JPH0293903A - Abnormality processing system for fuzzy controller - Google Patents

Abstract

PURPOSE:To optimize the control of abnormality and to easily execute the investigation or the like of a new rule by deciding the existence of abnormality when dissociation with the prescribed number of labels exists, deleting each label when the output value of the label exceeds a reference value and inhibiting the output of a fuzzy inference value when there exist plural output values more than the reference value. CONSTITUTION:Since abnormality is detected by deciding whether dissociation is more than the prescribed number of fuzzy labels or not when the fuzzy output of each rule has probability more than a certain reference value, abnormality detection can be executed by an extremely simple constitution. Whether the output of each label out of plural labels exceeds the reference level or not at the time of detecting abnormality, the label output is neglected when it is less than the reference value, and no control output is generated in a solution as no solution when there are several labels exceeding the reference value. Consequently, output evasion or optimum control at the time of abnormality can be attained.

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Field of Industrial Application The present invention relates to a method for handling an abnormality in a fuzzy controller used for fuzzy control.

(B) Conventional technology As a processing method for inference output abnormalities in conventional fuzzy controllers, the fuzzy output with multiple prominences associated with contradictory rule settings, etc. is judged to detect abnormality and to analyze the situation at that time. There are things you can memorize so you can check them later. However, with this method, is there a limit? ff1l
When an abnormality is detected during operation, the output from the fuzzy controller to the controlled object is performed in the same way as during normal operation.

(C) Problems to be Solved by the Invention The conventional abnormality processing method described above has an abnormality detection function, but since the output to the controlled object remains unchanged, there is no workaround when an abnormality occurs. Furthermore, the anomaly detection method is also complicated, leading to increased costs. Furthermore, there was a problem in that control under contradictory conditions required time to confirm abnormalities and consider new rules.

On the other hand, in the case of fuzzy control, how to perform the control?
′ control knowledge is often obtained from experienced engineers who operate the controlled object. However, when operating know-how is collected from multiple veterans, individual differences may emerge, and the best way to determine which one is best is to actually try controlling it based on that knowledge. Also, -
Knowledge gained from other experts may also be wrong.

Furthermore, there are cases where the acquired knowledge is used incorrectly when setting it up. In this way, when an abnormality in inference calculations due to fuzzy control rule settings or a situation in which it is unclear which is the correct answer occurs, it is necessary to be able to detect and deal with the situation. However, conventional abnormality handling methods have not been able to adequately meet this need.

The present invention was made in view of the above-mentioned problems, and provides an abnormality processing method for a fuzzy controller that optimizes control in the event of an abnormality with a simple configuration and facilitates consideration of new rules. It is intended to.

(d) Means and Effects for Solving the Problems In the abnormality processing method of the present invention, it is determined whether each rule output of the fuzzy inference calculation section is equal to or greater than a predetermined number of labels. For example, in the rule output, if the labels are extremely far apart like NL and PM, it is determined that there is an abnormality. If there is an abnormality, it is determined for each label whether the output value is greater than or equal to the reference value. If the value is less than the reference value, the label is deleted from the rule output.

If there are still multiple labels remaining before or after deletion, and both exceed the standard level, there is no solution.
This time, we will not output fuzzy output due to this rule.

(E) Examples The present invention will be explained in more detail with reference to Examples below.

FIG. 1 is a block diagram of a fuzzy controller in which the present invention is implemented. The fuzzy controller 10 includes a fuzzy inference calculation unit 11, a control rule input unit 12 that inputs a control rule (control rule) used for inference calculation in the fuzzy inference calculation unit 11, a control storage unit 13 that stores the control rule, and members. a ship function input section 14, a membership leap number storage section 15, a measured value input section 16 that inputs an input signal, such as a deviation or the amount of change in the deviation, to the fuzzy inference calculation section 11;
An inference abnormality input unit 17 that inputs label differences and level reference values for each label for determining an inference abnormality; The detection and output section is composed of a control situation display section 19 that displays the control situation in real time, and an inference abnormal situation storage section 20 that stores the abnormal situation at the time of an inference abnormality. Further, the abnormal situation stored in the inference abnormal situation storage section 20 is displayed on the inference abnormal situation display device 21. Which point in time the abnormal situation is to be displayed is specified by a command from the normal situation display command device 22 (↑1). The inference abnormal situation storage unit 20 stores data as shown in FIG.

Note that the inference abnormal situation display unit 20, the inference abnormal situation display device 21, and the inference abnormal situation display command device 22 may be realized by a personal computer. In that case, the inference anomaly detection and output unit and the Persol computer are connected via a communication line.

When controlling the controlled device (plant) 30 using the fuzzy controller 10 as shown in L,
The connections are made as shown in FIG.

FIG. 4 shows membership functions used in the fuzzy controller 10. In the figure, NL is negative and very large, NM is negative and quite large, NS is negative and somewhat large, ZR is zero, PS is positive and somewhat large, PM is positive and quite large, and PL is positive and very large. The large numbers are shown respectively. In addition,
Here, the label NL is (1), and the following numbers are the labels P.
Label numbers are attached up to L(7).

It is assumed that the control rule storage unit 13 stores the following rules.

(1) If Input A = PM, Then Output: IQ = PM (2) [Input A =' PM & Human power B = ZR1Th
en Output Q=PS (n) I f Input C=NS & Input D=NL, Th
en Output Q=NS Suppose that inputs A, B, C, and D shown in FIG. The output for each rule is
As shown in FIG. 6(A), FIG. 6(B), and FIG. 6(C). When these respective output values are combined without being corrected by abnormality processing, the output becomes as shown in FIG. 6(D).

Next, the processing operation of the inference anomaly detection and output unit 18 will be explained with reference to the flowchart shown in FIG.

First, the fuzzy inference calculation unit 11 receives an output for each rule obtained by applying the rule to each input value, and stores the numbers of all rule output labels (step 5TI). next,
Subtract the minimum label number from the maximum label number for each rule,
It is determined whether the difference label value is 2 or more (step 5T2).

If the calculated label difference value is not 2 or more, it is assumed that the output is relatively appropriate, and if the judgment is NO in step ST2, the process moves to step ST3, where the fuzzy outputs of the output label numbers in that rule are combined, that is, logically summed. In step 5T4), the controlled device 30 is de-fuzzyed.
Outputs operation signals to. Furthermore, the fuzzy output in step ST3 is transferred to the control status display section 19 and displayed.

Next, as a result of the determination in step ST2, if the difference value between the maximum label number and the minimum label number is 2 or more, it means that outputs with very far apart label numbers exist according to one rule, and in this case, it is abnormal. It is determined that there is an abnormality processing notification (step 5T5), and then it is determined whether the fuzzy output level of the maximum label number is equal to or higher than the reference value (step 5T6).
, if the fuzzy output of the maximum label number is equal to or higher than the reference level, this determination becomes YES, and then it is determined whether the level of the fuzzy output of the lowest label number is equal to or higher than base YW (a (step 5T7).Step ST6 And in step ST7, if the level of either the fuzzy output with the maximum label number or the fuzzy output with the minimum label number is below the reference value, the determination in each step becomes NO, and the fuzzy output below the reference value is deleted. , the output label number is also deleted (step 5T9), and step ST
Return to 2. As a result, fuzzy outputs that are less than the reference value are ignored, and in the following processing, the processing from step ST2 onwards will be similarly performed for other fuzzy labels. For example, if there are fuzzy outputs corresponding to two label numbers with a relatively large label number difference per rule, and one fuzzy output is less than the reference value, that label will be ignored and the other The fuzzy output of the label is de-fuzzyed and output to the controlled device 30. therefore,
In this case, even though abnormal processing is detected,
A relatively appropriate operation output can be produced.

Next, in step ST6 and step ST7, if the levels of both the fuzzy outputs of the maximum label number and the minimum label number are both greater than or equal to the base j'P value, it is determined that there is no solution, and pseudo data is set. 5T8), no output is made at this sampling time. Therefore,
Controlled device 30 subjected to defuzzification processing and output
There is no output to , that is, no fuzzy output at the current sampling time.

(f) Effects of the Invention According to this invention, abnormality detection is performed by determining whether or not the fuzzy output of each rule has a probability greater than a certain base value, but the deviation is greater than or equal to a predetermined number of fuzzy labels. , since it detects an abnormality, it can detect an abnormality with an extremely simple configuration, and can reduce costs compared to conventional systems. In addition, when an abnormality is detected, depending on whether the output of each of multiple labels exceeds the standard level, if the label output is below the standard value, it will be ignored, and if the label output exceeds the standard value, it will be ignored. If there are several labels, it is assumed that there is no solution and the control 'r:fff output for that solution is not output, so that output can be avoided or optimal control can be performed in the event of an abnormality. Furthermore, since it is equipped with a storage means for storing the abnormal situation and can check it later, there are other advantages such as it is easy to check and study contradictory rules.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a fuzzy controller in which the present invention is implemented, FIG. 2 is a diagram showing a data map that stores inference abnormal situations of the fuzzy controller, and FIG. FIG. 4 is a diagram showing an example of the membership function, and FIG. 5 is a diagram showing the membership function when a predetermined input is applied in the fuzzy controller of the embodiment. Diagrams showing the relationship, Figure 6 (A), Figure 6 (B
) and FIG. 6(C) respectively show the membership function of the rule that obtained the output when a predetermined rule is applied according to the input amount input in FIG. 5, and FIG. D
) are shown in Figure 6 (A), Figure 6 (B) and Figure 6 (C) above.
FIG. 7 is a diagram illustrating the membership function obtained by combining the respective outputs of . 10: Fuzzy controller, 11: Fuzzy inference calculation section, 17: Inference abnormality judgment criterion input section, 18: Inference abnormality detection & output section, 20: Hl: Theory abnormality situation storage section, 30: Control target device.

Claims (1)

[Claims] (1) For each rule output of the fuzzy inference calculation unit, determine whether there is a discrepancy of more than a predetermined number of labels, and if there is a discrepancy of more than a predetermined number of labels, it is determined that there is an abnormality. Determine whether the output value is greater than or equal to the reference value, and if there is an output value less than the reference value, delete that label, and if there are multiple output values greater than or equal to the reference value, output the fuzzy inference value. An abnormality processing method for a fuzzy controller that is characterized by preventing