JPH05342004A - Diagnostic device - Google Patents

Abstract

(57) [Summary] [Purpose] It is possible to explicitly specify the order of efficient diagnosis in the form of knowledge, which facilitates the description of knowledge and realizes more flexible processing. [Configuration] A knowledge base 1 in which knowledge for diagnosis is stored,
It has a diagnostic engine 2 that makes a diagnosis based on the diagnostic knowledge stored in the knowledge base 1, and a user interface 3 that functions as an interface for the operator in the diagnosis by the diagnostic engine 2, and is stored in the knowledge base 1. The knowledge for diagnosis that is related to causality is separated from the knowledge that determines the procedure of diagnosis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a diagnostic device for diagnosing a failure or the like using a knowledge base.

[0002]

2. Description of the Related Art Generally, in a failure diagnosis type expert system, diagnosis uses knowledge based on a causal relationship between a cause of a failure and a phenomenon caused by the failure, and a possible cause of failure is observed from an observed phenomenon. And in many systems of this type the diagnostic procedure is determined solely by this causal relationship. That is, there is no clear distinction between the causal relationship and the knowledge about the control of the diagnostic procedure, and the causal relationship directly determines the diagnostic procedure. For example, the information about the diagnosis target that is required during the diagnosis is queried by the user as a question, but since the occurrence of this question depends on the progress of the diagnosis, the questions are generated in the order according to the above causal relationship. It had been. Further, in order to improve the efficiency of diagnosis, the knowledge of the diagnosis based on the causal relationship is changed so as to improve the efficiency of diagnosis.

[0003]

However, the efficiency of diagnosis considering many factors individually is determined by combining many factors such as the cost of inspection, the installation condition of the diagnosis target, and the treatment ability of the operator. However, there is a problem in that it is extremely difficult to determine the optimum diagnostic procedure considering these factors individually only by the causal relationship.

An object of the present invention is to provide a diagnostic device capable of efficiently and easily performing an optimal diagnosis in which many elements are individually considered.

[0005]

In order to achieve the above-mentioned object, the invention according to claim 1 separates knowledge on the causal relationship of a failure and knowledge on the procedure control of diagnosis. This makes it possible to explicitly specify an efficient diagnosis order in the form of knowledge, facilitate description of knowledge, and realize more flexible processing.

More specifically, the operator (expert) often sees the diagnosis as a procedural flow chart, and considers the factors that influence the efficiency, which are taken into consideration in constructing this flow chart. , It is often difficult to extract it as knowledge. Therefore, as in the invention according to claim 2, as the knowledge for determining the diagnosis procedure, the knowledge for controlling the display order of the questions to the operator is adopted, and the flowchart-like knowledge regarding the display order of such questions is adopted. By making the knowledge base explicit, it becomes possible to describe the order of diagnosis more directly.

In the invention according to claim 3, the invention according to claim 1
In the diagnostic device described, the knowledge of causal relationships is configured such that cause candidates with the same consequent event are treated as a cause candidate group, and the diagnostic engine has the potential to be based on information provided by the operator. Diagnosis is performed by repeating selection of a cause candidate group and ordering of each cause candidate in the selected cause candidate group. At this time, the display order of questions is controlled for each cause candidate group. Bringing knowledge to the knowledge base, controlling the questions that influence the ordering of the cause candidates within one cause candidate group based on the knowledge, and moving the diagnostic focus to other cause candidate groups in response to the question To control. That is, in order to efficiently use the above-mentioned flow chart type knowledge and causal relationship knowledge, a causal relationship is used to specify a cause candidate, that is, a focus of diagnosis is performed, and a flowchart is used as a procedure for ordering between the candidates. Type query order control knowledge is used. By associating the question order control knowledge with each cause candidate group unit, it is not necessary to consider the question order of the entire diagnosis, and it is possible to limit the description to the relationship between local questions.

Further, as in the invention described in claims 4 and 5, by combining the movement of the focus according to the certainty factor, the movement between the groups is controlled by the certainty factor, and the local question order control is performed by the flow chart type knowledge. It becomes possible to do it.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an embodiment of a diagnostic device according to the present invention, and here, it is assumed that the diagnostic device is a fault diagnostic device that diagnoses a fault. The diagnostic apparatus according to the present embodiment interprets the knowledge base 1 in which the diagnostic knowledge, which is a description about the diagnostic method, and the diagnostic knowledge in the knowledge base 1 are interpreted, and based on the interpretation, determines the question to the operator and operates it. A diagnostic engine 2 that estimates a possible cause of failure from a user's response, and a user interface 3 that manages display of questions and results to the operator and input of a response from the operator.

The diagnostic engine 2 is a focus determination mechanism 4 for determining the focus of the failure cause candidate based on the causal relationship of the failure and the information given by the operator about the ordering of the focused failure cause candidates. And a question display control mechanism 6 that determines the question to be displayed next to the operator based on the information given by the operator.

The knowledge base 1 also has failure causal relationship knowledge 7, certainty factor adjustment knowledge 8 and question display order control knowledge 9.

Here, the failure causal relationship knowledge 7 expresses a causal relationship between the phenomenon and the cause of the failure, and has a network structure as shown in FIG. Referring to FIG. 2, one failure phenomenon, for example, “engine start failure” is a result of some failure cause, for example, “electrical system failure”, and one failure cause, for example, “electrical system failure” is more fundamental. A failure cause, for example, “ignition system failure” is the result, and the failure phenomenon, the failure cause, and the root cause thereafter are collectively referred to as a failure event. Further, a pair of relationships L between the phenomenon of failure and its cause or its cause and more fundamental cause is called a causal link, and both ends of the causal link L are called a result side and a cause side, respectively.

The certainty factor adjustment knowledge 8 is knowledge expressing the strength of the probabilistic relationship of each causal link L. FIG. 3 is a diagram showing an example of certainty factor adjustment knowledge 8 regarding a causal link of “electrical system failure → ignition system failure” in FIG. 2 and the causal relationship network. Each causal link L is included in the certainty factor adjustment knowledge 8.
The default strength (default certainty factor) and the conditions for changing the strength (default certainty factor) and the magnitude of the change are included. The strength of each causal link is given by the operator. The amount of change determined based on the obtained information is added to the default certainty factor. In the example of FIG. 3, the causal link “defective electrical system → defective ignition system” has a default confidence of 30%, and when the operator's answer to the question “spark test” is “OK”, This strength is reduced by 20% to 10%. On the other hand, when the operator's answer to the question “plug state” is “dirty”, the default certainty factor of 30% is further increased by 10%, and the causal link strength becomes 40%.

Further, in the failure causal relationship knowledge 7, a plurality of causal links having the same failure phenomenon as a result and the failure event on the side of the cause thereof are regarded as one group, and this is called a cause candidate group. The diagnosis engine 2 is adapted to make a diagnosis by repeating the selection of a possible cause group based on the information given by the operator and the ordering of the respective cause candidates in the group. ,
Define the selected cause group as diagnostic focus. Here, one causal candidate group corresponds to one resulting failure event, so each group and diagnostic focus is indexed by the resulting failure event.

Further, the question display order control knowledge 9 is associated with each cause candidate group, and the question affecting the certainty factor of the causal link belonging to the group is, for example, FIG.
As shown in, the question is expressed as a question-answer network. The network of FIG. 4 is the same as the causal network of FIG.
An example of a question-> answer network of a causal candidate group having a result of is shown.

As can be seen from FIG. 4, the question-to-answer network includes six types of nodes, a root node RN, a question node QN, a response node AN, a condition node CN, and a diagnostic focus movement designation node DN, and links between them. And specifies the display dependency between questions, and is a description that determines the display of another question in response to the operator's response to a question. Here, the question node QN means the display of a question sentence, and the response node AN is connected thereafter. The answer node AN is a condition for answering a question, and a question connected after the answer node where the condition is not satisfied is not displayed. The condition node CN is a condition for the establishment / non-establishment of each parent, and includes an & node that is satisfied when all parents are satisfied, and an OR node (not shown) that is satisfied when any parent is satisfied. The question displayed is
The contents of all question nodes QN that can be reached by following each link from the root node RN and following the established condition node CN or response node AN. The diagnostic focus movement designation node DN designates movement of the diagnostic focus when this node is reachable.

Next, the processing operation of the diagnostic device having such a configuration will be described. FIG. 5 is a flow chart showing a processing flow of diagnosis by the diagnosis apparatus of this embodiment. Referring to FIG. 5, the diagnosis starts when the operator selects any of the failure phenomena on the result side of the causal relationship network (step S1). When one failure phenomenon is selected by the operator, the focus determination mechanism 4 of the diagnostic engine 2 sets the selected failure phenomenon as an index, and the causal link having the failure event set as the index on the result side and its link. Knowledge base of failure events on the causative side 1
The failure candidate causal knowledge 7 is listed, and this cause candidate group is determined as the diagnostic focus (step S
2). Then, with the failure event set as the index as a key, the question display order control knowledge regarding this cause candidate group, that is, the diagnostic focus is searched from the question display order control knowledge 9 of the knowledge base 1 (step S
3). When the question order control knowledge regarding the diagnostic focus is retrieved, the question display control mechanism 6 and the certainty factor calculation mechanism 5 of the diagnostic engine 2 ask the operator with the question display order control knowledge and certainty factor adjustment knowledge corresponding to the diagnostic focus. Is displayed, the order of cause candidates in the group is determined and displayed according to the response (step S4). In this process, when the question display control mechanism 6 gives an instruction to move the diagnostic focus (step S5), a new diagnostic focus is decided according to the instruction of the question display control mechanism 6 and its index is set (step S6). .

FIG. 6 is a flow chart showing the details of steps S4 and S5. Referring to FIG. 6, in the process of determining the cause candidate order in step S4, the question display order control knowledge is searched to determine a new question (step S4).
11). If this is the diagnostic focus movement designation (step S12), the process proceeds to step S6 as described above, but if it is not the diagnostic focus movement designation, the question determined in step S11 is displayed to the operator (step S6). In step S13, the operator inputs a response (step S1).
4). Based on the response, the condition of the certainty factor knowledge corresponding to each causal link is checked (step S15), and the certainty factor that satisfies the condition is changed (step S16). When the certainty factor of the cause candidate is changed, the result is displayed to the operator.
Further, simultaneously with the calculation of these certainty factors, the question display order control knowledge 9 is referred to by the response to the question, the question to be newly displayed is determined, and the question is presented to the operator. The repetition of such question display and certainty factor change is continued until the designation of the diagnostic focus shift in the question display order control knowledge is reached.

When the movement of the diagnostic focus is instructed in step S12, that is, step S5, the question display control mechanism 6 is operated in step S6 as described above.
The new diagnostic focus is determined and its index is set in accordance with the instruction. At this time, the instruction of the question display control mechanism 6 is as follows: type T1 "set failure event explicitly specified in knowledge as new index", type T2
The failure cause candidate having the highest certainty is set as a new index, and the type T3 is "return to the previous index". First type T1
Is a case where a new diagnostic focus is explicitly set by the question display order control knowledge, and the designated failure event is used as a new index. The second type T2 is a case where movement of the diagnostic focus based on certainty is specified by the question display order control knowledge, and one of the candidate causes of failure in the current diagnostic focus has the highest certainty. Select and set it as the new index. The third type T3 is to return to the diagnostic focus before the current diagnostic focus is set. In the example of FIG. 4, the diagnostic focus movement designation nodes for “to control system failure” and “to ignition system” are type T1, and the diagnostic focus movement designation node for “confidence” is type T2.
"Return" diagnostic focus movement designation node is type T3
Is.

Thus, in step S6,
When a new diagnostic focus is set, the process returns to step S3 again, and the same processing as described above is repeated. That is, the failure cause group and the question sequence control knowledge corresponding to the new diagnostic focus are searched, and the failure cause candidates in those groups are displayed as the failure cause candidate together with the default certainty factor. At this time, the question control order knowledge is searched, the reachable question node is determined from the already input information, and the cause is displayed and the question is output at the same time. When the user responds to one of the outputted questions, the certainty factor adjustment knowledge is searched based on the information, and the certainty factor of each failure cause candidate is changed. At the same time, the question control knowledge is searched again including the newly input answer, the question to be newly displayed is determined and output. When the diagnostic focus movement designation node of the cause group is reached when the question control knowledge is searched for a new response, a new failure result is set.

As described above, in this embodiment, the knowledge about the causal relationship of the fault and the knowledge about the procedure control of the diagnosis are separated, so that the efficient order of diagnosis can be explicitly specified in the form of knowledge. It will be possible. This facilitates the description of knowledge and enables more flexible processing. In particular,
Even when many factors such as the cost of the inspection, the installation condition of the diagnosis target, and the treatment ability of the operator are combined, it is possible to easily determine the optimum diagnosis procedure considering these factors individually.

[0022]

As described above, according to the invention described in claim 1, the diagnostic procedure can be specified as knowledge independently of the causal relationship. You can specify the diagnostic flow,
It is possible to efficiently and easily perform the optimum diagnosis considering many elements individually.

According to the second aspect of the invention, since the knowledge for controlling the display order of the questions to the operator is adopted as the knowledge for determining the diagnosis procedure, it is possessed by a human expert. Flowing knowledge of existing diagnostics can be more easily incorporated into the system.

According to the third aspect of the invention, the knowledge about the causal relationship is configured such that the cause candidates having the same result side event are treated as the cause candidate group, and the diagnosis engine is given by the operator. Based on the information, diagnosis is performed by repeating the selection of possible cause group and the ordering of each cause candidate in the selected cause group, and at this time, the question is displayed for each cause group. Put knowledge to control the order in the knowledge base,
Based on the knowledge, it controls the question that influences the ordering of the cause candidates in one cause candidate group, and further controls the shift of the diagnostic focus to another cause candidate group in response to the question. Therefore, it is not necessary to describe the knowledge of the question display order with the entire diagnosis as one flow, and the descriptiveness of the knowledge and the ease of acquiring the knowledge can be improved.

According to the inventions of claims 4 and 5,
Knowledge about the certainty factor representing the likelihood of each cause candidate in the cause candidate group is also given to the knowledge base, and the movement of the focus of diagnosis to another cause candidate group is determined according to the certainty factor. Therefore, the order of the questions can be limited to a local description and the overall flow can be controlled according to the certainty factor, and the order of the questions can be determined without considering all combinations of conditions in advance.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of a diagnostic device according to the present invention.

FIG. 2 is a diagram showing an example of a network structure of failure causal relationship knowledge.

FIG. 3 is a diagram showing an example of certainty factor adjustment knowledge.

FIG. 4 is a diagram showing an example of a question → answer network of one cause candidate group.

5 is a flowchart showing a flow of a diagnostic process of the diagnostic device of FIG.

6 is a flowchart showing a more detailed flow of processing relating to order determination of cause candidates and movement of diagnostic focus in the diagnostic processing shown in FIG.

[Explanation of symbols]

 1 Knowledge base 2 Diagnostic engine 3 User interface 4 Focus determination mechanism 5 Certainty factor calculation mechanism 6 Question display control mechanism 7 Failure causal relationship knowledge 8 Certainty factor adjustment knowledge 9 Question display order control knowledge

Claims (5)

[Claims] 1. A knowledge base in which diagnostic knowledge is stored, a diagnostic engine for performing diagnosis based on the diagnostic knowledge stored in the knowledge base, and an interface for an operator in the diagnostic engine. In a diagnostic device having user interface means,
The diagnostic knowledge stored in the knowledge base has a knowledge about a causal relationship and a knowledge for determining a diagnosis procedure separated from each other. 2. The diagnostic device according to claim 1, wherein knowledge for controlling a display order of questions to an operator is adopted as knowledge for determining the diagnosis procedure, and the diagnostic engine displays the questions. A diagnostic device characterized in that it controls the display order of questions to the operator based on knowledge for controlling the order. 3. The diagnostic device according to claim 1, wherein the knowledge regarding the causal relationship is configured such that cause candidates having the same result side event are treated as a cause candidate group, and the diagnosis engine is operated by an operator. Based on the information given by, the selection of a possible cause group and the ordering of each cause candidate in the selected cause candidate group are repeated to make a diagnosis. Knowledge that controls the display order of questions for each cause candidate group is given to the knowledge base, the question that influences the ordering of the cause candidates in one cause candidate group is controlled based on this knowledge, and further according to the response to the question A diagnostic device characterized by being adapted to control the movement of the diagnostic focus to another cause candidate group. 4. The diagnostic device according to claim 3, further comprising: providing a knowledge base with knowledge about a certainty factor representing the strength of possibility of each cause candidate in the cause candidate group, and the diagnosis engine is configured to A diagnostic apparatus, characterized in that the movement of the focus of diagnosis to a cause candidate group is determined according to the certainty factor. 5. The diagnostic device according to claim 4, wherein the knowledge about the certainty factor changes a default strength of a causal link indicating a pair of relation between a result and a cause candidate, and the default strength. The certainty factor and the description about the magnitude of change are included, and the certainty factor can be changed by adding the amount of change determined based on the information given by the operator to the default certainty factor. A diagnostic device characterized by being present.