JPH0831199B2 - Development support device for abnormality diagnosis system - Google Patents

Description

Description: [Industrial field of use] The present invention relates to a support device applied to the development of an abnormality diagnosis system for plants and machinery.

[Prior Art] In the fault diagnosis of plants and machinery, various kinds of element methods such as upper and lower limit check method, balance check method, redundant signal comparison method, model comparison method, knowledge engineering method have been developed and applied. There is. However, there was no objective criterion as to which elemental method could be used to make the required diagnosis with the required speed and certainty, and the choice was entirely left to human judgment.

 Next, the outline of the conventional diagnostic element technology will be described.

(1) The upper and lower limit check method is a method of detecting an abnormality by monitoring that a process amount such as pressure or temperature measured by a plant or machinery deviates from a predetermined normal range. Not the process amount itself,
There are also improved methods such as monitoring the rate of change.

(2) The balance check method is to detect abnormalities by observing whether or not a plurality of related variables such as mass balance, energy balance, etc., in which the inflow rate and the outflow rate must be balanced, are balanced. This is a method for detecting and identifying the location of an abnormality. Normally a static balance check is done.

(3) Redundant signal comparison method obtains multiple redundancy and signals for one quantity by measuring the same physical quantity with multiple measuring instruments or calculating the same calculation with multiple different calculation methods. , It is a method of diagnosing abnormalities in measuring instruments, calculators and algorithms by comparing these.

(4) The model comparison method has a model that shows the behavior of a plant or machinery in a normal state, and diagnoses an abnormality by monitoring the difference between the calculated physical quantity and the calculated physical quantity. Is a method of performing. A dynamic characteristic model is usually used as the model.

(5) The knowledge engineering method is to computerize knowledge used by an abnormality diagnosis expert in abnormality detection, cause identification, and countermeasure determination, and a thinking method using the knowledge in the form of a knowledge base and an inference program, respectively. This is a method for realizing the expert's diagnosis process by computer.

[Problems to be Solved by the Invention] The conventional drawbacks are listed below.

(1) The required speed cannot be obtained because an inappropriate method is selected.

(2) A wrong diagnosis is made because an inappropriate method is selected.

(3) Since an appropriate method cannot be selected, there is a possibility of causing serious damage or an accident due to failure to detect an abnormality, identify a cause, or determine a countermeasure.

[Means for Solving Problems] The present invention uses an upper and lower limit check method as a diagnostic element method,
The balance check method, the redundant signal comparison method, the model comparison method, and the knowledge engineering method are taken up, and knowledge that associates the characteristics of each method with both the physical characteristics of the diagnostic target and the diagnostic event and the combination of the element methods We will support the development of an optimal abnormality diagnosis system by using inference technology of knowledge engineering while making the knowledge base of integration into a knowledge base and interacting with the user.

[Operation] Applicability of each method can be appropriately judged by the above knowledge, and optimal combination and integration can be made by the knowledge of the above knowledge base. It has the effect of providing moderate support.

[Embodiment] 1) Device Configuration An embodiment of the present invention is shown in FIG.

A central processing unit (hereinafter abbreviated as CPU) 1 is a database 2
Via the shared memory 3, the input device 4 and the output device 5
Connected with. Further, a distributed processing unit (hereinafter abbreviated as DPU) 10, 20, 30, 40, and 50 is connected to the data bus 2, and each DPU stores a memory (MA) for storing knowledge about characteristics of the diagnostic method. 11,21,31,41 and 51 and memory for accumulating past usage experience (MB) 12,22,32,4
2 and 52 are connected. To the CPU 1, a memory (MC) 6 for storing knowledge about integration of methods is connected.

The CPU 1 inputs data from the input device 4 via the data bus 2 and issues a processing start command to each DPU 10, 20, 30, 40 and 50 via the data bus 2. Each DPU 10,20,30,40 and 50 receives this instruction and receives memory 11,21,31,41 and 5 respectively.
Knowledge base stored in 1, and memory 12,22,32,
Inference processing is performed using the experience data stored in 42 and 52, and the respective processing results are output to the shared memory 3 via the data bus 2. The CPU 1 integrates the contents of the shared memory 3 with reference to the knowledge base stored in the memory 6, outputs the integrated result to the output device 5 via the data bus 2, and simultaneously writes it to the shared memory 3.
The respective DPUs 10, 20, 30, 40 and 50 refer to this to update the contents of the memories 12, 22, 32, 42 and 52 storing the usage experience, respectively.

2) Processing contents of inference mechanism The inference mechanism is, as shown in FIG.
And five applicability determination units 110, 120, 130, 140 that determine the applicability of the upper and lower limit check method, the balance check method, the redundant signal comparison method, the model comparison method, and the knowledge engineering method.
And 150, which are integrated via a working memory area 103 which stores the processing results of the respective parts.

The processing of the inference supervision section is executed on the CPU 1 using the knowledge in the knowledge base memory 6. Five applicability judgment units 110, 12
Knowledge base memory for processing 0, 130, 140 and 150 respectively
Knowledge in 11, 21, 31, 41 and 51 and experience data memory 12,
DPU10,20,30,40 using data in 22,32,42 and 52
And run on 50.

The working memory area 103 is provided on the shared memory 3.

(1) Processing of the inference supervision section The inference supervision section 101 executes the following processing upon activation of this apparatus.

(1.1) An instruction to execute the applicability judgment process described in (2) is issued to each applicability judgment unit 110, 120, 130, 140 and 150.

(1.2) Check whether or not the applicability judgment processing of all element methods has been completed, and if so, jump to step (1.6) described later.

(1.3) If it is not completed in the above step (1.6), check whether there is a question request (step (2.4) described later) from each applicability judgment unit, and if there is no question request, the above step Return to (1.2).

(1.4) If there is a question request, the question is posed through the output device 5 and the answer is received through the input device 4.

(1.5) Write the answer in the working memory area 103.

(1.6) Referring to the degree of applicability of each element method written in the working memory area 103 according to the method integrated knowledge base, a draft of the integrated abnormality diagnosis system is created in the working memory area 103.

(1.7) The draft created in step (1.6) above is output via the output device 5, and at the same time the user is asked whether or not the draft needs to be modified.

(1.8) The necessity of the correction and the correction item when the correction is necessary are input through the input device 4.

(1.9) The final plan with necessary corrections is recorded in the working memory area 103.

(1.10) The final plan created in step (1.9) above is output via the output device 5.

(1.11) Finally, the applicability determination units 110, 120, 130, 140, and 150 are issued an execution command for the experience data update process.

(2) Process of Applicability Judgment Unit The applicability judgment units 110, 120, 130, 140, and 150 of the respective diagnostic element methods have the same processing contents except that the knowledge base and empirical data to be used are different.

(2.1) When an execution command is received from the inference control unit 101, the process branches to step (2.2) or (2.8) described later depending on whether it is the applicability determination process or the experience data update process.

(2.2) Applicability judgment is performed using the applicability judgment knowledge base for three purposes: abnormality detection, cause identification, and countermeasure determination.

(2.3) First, take out the condition items for judging applicability.

(2.4) If there is a condition for which no value is given,
A query request is issued to the inference control unit 101 via the working memory area 103 to obtain a value.

(2.5) If values are given to all conditions, knowledge of how to determine applicability is used to detect abnormality in each element method,
Determine the applicability for each cause identification and countermeasure decision. Here, the degree of applicability is +1.0 is the most appropriate, -1.0 is the most unsuitable, 0 cannot be determined, and the value between +1.0 and -1.0 is appropriate / unsuitable according to its size. Shall be shown.

(2.6) The past experience data is retrieved from the experience data memory.

(2.7) The applicability determined in step (2.5) above and the experience data extracted in step (2.6) above are written to the working memory area 103, and the applicability determination process ends.

(2.8) Refer to the final plan in the working memory and update the number of times each element method is used and the number of combinations with other methods.

3) Configuration example of knowledge base The knowledge base is a knowledge base that determines the applicability of the upper and lower limit check method, the balance check method, the redundant signal comparison method, the model comparison method, and the knowledge engineering method, and combinations of these methods. It consists of a knowledge base on integration of methods.

(1) Applicability judgment knowledge base Tables 1 to 5 show examples of applicability judgment knowledge base of each diagnostic element method. This knowledge base consists of knowledge about the following:

A) Conditional items for judging applicability of elemental methods. These are categorized into the prerequisites for applying the method, the usage environment of a computer, the measurement of necessary data, and the definition of normal values that are the basis of abnormality diagnosis. It

B) A method of determining the applicability of the method by combining the condition items in the previous section. This method consists of a decision condition for each purpose of abnormality detection, cause identification, and countermeasure determination, and a pair of applicability when the condition is satisfied. More than one is given. However, the determination conditions are determined in order from the top, and if one is found to be satisfied, the applicability of the pair is set as the determination result, and the processing ends. That is, the determination conditions below that are ignored.

(2) Method integrated knowledge base Table 6 shows an example of the method integrated knowledge base. This knowledge base is the knowledge for integrating the results of determining the applicability of each element method using the knowledge base described in the above item (1).

4) Contents of Working Memory Area The working memory area 103 is composed of a question response data section 201 and an integrated system data section 202 as shown in FIG.

(1) As shown in FIG. 3, the question / answer data section 201 is capable of storing a plurality of pairs of question data and answer data. The question data has a value and the answer data has a value. The question is not present when a question is requested, and the answer data has a value when a response to the question is given.

(2) The integrated system data section 202 is further divided into a rough draft section 203 and a final draft section 204, each having the same data structure.

 The data structure is as shown in Fig. 3. a) Method name used for abnormality detection, degree of application and number of uses b) Method name used for cause identification, degree of application and number of uses c) Method name used for determining countermeasures And the applicability and the number of times of use can be stored.

[Effects of the Invention] The following effects are achieved by using the development support device for the abnormality diagnosis system according to the present invention described above.

(1) Applicability of each diagnostic element method and their combination
By performing computerized automation processing using the knowledge about integration, it is possible to develop a diagnostic system with high diagnostic accuracy and high speed.

(2) When a new method is devised, it is possible to easily expand the function of the apparatus simply by creating an applicability judgment knowledge base for the method and adding a part of the integrated knowledge base.

(3) Even when the applicability judgment result of some methods or the integration result of the methods is not preferable, the related knowledge is localized in the applicability judgment knowledge base or integrated knowledge base of each method and maintained. Easy to improve.

Data for this maintenance / improvement can be retrieved from the experience data memory.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a system configuration in an embodiment of the present invention, FIG. 2 is a block diagram showing a configuration of an inference mechanism in the above embodiment, and FIG. It is a figure which shows the structure of the working memory area in the said Example. 1 ... Central processing unit (CPU), 2 ... Data bus (DATA
-BUS), 3 ... Shared memory, 4 ... Input device, 5 ...
Output device, 6 ... Knowledge-based memory (MC), 10, 20, 30, 4
0,50 …… Distributed processor, 11,21,31,41,51 …… Knowledge base memory (MA), 12,22,32,42,52 …… Experience data memory (MB), 101 …… Inference control Part, 103 ... Working memory, 110 ... Upper / lower limit check method applicability judgment part, 120 ... Balance check method applicability judgment part, 130 ... Redundant signal comparison method applicability judgment part, 140 ... Model comparison method application part Sex determination part, 1
50 …… Knowledge engineering method applicability judgment part, 201 …… Question answer data part, 202 …… Integrated system data part, 203 …… Draft part, 204 …… Final draft part.

Claims (1)

[Claims] 1. An upper and lower limit check method, a balance check method,
The same number of first storing means as the number of the above-mentioned element methods storing the knowledge about the applicability of the abnormality diagnosing element methods such as the redundant signal comparing method, the model comparing method and the knowledge engineering method, and the first storing means are stored in these storing means. The same number of distributed processing devices as the number of the above-mentioned first storage means for calculating the application effectiveness of each element method using knowledge, and the above-mentioned first storage means for storing the usage experience of each method The same number of second storage units, a central processing unit that integrates the calculation results in the distributed processing unit, an input unit that inputs information to the central processing unit, and a processing result of the central processing unit. An apparatus for supporting the development of an abnormality diagnosis system, comprising: an output device for outputting information; and a bus for transmitting and receiving information between the above devices.