, 2, 3, ..., number of elements (number of components)
k_e_L_i, k_e_H_i, k_i_sigma are set for each element (each component)

<Determining part (FIG. 23)>
In this process, based on the determination criterion 1 obtained by the determination criterion update unit B, it is determined whether the newly obtained data is normal (abnormal). Specifically, it is shown in FIG.

・ Statistically process each element of diagnosis target data (vector) (average, variance)
・ (Average value of i-th element) ≦ e_Li
Or (average value of i-th element) ≧ e_Hi
Or (dispersion of i-th element) ≧ w_sigma_i
In this case, f_ano_1 = 1.

・ In other cases, set f_ano_1 = 0.

In the present embodiment, as a statistical process, a method for obtaining an average value and a variance value assuming a normal distribution is used. However, there is a method for obtaining an arbitrary distribution that does not assume a normal distribution, such as the MCMC method (Markov chain Monte Carlo method). Such a method may be used.

This configuration is also applied to a system composed of a central processing unit represented by a cloud and a plurality of terminals. In this case, the determination criterion update unit A (190) and the determination criterion update unit B (200) may be processed in the cloud, and the calculation results of the control units in a plurality of terminals may be used as data used at that time. Further, the determination unit (210) may be implemented in each terminal and detect an abnormality in the control unit of each terminal.

As described in the text, the data used in the judgment standard update unit A (190) and the judgment standard update unit B (200) is different from the data used in the judgment unit (210). For example, the determination reference update unit A (190) and the determination reference update unit B (200) perform processing using data when there is no problem in the processing of the control unit 21 (when there is a track record). The determination unit (210) is implemented to detect whether there is a problem in the processing contents of the control unit 21 (when there is no record) and whether or not the function (software) is abnormal. As an example when there is a possibility that there is a problem (when there is no record), a case where the processing content of the control unit 21 is updated is considered.

・ Function ・ Diagnosis unit 1: Statistical processing Using data, only significant information (average value, variance value, etc.) indicating statistical characteristics is left for data with a large number of data points. Generated diagnostic criteria. Diagnosis results vary according to the characteristics of the data used when generating the reference.

・ Operation ・ Diagnostic unit 1: Compression (frequency analysis)
Identify normal / abnormal (dangerous) based on diagnostic criteria generated using data.
It may be possible to detect unknown normality / abnormality (danger) hidden in data. Alternatively, an abnormality (danger) can be predicted.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

[Embodiment 5]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses machine learning (clustering). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method.

When the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, a notification is given. Further, when the diagnosis result of the diagnosis unit 1 is normal and the diagnosis result of the diagnosis unit 2 is abnormal, the internal calculation value of the control function is restricted within a predetermined range. Further, when the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is abnormal, the control function is stopped.

That is, in the present embodiment, compared with the first embodiment, the control content of the control function (control unit) is controlled normally or according to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2. The point of changing to one of emergency controls 1 to 4 is different. Hereinafter, the configuration and operation of a control device according to Embodiment 5 of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing the entire control apparatus, but since it is the same as that of the first embodiment, it will not be described in detail.

FIG. 24 shows the contents of processing written in the ROM 13. As described above, the control unit 31 calculates an output signal such as a control signal necessary for control using the input signal written in the RAM 14 via the input circuit 16 and the input / output port 17. As described above, the output signal is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Each of the diagnosis unit 1 (22) and the diagnosis unit 2 (23) detects an abnormality of the control unit 31 using parameters calculated by the control unit 31. The parameter calculated by the control unit 31 may be an intermediate calculation result, or may be an input signal or an output signal to the control unit 31. The control change unit 32 changes the control based on the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23).

In other words, the control change unit 32 executes a process of changing the control of the control unit 31 as a process corresponding to the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23). To do.

Parameter transmission / reception among the control unit 31, the diagnosis unit 1 (22), the diagnosis unit 2 (23), and the control change unit 32 is generally sent via the data bus via the RAM 14. The abnormality notification is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Details of each process will be described below.

<Diagnostic section 1 (FIG. 3)>
FIG. 3 is a diagram showing the entire diagnosis unit 1 (22), which includes the following calculation units and is the same as in the first embodiment.

・ Criteria update part A (100)
・ Criteria update part B (110)
・ Decision part (120)
Hereinafter, the detail of each part of the diagnostic part 1 is demonstrated.

<Judgment Criteria Update Unit A (FIG. 4)>
In this process, clustering information is calculated using data. Specifically, although it is shown in FIG. 4, it is not described in detail because it is the same as the first embodiment.

<Judgment Criteria Update Unit B (FIG. 6)>
In this process, the data range is set for each clustered data set using the above-described clustering information calculated by the determination criterion update unit A, and the result is output as range information. Specifically, although it is shown in FIG. 6, it is not described in detail because it is the same as the first embodiment.

<Decision part (FIG. 8)>
In this process, it is determined whether newly obtained data is normal (abnormal) based on the above-described range information. Specifically, although shown in FIG. 8, it is the same as that of the first embodiment, and therefore will not be described in detail.

<Diagnostic section 2 (FIG. 9)>
The processing contents of the diagnosis unit 2 (23) are specifically shown in FIG. 9, but will not be described in detail because they are the same as those in the first embodiment.

<Control change unit (FIG. 25)>
The processing content of the control change part 32 is shown. Specifically, it is shown in FIG.

・ F_ano_1 = 0 and f_ano_2 = 0
When c_mode = 0
・ F_ano_1 = 1 and f_ano_2 = 0
When c_mode = 1
・ F_ano_1 = 0 and f_ano_2 = 1
When c_mode = 2
・ F_ano_1 = 1 and f_ano_2 = 1
C_mode = 3
Note that the ROM 13 (storage device) of the control device 11 stores a control table (database) as shown in FIG. 35A. The control table stores a combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 and a process corresponding to this combination in association with each other. The control change unit 32 reads a process corresponding to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 from the control table, and executes the read process (a process for changing the control of the control unit 31).

<Control unit (FIG. 26)>
The processing content of the control part 31 is shown. Specifically, it is shown in FIG.

・ When c_mode = 0 Normal control is performed.

It should be noted that the contents of normal control are existing control and depend on the control target and control system.

-When c_mode = 1 Emergency control 1 (41) is executed.

・ When c_mode = 2: Perform emergency control 2 (42).

・ When c_mode = 3: Perform emergency control 3 (43).

FIG. 35B is an example of a time chart of f_ano_1, f_ano_2, and c_mode. The control unit 31 changes the control to either normal control or emergency control 1 to 3 according to c_mode.

<Emergency control 1 (FIG. 27)>
The contents of emergency control 1 (41) are shown. Specifically, it is shown in FIG.

・ When c_mode = 1
Anomaly is reported as f_ano_F = 1.

・ Other than that,
f_ano_F = 0
<Emergency control 2 (FIG. 28)>
The contents of emergency control 2 (42) are shown. Specifically, it is shown in FIG.

・ When c_mode = 2
Change the data / parameter value to the nearest normal range.

・ Other than that, no change is made.

When c_mode = 2, since the diagnosis unit 2 makes an abnormality determination, x1 and x2 may be restricted to the following ranges from the diagnosis unit 2 (23) (FIG. 9).

k_x1_L <x1 <k_x1_H
k_x2_L <x2 <k_x2_H
In other words, when the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is normal, the control change unit 32 changes the target data to the data of the nearest cluster.

<Emergency control 3 (FIG. 29)>
The contents of emergency control 3 (43) are shown. Specifically, it is shown in FIG.

・ When c_mode = 3
Stop control.

・ Other than that,
Normal control is implemented.

Or, when c_mode = 3, fail-safe processing may be performed. The fail-safe process is an existing control and depends on the control target, the control system, and the like, and thus will not be described in detail here.

In the present embodiment, an example of diagnosis using two-dimensional data is shown. However, as described in the first embodiment, both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) can be expanded in dimension. It should be added that it does not apply only to two-dimensional data.

-Action-Diagnosis unit 1: Machine learning Identify normal / abnormal (danger) based on diagnostic criteria generated using data.

It may be possible to detect unknown normal / abnormality (danger) hidden in the data.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

As described above, according to the configuration shown in the present embodiment, the function of the control unit 21 is diagnosed using both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) having different functions / actions / effects. Since the control is changed based on the diagnosis result of 1 and the diagnosis unit 2, the control according to the nature of the abnormality is possible, and the performance / reliability of the control system is increased.

[Embodiment 6]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the input value to the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses machine learning (clustering). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method. When the diagnosis unit 1 or the diagnosis unit 2 determines that the input value to the control function is abnormal, the input value to the control function is set to a different value or the operation of the control function is not performed (stopped).

That is, the present embodiment is different from the first embodiment in that normality or abnormality of the input value to the control function (control unit) is determined. Also, the control content of the control function (control unit) is changed to either normal control or emergency control 2 according to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2.

FIG. 30 shows the contents of processing written in the ROM 13. As described above, using the input signal written in the RAM 14 through the input circuit 16 and the input / output port 17, the control unit 21 calculates an output signal such as a control signal necessary for control. As described above, the output signal is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. The diagnosis unit 1 (22) and the diagnosis unit 2 (23) each detect an input abnormality to the control unit 21 by using an input value to the control unit 21. The control change unit 51 changes the control based on the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23). The parameter transmission / reception among the control unit 21, the diagnosis unit 1 (22), the diagnosis unit 2 (23), and the control change unit 51 is generally sent via the data bus via the RAM 14. The abnormality notification is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Details of each process will be described below.

<Control change part (FIG. 31)>
The content of the control change part 51 is shown. Specifically, it is shown in FIG.

・ When f_ano_1 = 1 or f_ano_2 = 1
Change the data / parameter value that is the input value to the control unit to the nearest normal range.

・ Other than that, no change is made.

Alternatively, when f_ano_2 = 1, the diagnosis unit 2 makes an abnormality determination. Therefore, from the diagnosis unit 2 (23) (FIG. 9), x1 and x2 may be restricted to the following ranges.

k_x1_L <x1 <k_x1_H
k_x2_L <x2 <k_x2_H
Note that the ROM 13 (storage device) of the control device 11 stores a control table (database) as shown in FIG. 36A. The control table stores a combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 and a process corresponding to this combination in association with each other. The control change unit 51 reads a process corresponding to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 from the control table, and executes the read process (a process for changing the control of the control unit 21).

FIG. 36B is an example of a time chart showing the operations of f_ano_1, f_ano_2, and the control unit 21. The control unit 21 changes the control to emergency control 2 when f_ano_1 = 1 or f_ano_2 = 1. Note that the control change unit 51 may stop the control unit 21 when the diagnosis result of the diagnosis unit 1 is abnormal and the diagnosis result of the diagnosis unit 2 is abnormal.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

As described above, according to the configuration shown in the present embodiment, the diagnosis unit 1 (22) and the diagnosis unit 2 (23) having different functions / actions / effects are used to diagnose the input value to the control unit 21, If it is determined that there is an abnormality, the input value is restricted to the normal range, so that the risk of unexpected operation abnormality of the control unit 21 can be suppressed and highly reliable control can be performed.

[Embodiment 7]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses machine learning (clustering). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method.

Also, the control target is an internal combustion engine, and the output value of the control function is the air amount, fuel injection amount, and ignition timing of the internal combustion engine.

That is, in the present embodiment, the control device of the fifth embodiment is applied to the control of the internal combustion engine. Hereinafter, the configuration and operation of a control device according to Embodiment 7 of the present invention will be described with reference to the drawings.

FIG. 32 shows the contents of processing written in the ROM 13. As described above, the control unit 61 uses the input signal written in the RAM 14 via the input circuit 16 and the input / output port 17 to control signals (target ignition timing, target fuel injection amount, target air) required for control. The output signal is calculated. As described above, the output signal is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. The diagnosis unit 1 (22) and the diagnosis unit 2 (23) detect an abnormality of the control unit 61 using parameters (target ignition timing, target fuel injection amount, target air amount) calculated by the control unit 61. The parameter calculated by the control unit 61 may be an intermediate calculation result, or may be an input signal or an output signal to the control unit. The control change unit 32 changes the control based on the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23). Parameter transmission / reception among the control unit 61, the diagnosis unit 1 (22), the diagnosis unit 2 (23), and the control change unit 32 is generally sent via the data bus via the RAM 14. The abnormality notification is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Details of each process will be described below.

<Decision part (FIG. 8)>
In this process, it is determined whether newly obtained data is normal (abnormal) based on the above-described range information. Specifically, although shown in FIG. 8, it is the same as that of the first embodiment, and therefore will not be described in detail. In FIG. 8, two-dimensional results are shown, but the parameters (target ignition timing, target fuel injection amount, target air amount) calculated by the control unit 61 are three-dimensional and are expanded to three dimensions. This process is performed.

<Diagnostic section 2 (FIG. 15)>
The processing content of the diagnosis part 2 (23) is shown. Specifically, although it is shown in FIG. 15, it is not described in detail because it is the same as the second embodiment.

<Control change unit (FIG. 25)>
The processing content of the control change part 32 is shown. Specifically, although shown in FIG. 25, since it is the same as Embodiment 5, it does not elaborate.

<Control unit (FIG. 26)>
The processing content of the control part 61 is shown. Specifically, it is shown in FIG.

・ When c_mode = 0 Normal control is performed.

Note that the contents of normal control are existing engine control, and will not be described in detail here.

-When c_mode = 1 Emergency control 1 (41) is executed.

・ When c_mode = 2: Perform emergency control 2 (42).

・ When c_mode = 3: Perform emergency control 3 (43).

・ When c_mode = 1
Anomaly is reported as f_ano_F = 1.

As an abnormality notification method, in the case of an internal combustion engine for automobiles, an abnormal lamp may be turned on so that the driver can understand.

・ Other than that,
f_ano_F = 0
<Emergency control 2 (FIG. 28)>
The contents of emergency control 1 (42) are shown. Specifically, it is shown in FIG.

・ Other than that, no change is made.

When c_mode = 2, since the diagnosis unit 2 makes an abnormality determination, x1 and x2 may be restricted to the following ranges from the diagnosis unit 2 (23) (FIG. 9).
k_x1_L <x1 <k_x1_H
k_x2_L <x2 <k_x2_H

That is, when c_mode = 2, by limiting the target fuel injection amount, target air amount, and target ignition timing to the internal combustion engine to predetermined ranges, abnormal operation of the internal combustion engine can be prevented and operation can be continued. It is.

・ When c_mode = 3
Stop control.

That is, since both the diagnosis unit 1 and the diagnosis unit 2 have determined that there is an abnormality, it is determined that it is dangerous to continue the control, the control is stopped, and the operation of the internal combustion engine is stopped.

・ Other than that,
Normal control is implemented.

Or, when c_mode = 3, fail-safe processing may be performed. Fail-safe processing for internal combustion engines is an existing control and will not be described in detail here.

In the present embodiment, an example of diagnosis using three-dimensional data is shown. However, as described in the text of the first embodiment and the present embodiment, both of the diagnosis unit 1 (22) and the diagnosis unit 2 (23) are dimension. It should be noted that these extensions are possible and not applicable only to three-dimensional data.

-Action-Diagnosis unit 1: Machine learning Identify normal / abnormal (danger) based on diagnostic criteria generated using data.
It may be possible to detect unknown normality / abnormality (danger) hidden in data.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

As described above, according to the configuration shown in the present embodiment, the function of the control unit 61 is diagnosed by using both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) having different functions / actions / effects, and the diagnosis unit Since the control is changed based on the diagnosis result of 1 and the diagnosis unit 2, the control according to the nature of the abnormality is possible, and the performance / reliability of the control system of the internal combustion engine is increased.

[Embodiment 8]
In the present embodiment, the first diagnosis unit and the second diagnosis unit determine whether the internal calculation value of the control function is normal or abnormal. The means for updating the determination criteria of the diagnosis unit 1 uses machine learning (clustering). The diagnosis unit 2 performs a diagnosis based on a rule-based method, and a predetermined criterion for the diagnosis unit 2 is described using a rule-based method.

Also, the controlled object is the rolling process of the steel plant, and the output values of the control function are the target tension, the target reduction position, and the target roller speed. In other words, the control device calculates an output signal used for controlling the tension, the reduction position, or the rolling material moving speed of the rolling device, and controls the rolling device.

That is, in this embodiment, the control device of the fifth embodiment is applied to the control of the rolling device. Hereinafter, the configuration and operation of the control apparatus according to the eighth embodiment of the present invention will be described with reference to the drawings.

FIG. 33 shows the contents of the process written in the ROM 13. As described above, using the input signal written in the RAM 14 via the input circuit 16 and the input / output port 17, the control unit 61 uses the control signals (target tension, target pressure reduction position, target roller speed) necessary for control. The output signal is calculated. As described above, the output signal is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. The diagnosis unit 1 (22) and the diagnosis unit 2 (23) detect an abnormality of the control unit 61 using parameters (target tension, target reduction position, target roller speed) calculated by the control unit 61. The parameter calculated by the control unit 61 may be an intermediate calculation result, or may be an input signal or an output signal to the control unit. The control change unit 32 changes the control based on the combination of the diagnosis result 1 of the diagnosis unit 1 (22) and the diagnosis result 2 of the diagnosis unit 2 (23). Parameter transmission / reception among the control unit 61, the diagnosis unit 1 (22), the diagnosis unit 2 (23), and the control change unit 32 is generally sent via the data bus via the RAM 14. The abnormality notification is output as a signal to the outside through the RAM 14, the input / output port 17, and the output circuit 18. Details of each process will be described below.

<Decision part (FIG. 8)>
In this process, it is determined whether newly obtained data is normal (abnormal) based on the above-described range information. Specifically, although shown in FIG. 8, it is the same as that of the first embodiment, and therefore will not be described in detail. In FIG. 8, two-dimensional results are shown, but the parameters (target tension, target reduction position, target roller speed) calculated by the control unit 61 are three-dimensional and are expanded to three dimensions. Processing is performed.

<Control unit (FIG. 26)>
The processing content of the control part 71 is shown. Specifically, it is shown in FIG.

・ When c_mode = 0 Normal control is performed.

Note that the contents of the normal control are the existing rolling process control and will not be described in detail here.

-When c_mode = 1 Emergency control 1 (41) is executed.

・ When c_mode = 2: Perform emergency control 2 (42).

・ When c_mode = 3: Perform emergency control 3 (43).

・ When c_mode = 1
Anomaly is reported as f_ano_F = 1.

¡As a method of notifying the abnormality, in the case of rolling process control, it is conceivable to turn on the abnormal light in the central control room so that the process supervisor can understand.

・ Other than that,
f_ano_F = 0
<Emergency control 2 (FIG. 28)>
The contents of emergency control 142 are shown. Specifically, it is shown in FIG.

・ Other than that, no change is made.

k_x1_L <x1 <k_x1_H
k_x2_L <x2 <k_x2_H
That is, when c_mode = 2, by restricting the target tension, target reduction position, and target roller speed of the rolling process to predetermined ranges, abnormal operation of the rolling process can be prevented and the operation can be continued.

・ When c_mode = 3
Stop control.

That is, since both the diagnosis unit 1 and the diagnosis unit 2 have determined that there is an abnormality, it is determined that it is dangerous to continue the control, the control is stopped, and the operation of the rolling process is stopped.

・ Other than that,
Normal control is implemented.
Alternatively, when c_mode = 3, fail-safe processing may be performed. Fail-safe processing for the rolling process is an existing control and will not be described in detail here.

The difference in function / action / effect between the diagnosis unit 1 (22) and the diagnosis unit 2 (23) will be described below.
-Function-Diagnosis unit 1: Machine learning Using data, for example, the data is divided into each set by clustering on the basis of a short distance. Diagnosis results vary according to the characteristics of the data used when generating the reference.

-Diagnosis unit 2: Rule base A known abnormality (danger) can be avoided.

As described above, according to the configuration shown in the present embodiment, the function of the control unit 71 is diagnosed using both the diagnosis unit 1 (22) and the diagnosis unit 2 (23) having different functions / actions / effects, and the diagnosis unit Since the control is changed based on the diagnosis result of 1 and the diagnosis unit 2, the control according to the nature of the abnormality is possible, and the performance / reliability of the control system of the rolling process is increased.

Note that the present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above-described embodiment has been described in detail for easy understanding of the present invention, and is not necessarily limited to the one having all the configurations described. Further, a part of the configuration of an embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of an embodiment. In addition, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

In addition, each of the above-described configurations, functions, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit. Each of the above-described configurations, functions, and the like may be realized by software by interpreting and executing a program that realizes each function by a processor (CPU). Information such as programs, tables, and files for realizing each function can be stored in a recording device such as a memory, a hard disk, or an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

In the first to eighth embodiments, the CPU 12 (control change unit 24) reads the process corresponding to the combination of the diagnosis result of the diagnosis unit 1 and the diagnosis result of the diagnosis unit 2 from the ROM 12 (control table) and executes the read process. To do. As described above, the cooperation of hardware resources (CPU 12, ROM 13, etc.) has an advantageous effect that the control can be changed according to the diagnosis result while improving the reliability of the diagnosis.

In addition, the following aspects may be sufficient as embodiment of this invention.

(1) a first diagnosis unit for diagnosing a control function according to a determination criterion updated based on at least an input value to the control function, an internal operation value of the control function or an output value of the control function, and a predetermined determination According to the criteria, independently of the first diagnosis unit (in parallel processing), a second diagnosis unit that diagnoses the control function, a diagnosis result of the first diagnosis unit, and a diagnosis result of the second diagnosis unit A control apparatus comprising: a control change unit that changes the control content of the control function according to the combination.

(2) The control device according to (1), wherein the first diagnosis unit determines at least normality or abnormality of an input value to the control function, an internal calculation value of the control function, or an output value of the control function.

(3) The control device according to (1), wherein the second diagnostic unit determines at least normality or abnormality of an input value to the control function, an internal calculation value of the control function, or an output value of the control function.

(4) In (1), the means for updating the determination criteria of the first diagnosis unit is configured to input (some) characteristic or property of an input value to the control function, an internal calculation value of the control function, or an output value of the control function. A control apparatus characterized by being a calculation method for extraction.

(5) The control device according to (1), wherein the means for updating the determination criterion of the first diagnosis unit uses at least machine learning.

(6) The control device according to (1), wherein the means for updating the determination criterion of the first diagnosis unit is at least a dimension reduction process.

(7) The control apparatus according to (1), wherein the means for updating the determination criterion of the first diagnosis unit is at least a data compression process.

(8) The control device according to (1), wherein the means for updating the determination criterion of the first diagnosis unit is at least statistical processing.

(9) In (1), the second diagnosis unit diagnoses by a rule-based method, and the predetermined criterion of the second diagnosis unit is described by a rule-based method apparatus.

(10) The control device according to (10), wherein a notification is made when the diagnosis result of the first diagnosis unit is abnormal and the diagnosis result of the second diagnosis unit is normal.

(11) In (1), when the diagnosis result of the first diagnosis unit is normal and the diagnosis result of the second diagnosis unit is abnormal, the input value to the control function, the internal calculation value of the control function, or the control function A control device that restricts an output value within a predetermined range.

(12) In (1), when the diagnosis result of the first diagnosis unit is abnormal and the diagnosis result of the second diagnosis unit is abnormal, the control function is switched to a fail-safe mode or stopped. apparatus.

(13) In (1), when the first diagnosis unit or the second diagnosis unit determines that the input value to the control function is abnormal, the input value to the control function is set to a different value or the control function A control device characterized by not performing (stopping) the operation.

(14) The control device according to (1), wherein the control target is an internal combustion engine, and the output value of the control function is at least an air amount, a fuel injection amount, and an ignition timing of the internal combustion engine.

(15) The control device according to (1), wherein a control target is a rolling process of a steel plant, and an output value of the control function is at least a tension, a reduction position, and a rolling material moving speed in the rolling process.

DESCRIPTION OF SYMBOLS 1 ... Engine 2 ... Fuel injection valve 3 ... Electronic throttle 4 ... Spark plug 5 ... Rolling process 11 in steel plant ... Control device or data processing device 12 ... CPU of control device or data processing device
13 ... ROM of control device or data processing device
14 ... RAM of control device or data processing device
15 ... Data bus 16 of control device or data processing device ... Input circuit 17 of control device or data processing device ... Input / output port 18 of control device or data processing device ... Output circuit 21 of control device or data processing device ... Write to ROM Control unit 22 to be executed: diagnostic unit 1 written in ROM
23 ... Diagnostic unit 2 written in ROM
24: Control change unit 31 written in ROM ... Control unit 32 written in ROM ... Control change unit 41 written in ROM ... Emergency control 1
42 ... Emergency control 2
43 ... Emergency control 3
51 ... Control change unit 61 written in ROM ... Control unit 71 ... Control unit 100 ... Judgment reference update unit A
101 .. Inside of judgment criterion update unit A (machine learning: k-means method)
110 ... Judgment standard update part B
111... Internal 120 of the judgment criterion update unit B... Judgment unit 121.
131 ... Inside of the judgment reference update unit A (dimensional reduction: sparse modeling)
140... Criteria update unit B
141 ... Inside 150 of the judgment reference updating unit B ... Judging unit 151 ... Inside the judgment unit 160 ... Judgment standard updating unit A
161. Inside of the judgment reference update unit A (compression: frequency analysis)
170 ... Judgment standard update part B
171 ... Internal 180 of the judgment reference updating unit B ... Determination part 181 ... Internal of the judgment part 190 ... Determination standard updating part A
191 ... Inside of the criterion update unit A (statistical processing: average value and variance value)
200 ... Judgment standard update part B
201 ... Inside of judgment criterion update unit B 210 ... Determination unit 211 ... Inside of judgment unit

Claims

A control unit that calculates an output signal used for control based on an input signal;
The first determination criterion is updated at a first timing based on target data indicating data handled by the control unit, and the target data at the second timing after the first timing is diagnosed according to the updated first determination criterion. A first diagnosis unit;
A second diagnosis unit for diagnosing the target data at the second timing according to a predetermined second determination criterion;
A control change unit that executes a process corresponding to a combination of the diagnosis result of the first diagnosis unit and the diagnosis result of the second diagnosis unit;
A control device comprising:
The control device according to claim 1,
The process corresponding to the combination of the diagnosis result of the first diagnosis unit and the diagnosis result of the second diagnosis unit includes a process of changing the control of the control unit.
The control device according to claim 1,
An input device for inputting the input signal;
A storage device that stores a combination of a diagnosis result of the first diagnosis unit and a diagnosis result of the second diagnosis unit and a process corresponding to the combination;
An arithmetic unit having the control unit, the first diagnosis unit, the second diagnosis unit, and the control change unit;
An output device for outputting the output signal,
The control change unit
A control device that reads a process corresponding to a combination of a diagnosis result of the first diagnosis unit and a diagnosis result of the second diagnosis unit from the storage device and executes the read process.
The control device according to claim 1,
The target data is
The control device, wherein the control signal is the input signal, a calculation result of the control unit that is calculating the output signal from the input signal, or data indicated by the output signal.
The control device according to claim 4,
The first diagnosis unit includes:
Extracting the characteristics of the target data;
The control device updates the first determination criterion based on the extracted characteristic.
The control device according to claim 1,
The first diagnosis unit includes:
The control device updates the first determination criterion using machine learning.
The control device according to claim 1,
The first diagnosis unit includes:
The control device, wherein the first determination criterion is updated using a dimension reduction process.
The control device according to claim 1,
The first diagnosis unit includes:
The control apparatus updates the first determination criterion using a compression process.
The control device according to claim 1,
The first diagnosis unit includes:
The control device updates the first determination criterion using statistical processing.
The control device according to claim 1,
The control change unit
An abnormality is reported when the diagnosis result of the first diagnosis unit is abnormal and the diagnosis result of the second diagnosis unit is normal.
The control device according to claim 6,
The first diagnosis unit includes:
Clustering the target data using machine learning,
The control change unit
When the diagnosis result of the first diagnosis unit is abnormal and the diagnosis result of the second diagnosis unit is normal, the target data at the second timing is changed to data of the nearest cluster. Control device.
The control device according to claim 1,
The control change unit
When the diagnosis result of the first diagnosis unit is abnormal and the diagnosis result of the second diagnosis unit is abnormal, the control unit is caused to execute fail-safe processing or the control unit is stopped. Control device.
The control device according to claim 1,
The target data is
The data indicated by the input signal,
The first diagnosis unit includes:
Clustering the target data using machine learning,
The control change unit
When the diagnosis result of the first diagnosis unit is abnormal or the diagnosis result of the second diagnosis unit is abnormal, the target data at the second timing is changed to data of the nearest cluster, or the control unit The control device characterized by stopping.
The control device according to claim 1,
The controller is
A control device that controls an internal combustion engine by calculating an output signal used for controlling an air amount, a fuel injection amount, or an ignition timing of the internal combustion engine.
The control device according to claim 1,
The controller is
A control device that calculates an output signal used to control a tension, a rolling position, or a rolling material moving speed of a rolling device, and controls the rolling device.