JPH04338828A - Fault diagnostic system - Google Patents

Abstract

PURPOSE:To effectively perform the diagnoses for study of the causes of the faults of equipments and the diseases by using the diagnostic knowledges acquired with a 'diagnositic knowledge construction system'. CONSTITUTION:A check instruction part 3 selects a set of causes and symptoms sorted out of a cause group shown by the identification number received from an assurance calculation part 6 by reference to a diagnostic knowledge store part 2. The selected set of causes and symptoms is sent to a check part 4. The part 4 carries out the check and repeats the calculation of assurance through the part 6 and the extraction of a convincing cause group of a fault. Then the part 3 completes a diagnosis as long as the set of causes and symptoms received from the part 6 is identical with the finally sorted set, that is, if the further sorting of the set is impossible. If the cause group of the set of causes and symptoms is identical with the cause group of a relevant fault, the diagnostic result, i.e., the studied cause group is displayed at a display part 8. In such a way, the cause of a fault is automatically obtained.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for diagnosing disorders of various devices and bodies.

0002

[Background Art] In recent years, in the information processing and communication industries, as well as medicine, diagnostic expert systems have begun to be constructed that convert the diagnostic knowledge actually thought by experts into data and use it to diagnose disorders and diseases. The importance of these expert systems is already clear.

[0003] In order to construct the diagnostic knowledge referred to in conventional fault diagnosis methods, experts in diagnosing the corresponding fault etc. express their own diagnostic procedures in flowcharts, and then an expert system developer implements the expert system. The system was manually converted into a language that can be used by the equipment used, and a separate system was developed to interpret this manually converted diagnostic knowledge and perform diagnosis. Therefore, there is no example of a system that automatically executes a series of processes from constructing diagnostic knowledge to diagnosis and investigating the cause after inputting a simple correspondence between causes and symptoms.

0004

[Problem to be solved by the invention] There is a large gap between the words and expression formats used for diagnosis by experts in each field and the programming languages and expression formats used for knowledge processing, and the conversion work, that is, the diagnosis expert system. Building diagnostic knowledge is technically complex and requires a large amount of time. Furthermore, we have separately developed a system that uses this constructed diagnostic knowledge to diagnose disorders and diseases. Knowledge construction and diagnostic system development are performed in separate processes, and if the form of diagnostic knowledge is different, new systems can be developed. This has caused inconveniences such as the need to construct a diagnostic system.

[0005] An invention that solves the above-mentioned problems in knowledge construction is the invention "diagnostic knowledge construction method" which was filed for patent on the same day as the present application by the same applicant. We propose a method for performing diagnosis using the diagnostic knowledge created by the method. Therefore, since the knowledge used for diagnosis according to the present invention is always in the same format, problems in redevelopment and maintenance of the diagnostic system due to differences in knowledge format are resolved, and furthermore, it is possible to easily deal with causes and symptoms. This makes it possible to automate a series of processes from identification to cause investigation.

[0006]

[Means for Solving the Problems] The disorder diagnosis method of the present invention shows the presence or absence of a relationship between individual causes and symptoms, or the strength of the relationship, in a two-dimensional table centered on cause groups and symptom groups of disorders. Weights are assigned to extract cause-symptom pairs that have a relationship between some of the cause groups and some of the symptom groups in the table, and all causes in the table are related to some cause-symptom pairs. The causes are classified by forming pairs of causes and symptoms that belong to the group, and a reduced table is created by removing related symptom groups from the table consisting of the classified cause groups, and this table is used. Furthermore, in a method of diagnosing a disorder using diagnostic knowledge obtained by repeatedly classifying causes by extracting pairs of causes and symptoms based on the relationship between causes and symptoms described above,

[0007]
For the cause groups classified from the same cause group, tests are conducted to confirm the presence or absence of symptom groups related to the cause group in the order of classification, and the certainty of the cause group is determined by the presence or absence of each confirmed symptom. The method is characterized in that the cause of the failure is investigated by repeatedly determining the causes and selecting the group of causes with the highest probability.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Examples of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an embodiment of the fault diagnosis method of the present invention. In Figure 1, the diagnostic knowledge construction unit 1 repeatedly extracts pairs of causes and symptoms that have a strong relationship from a two-dimensional table with causes and symptoms as axes. It has the function of constructing diagnostic knowledge consisting of sets of . The function of constructing diagnostic knowledge consisting of sets of causes and symptoms from a two-dimensional table is described in detail in the aforementioned "Diagnostic Knowledge Construction Method" by the same applicant.

The diagnostic knowledge storage unit 2 stores sets of cause groups and symptom groups extracted by the diagnostic knowledge construction unit 1. In the diagnostic knowledge storage unit 2, relationships between causes and symptoms are generally described in the format ((N)/(X)/(Y)). here,
(N) indicates the identification number of the cause-symptom pair, and also represents the order in which the pair was extracted. Further, (X) indicates a cause group, and (Y) indicates a symptom group. For example, in general, as an example of a cause-symptom pair, ((i, j, ~, k
,l)/(xi+1,xi+2,~,xi+a)
Knowledge is accumulated in the following format: /(yj+1,yj+2,~,yj+b)). That is, ((i, j, ~, k
,l)/(xi+1,xi+2,~,xi+a)
/(yj+1 ,yj+2 ,~,yj+b )) is (
(i, j, ~, k)/(X)/(Y)) is extracted, and then subdivided within the range of cause group (X), resulting in the L-th subset. At this time, the cause group (X) is the cause group (xi+1
, xi+2 , ~, xi+a ). (yj+1, yj+2, ~, yj+b) represents a symptom group that is all related to the cause group (xi+1, xi+2, ~, xi+a).

In FIG. 1, the test instruction unit 3 selects and reads pairs of causes and symptoms from the diagnostic knowledge stored in the diagnostic knowledge storage unit 2 in descending order of identification number. The test instruction unit 3 performs the following for each symptom of the selected group of symptoms:
The inspection unit 4 instructs the inspection unit 4 to perform a test to confirm the presence or absence of the symptoms.
give instructions to The set selected by the test instruction unit 3 is a set in which the cause groups constituting the set are extracted from the same cause group. In other words, it selects the set 1 = 1, 2, -, p with identification numbers (i, j, ~, k, l), and instructs to perform a test for the presence or absence of the symptom group of each group. .

The test section 4 refers to the test method description section 5 for the test method for the symptom group received from the test instruction section 3, and performs the test according to the test method. The test method description section 5 describes methods for testing individual symptoms and their occurrence. For example, the symptom group (y1, y2, ~, y
m), test equipment that automatically performs inspections, program names and commands for running test programs,
Program names and commands for displaying questions to the troubleshooter or test instructions (c1, c2, c
m) format. The inspection unit 4 receives an inspection command ci that is instructed by the inspection instruction unit 3 and corresponds to the symptom yi.
The corresponding inspection function prepared in the inspection section 4 is operated using the . The inspection command ci corresponds to a call statement that starts an inspection program prepared as a so-called subroutine. The testing unit 4 receives test results indicating whether or not the corresponding symptoms have occurred from the testing function that issued the instruction to perform the test.

The testing section 4 sends the test results to the certainty factor calculation section 6 for each set of causes and symptoms. The information sent from the inspection unit 4 to the certainty calculation unit 6 is, for example, ((i, j, ~
,k,l)/(xi+1 ,xi+2 ,~,xi+a
)/(yj+1 , yj+2 , ~, yj+b))/
The information is in the format (○, ×, ~, ○)), and the identification number of the cause-symptom pair, the cause group, the symptom group, and the presence or absence of each symptom are indicated by ○×.

The certainty calculation unit 6 calculates the certainty of the cause-symptom pair indicated by the identification number (i, j, -, k, l) based on the presence or absence of each symptom. The confidence calculation unit 6
With reference to the cause/symptom relationship table 7, the certainty factor is calculated for each cause included in the cause/symptom pair received from the testing unit 4, and the average of the certainty factors for all causes in the group is calculated. Cause/symptom relationship table 7, as shown in Figure 2, is a two-dimensional table with causes and symptoms as axes, and the columns corresponding to cause i and symptom j in the table are weighted with Wij. and
If it is unrelated, it is 0. At this time, the certainty factor CF(xi+k) of cause xi+k is, for example, 1=>Wij>
Weighted as =0, 1-(1-Wi+kj+1*n
)*(1-Wi+kj+2*n)*~*(1-Wi+k
The confidence level is determined by the calculation (j+b*n). Here, n is 1 if the corresponding symptoms exist, and 0 if there are no symptoms.
It is. Furthermore, the certainty calculation unit 6 calculates a set of causes and symptoms (
(i, j , ~, k, l)/(xi+1 , xi+2
, ~, xi+a )/(yj+1 ,yj+2 , ~,
Confidence CF(i,j,~,k,l) of yj+b))
is (CF(xi+1)+CF(xi+2)+～+
It is calculated by the calculation CF(xi+a))/a. The confidence calculation unit 6 calculates the confidence of the calculated cause and symptom pair CF(i,j,~,k,l), CF(i,j,~,
k, 2), ~, CF(i, j, ~, k, p), extract the set with the highest confidence and use its identification number (i, j, ~
, k, l) to the inspection instruction section 3.

As described above, the test instruction unit 3 refers to the diagnostic knowledge storage unit 2 and selects the cause and symptom pair classified from the cause group indicated by the identification number received from the certainty calculation unit 6. Then, the selected cause and symptom set is sent to the inspection section 4, and the inspection section 4 executes the inspection, the confidence calculation section 6 repeats the confidence calculation, and the extraction of the cause group that is most likely to be the cause group of the failure. In this way, if the cause and symptom pair received by the test instruction unit 3 from the certainty calculation unit 6 is a finally classified group, that is, a cause that cannot be further classified, the diagnosis is completed and the cause and symptom pair is classified. The cause group is the cause group sought for the target disorder, and the display section 8 shows the diagnosis result.
In other words, the group of causes that have been investigated is displayed. In this way, the cause of the failure can be automatically determined.

[0015]

[Effects of the Invention] As explained above, the fault diagnosis method of the present invention utilizes the diagnostic knowledge obtained in the above-mentioned "diagnostic knowledge construction method" to investigate the causes of equipment failures and diseases. This has the effect of allowing efficient diagnosis.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of a fault diagnosis method of the present invention.

FIG. 2 is a diagram showing the relationship between causes and symptoms in a two-dimensional table format.

[Explanation of symbols]

1 Diagnostic knowledge construction department 2 Diagnostic knowledge storage section 3 Inspection instruction department 4 Inspection Department 5 Inspection method description section 6 Confidence calculation unit 7 Cause/symptom relationship table 8 Display section 20 Cause/symptom relationship table

Claims (1)

[Claims] Claim 1: A two-dimensional table centered on cause groups and symptom groups of disorders is given a weight indicating the presence or absence of a relationship between each cause and symptom, or the strength of the relationship, and Extract cause-symptom pairs that are related to some of the groups in the table, and form cause-symptom pairs such that all causes in the table belong to some cause-symptom pair. By classifying the causes, a reduced table is created by removing related symptom groups from the table consisting of the classified cause groups, and this table is used to further identify the causes based on the relationship between the causes and symptoms described above. In a method of diagnosing disorders using diagnostic knowledge obtained by repeatedly classifying causes by extracting sets of symptoms, for cause groups classified from the same cause group, cause groups are classified in the order of classification. By repeatedly carrying out tests to confirm the presence or absence of symptom groups related to symptoms, determining the probability of the cause group based on the presence or absence of each confirmed symptom, and selecting the cause group with the highest probability. , a fault diagnosis method characterized by investigating the cause of the fault.