JPH02146629A - Trouble diagnostic device and inference device - Google Patents

Abstract

PURPOSE:To facilitate the description of a deep knowledge which is difficult to describe and to facilitate the production and the maintenance of a knowledge base by using a mechanism which sorts finely the functions and structures of a diagnostic object to changing these sorted functions and structures into the hierarchies based on the including relation and also inferring the state of the diagnostic object based on the data on the hierarchies. CONSTITUTION:A logical structure memory part 16 sorts finely the functions and structures of a diagnostic object to change these functions and structures into the hierarchies based on the including relation and holds the data on these hierarchies. A logical structure operating part 15 estimate the detailed states of the functions and structures of the diagnostic object, i.e., the node of the lowest rank node of the data structure based on the node state of the data structure held by the part 16 as well as the data structure. Thus it is possible to diagnose such troubles that so far could not diagnosed by the experiential knowledge only. Then the description of knowledges can be facilitated and at the same time the number of rules can be decreased. Thus the diagnostic ability is improved.

Description

[Detailed Description of the Invention] Industrial Application Field The present invention relates to a failure diagnosis device for diagnosing failures of equipment and devices that can finely classify functions and structures and hierarchize them based on inclusion relationships. .

BACKGROUND OF THE INVENTION In recent years, failure diagnosis systems using computers have been actively developed. In such a system, the specialized knowledge of a skilled engineer can be reduced to ``if 0.''.

Then 09. It is. It is often expressed in the form of production rules such as '', and the knowledge base is a collection of production rules.

In the case of forward inference using production rules, a question is first asked to obtain information about the diagnosis target, and the answer to the question is stored as a fact in the work area.

Then, a rule whose condition part is satisfied by that fact is selected, and the conclusion part of the selected rule is added to the work area as a new fact. Thereafter, rules whose condition parts are satisfied by the facts in this work area are sequentially executed, and the process is repeated until there are no more executable rules.

Backward reasoning is often used in fault diagnosis. In the case of fault diagnosis, since the number of fault locations that can be concluded is limited, fault diagnosis is performed by forming a hypothesis regarding the fault location and performing the process of proving it by backward reasoning. During this certification process, the user of the failure diagnosis system is asked questions regarding the diagnosis target at any time.

Furthermore, failure diagnosis may be performed by bidirectional reasoning in which backward reasoning is performed using a conclusion drawn by forward reasoning as a hypothesis.

Other knowledge expressions other than production rules include:
Frame representation is often used.

Problems to be Solved by the Invention When considering the knowledge possessed by experts in troubleshooting electronic equipment such as TVs and VTRs, it is important to understand that the knowledge acquired through experience and the knowledge of the subject of diagnosis (for example, the equipment configuration of electronic equipment) knowledge of functional relationships and signal flows). However, when fault diagnosis is performed using a computer, the empirical knowledge (shallow knowledge) of fault repair experts is usually expressed in the form of production rules, or the structure of the target to be diagnosed is expressed as a frame. For this reason, the following problems occurred.

When knowledge is expressed using production rules, the knowledge becomes flat and the knowledge base becomes extremely complex and huge, making maintenance difficult. This is because production systems cannot deal with failures that are not described as rules at all, so the only way to improve diagnostic ability is to increase the number of registered rules. Furthermore, when performing failure diagnosis, it is necessary to ask the repair person about information regarding the diagnosis target at the time of inference. However, the production system does not have a framework for asking questions, so it is necessary to write rules for asking questions.
This causes the number of rules to increase.

On the other hand, there is a method that uses not only empirical knowledge (shallow knowledge) but also deep knowledge to infer the failure location by expressing the structure of electronic equipment in a frame.

However, in the case of frames, the description is complex, and only knowledge engineering and computer experts, so-called KEs, can describe the knowledge well.

As described above, in the conventional technology, the method of expressing knowledge is complicated and does not match the form of knowledge possessed by experts in the repair of electronic equipment, making it difficult to create and maintain a knowledge base.

In view of the above-mentioned problems, the present invention provides a failure diagnosis device that provides knowledge expressions that suit the form of knowledge possessed by failure repair experts and is equipped with an inference mechanism suitable for each knowledge expression. be.

Means for Solving the Problems In order to solve the above problems, the invention according to claim 1 provides a logical structure memory that holds data that finely categorizes the functions and structures of a diagnostic target and hierarchizes them based on inclusion relationships. section and the state of the node of the data structure held in the logical structure storage section as input, and using the data structure held in the logical structure storage section, determine the detailed function of the diagnosis target, which is the node at the lowest layer of the structure. This is an inference device that has a logical structure operation unit that estimates the state of the structure.

The invention according to claim 2 also provides an input section for inputting information such as symptoms and observed values related to a diagnosis target, and a rule for storing causal relationships between information such as symptoms and observed values of the diagnosis target and a failure function and a failure range. a storage unit, a forward inference unit that estimates a failure function and a failure range using the causal relationships in the rule storage unit based on the information input by the input unit; The present invention is a fault diagnosis device equipped with the above-mentioned inference device that estimates the state of detailed functions and structures of a diagnosis target using data in which the functions and structures of the diagnosis target are hierarchized as input.

Furthermore, a backward inference unit performs inference with the detailed functions and structures derived by the inference device as a goal, an interrogation unit inquires about information related to the diagnosis target that is necessary when inference is made by the backward inference unit, and the inquiry unit newly inputs information. An overall control unit may be added to enable forward inference to be made again based on the information regarding the diagnostic target.

Another solution is to use a question card storage section that stores question items for asking the repairer about the condition related to the diagnosis object, categorized into related items, and the data stored in this question card storage section. A fault diagnosis device can also be considered that includes a question card control unit for inputting the state related to the object, a rule storage unit for storing causal relationships between information such as symptoms and observed values of the diagnosis target, and the failure function and failure range, and a forward reasoning unit.

Furthermore, this fault diagnosis device may include an inference mechanism having the logical structure storage section and the logical structure operation section.

and an input section that includes the question card storage section and the question card operation section, a forward inference section that estimates the state of the object using the information from the input section and the causal relationship held in the rule storage section, and a forward inference section. The logical structure operation unit and the logical structure operation unit estimate the detailed function and structure state of the diagnosis target based on the data held in the logical structure storage unit, using the conclusion part of the rule in the logical structure storage unit as input. By using a diagnostic device having a backward inference section that performs inference with the result as a goal, even greater effects can be obtained.

Operation The present invention enables the knowledge necessary for fault diagnosis of electronic equipment to be constructed on a computer in a form suitable for the form thereof, by virtue of the above-described configuration. First, by finely classifying the functions and configurations to be diagnosed and layering them using inclusion relationships, it is possible to easily build the deep knowledge of repair experts on a computer, rather than using complex knowledge expressions such as frames. .

This knowledge is then used to estimate more detailed functional and structural conditions of the diagnostic target, making it possible to make inferences based on deep knowledge. This makes it possible to diagnose faults that could not previously be diagnosed using only empirical knowledge.

Furthermore, by combining the above knowledge representation and inference devices with forward reasoning and backward reasoning in conventional production systems, it is possible to easily describe knowledge, reduce the number of rules, and improve diagnostic ability. can.

Furthermore, by classifying question items into related items, organizing them in the form of question cards, and providing a mechanism to control the question cards, rules related to questions can be removed from the production rules, and the number of rules can be reduced. can be reduced.

Furthermore, since multiple questions can be asked at once in the form of question cards, the hassle of stopping reasoning after each question is eliminated.

In a production system, fewer rules means that the knowledge base is easier to write and maintain.

EXAMPLE Hereinafter, an example of claim 1 of the present invention will be described with reference to FIG. FIG. 2 is a diagram showing the configuration of an inference device according to an embodiment of claim 1 of the present invention. In FIG. 2, 21 is a logical structure operation section, and 22 is a logical structure storage section. The logical structure storage unit 22 holds data in which functions and structures to be diagnosed are hierarchically classified according to their inclusion relationships. A part of this data structure in a VTR is shown in FIG. In FIG. 3, each rectangle is called a node, and the arrow connecting the rectangles represents an inclusion relationship. In other words, the signal circuit in a VTR consists of a recording signal circuit, a playback signal circuit, and an E-E signal circuit, and a recording signal circuit consists of a recording video signal circuit, a recording audio signal circuit, and a recording color signal circuit. .

Among the nodes in FIG. 3, the lowest layer nodes such as the luminance signal circuit and the video head are called terminal nodes, and represent the detailed functions and structure of the object to be diagnosed. In this way, the data structure representing the relationship between nodes does not necessarily have to be a tree structure, but may be a network structure. The logical structure operation section 2I infers the state of this terminal node.

FIGS. 4 and 5 show an algorithm in one embodiment of the logical structure operation section 2I. The logical structure operation section has as input,
Takes the state of any node. The status of a node is either "normal" or "abnormal."

If the node's state is 'normal', check the existing state of that node; if it is 'normal', do nothing. If the existing state is undetermined, the state of that node is set to ``normal,'' and recursive execution is performed using as input that the state of the child node is ``normal.'' If the existing state of a node was ``Abnormal'', change the state of that node to ``Normal'' and change the state to the parent node that previously set the node to ``Abnormal''. Then, recursive execution is performed using the input that the state of the child node is "normal". This is repeated until the terminal node is reached.

If the state of the node in the input is "abnormal", the fifth
The state is estimated according to the algorithm shown in the figure.

First, check the existing state of the node and check if it is "normal".
If so, do nothing.

If the existing state is "abnormal" or undetermined, the state of that node is set to "abnormal", and recursive execution is performed using the input that the state of the child node is "abnormal". At this time, if the status of none of the child nodes becomes ``abnormal'', it is assumed that the node itself is not ``abnormal'', and the status of the node is reset to ``normal'', and the status of the parent node is changed. Let me know what has changed. This is repeated until the terminal node is reached. Furthermore, the set of terminal nodes estimated to be "abnormal" by recursive execution is compared with the set of terminal nodes that were already "abnormal" before the logical structure operation unit operates, and the difference set includes the Change the state of the terminal node to "normal" and notify the parent node that the state has changed.

For example, if the situation is as shown in Figure 6, node 3 is “
When there is an input “abnormal”, the set of terminal nodes that initially have the status “abnormal” is (node 4, node 5)
Since the set of terminal nodes that have become abnormal due to the input is (node 5, node 8), the difference set (node 4, node 8) is changed to the "normal" state.

A parent node that is notified of a state change from a child node changes its own state to "normal" once the state of all child nodes has been changed to "normal", and then changes the state of the parent node to "normal". Notify the node that the state has changed.

By estimating the state of the diagnostic target using the functions and structures of the diagnostic target in this way, it becomes possible to diagnose faults that could not be diagnosed using only empirical knowledge (shallow knowledge). Furthermore, by changing the algorithm of the logical structure operation section, various types of diagnosis can be performed.

FIG. 1 shows the configuration of an embodiment of a diagnostic device incorporating an inference device according to claim 1. In FIG. 1, 11 is an overall control section, 12 is a forward reasoning section, 13 is a backward reasoning section, 14 is a rule storage section, 15 is a logical structure operation section, 16 is a logical structure storage section, 17 is a question card control section, and I8 is a question card storage section. When the failure diagnosis is started, the overall control section 11 first calls the question card control section 17. The question card control section 17 searches the data held in the question card storage section I8, displays a series of questions, and accepts input from the user. When the user's input to the question is completed, the overall control unit 11 then calls the forward inference unit 12, and forward inference is performed. In forward inference, if the node name of the logical structure and its state are written in the conclusion section of the fired rule, the logical structure operation section 15 is called using the node name and its state as input. The logical structure operation section I5 changes the data in the logical structure storage section 16 using the above-mentioned algorithm. This cycle is repeated until there are no more question cards to display and no more rules to fire. At this point, the failure location to be diagnosed has been narrowed down to some extent in the form of the state of the terminal node of the data in the logical structure storage unit 16.

In order to further narrow down the estimated failure locations, the overall control unit calls the backward reasoning unit I3 with the hypothesis that the state of some of the estimated failure locations is “normal”.The backward reasoning unit 13 calls the rule storage unit 14. Search for a rule that has a goal in its conclusion part among the firing rules held in , and check whether the condition part holds true.If there is a question item in the condition part of the rule, check that question item. is used as an argument to call the question card control unit 17.The question card control unit I7 searches the question card storage unit 18, displays a series of questions including the question item given as an argument, and receives input from the user. If the condition part of the rule is not a question item, backward reasoning continues with that item as a subgoal.If new information regarding the diagnosis target is obtained in this way, the overall control unit calls the forward reasoning unit 12 again, Returning to the initial inference cycle, this algorithm is illustrated in FIG.

A condition for whether or not to display the series of question items may be added to the series of question items held in the question card storage section.

Effects of the Invention As described in 2., the present invention provides a mechanism for finely classifying the functions and structures of a diagnostic target, hierarchizing them based on inclusion relationships, and inferring the state of the diagnostic target using the data. The following effects can be obtained.

(1) In-depth knowledge, which has been difficult to describe in a knowledge base until now, such as the function and structure of a diagnostic target, can be easily described, and the creation and maintenance of the knowledge base becomes easier.

(2) Since the state of the diagnostic target is estimated using deep knowledge such as the function and structure of the diagnostic target, it becomes possible to diagnose even failures that could not be diagnosed with empirical knowledge (shallow knowledge).

Furthermore, by combining the above inference mechanism with forward inference and backward inference, the following effects can be obtained.

(3) Since it is no longer necessary to describe the functions and structures of the diagnostic target as production rules, the types of knowledge described as production rules are unified, and the knowledge outlook is improved.

(4) The number of rules can be reduced, making it easier to create and maintain a knowledge base written in production rules.

Furthermore, the following effects can be obtained by providing a mechanism for classifying and retaining questions for inputting information regarding a diagnosis target by related items and controlling the questions.

(5) Since it is no longer necessary to write rules for asking questions, the types of knowledge described as production rules are unified, and the visibility of knowledge is improved. (8) - Since users can ask multiple questions at once, users The number of questions asked will decrease.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a fault diagnosis device according to an embodiment of the present invention, FIG. 2 is a block diagram of an inference device according to an embodiment of the present invention, and FIG. FIG. 6 is a flowchart showing the inference procedure; FIG.
The figure is a flowchart 1 of a diagnostic procedure in a failure diagnostic device according to an embodiment of the present invention. 11... Overall control unit, 12... Forward reasoning unit, 13
... backward reasoning section, 14... rule storage section, 15.
...Logical structure operation section! 6...Logical structure storage section, 17
...Question card control unit, I8...Question card storage unit, 2I...Logical structure operation unit, 22...Logical structure storage unit. Name of agent: Patent attorney Shigetaka Awano

Claims (1)

[Scope of Claims] (1) A logical structure storage section that classifies the functions and structures of a diagnostic target and holds data in which they are hierarchically organized based on inclusion relationships, and a data structure held in this logical structure storage section. An inference device that includes a logical structure operation section that takes the state of a node as input and uses a data structure held in a logical structure storage section to estimate the state of a function or structure to be diagnosed, which is a node at the lowest layer of the structure. (2) an input unit for inputting information such as symptoms and observed values related to a diagnosis target; a rule storage unit for storing causal relationships between information such as symptoms and observed values for the diagnosis target and a failure function and failure range; a forward inference section that estimates a failure function or a failure range based on the information input by the forward inference section and using the causal relationship in the rule storage section; a logical structure storage section according to claim 1; 2. A fault diagnosis device having a logic structure operation section according to claim 1, wherein the fault function and fault range are input. (3) a backward inference unit that performs inference with the detailed functions and state of the structure derived by the logical structure operation unit as a goal; an interrogation unit that inquires about information regarding the diagnosis target that is necessary when inference is made by the backward inference unit; and the interrogation unit. 3. The failure diagnosis apparatus according to claim 2, further comprising an overall control section for enabling forward inference to be made again based on information regarding the diagnosis target newly inputted by. (4) A question card storage section that stores and stores question items for asking the repair person about information such as symptoms and observed values related to the diagnosis target, categorized into related items, and the data stored in this question card storage section. Claim 2: A question card control unit for inputting information such as symptoms and observed values related to a diagnosis target using the control unit.
3. A fault diagnosis device comprising the rule storage unit as described above and the forward inference unit as claimed in claim 2. (5) The failure diagnosis device according to claim 2, wherein the question card storage section and the question card control section according to claim 4 are used as input sections. (8) The fault diagnosis device according to claim 3, wherein the question card storage section and the question card control section according to claim 4 are used as an input section for forward inference and an interrogation section for inquiring information necessary for backward inference.