JP2856081B2 - Failure diagnosis system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a failure diagnosis system used for specifying a failure site in a system including a device, a plant, and the like.

[0002]

2. Description of the Related Art In recent years, these electric power systems have become increasingly complex and large-scale, and early restoration after a failure has occurred is an important issue. For this reason, it is necessary to determine a failure section after a power failure, and as a diagnostic method, a method based on a sign has been conventionally used in many cases. This is based on the premise that a knowledge base such as specialized knowledge and experience is constructed, and it is more difficult to create such a database in a large-scale power system.

As a solution to this problem, for example, see IEEJ Transactions on Papers B, April 1990, p. 258-266 propose a new diagnostic method based on a model. This models the function of the system to be diagnosed in the form of discrete states of each component and its transition, and performs diagnosis based on this model, and has the following advantages. (1) Since the diagnosis algorithm does not depend on a specific object, a fault diagnosis system can be created only by creating a model of the structure or function of the system to be diagnosed. (2) Since the correctness of the diagnosis is determined only by the model, the failure diagnosis system itself can be verified only by inspecting the model. (3) By improving the accuracy of the model, the diagnostic ability can be correspondingly improved.

[0004] More specifically, diagnosis is performed by generating a hypothesis of a fault from observation information obtained from the system and verifying the hypothesis by simulation. The simulation sequentially executes each transition of a transition network that models a system to be diagnosed, and outputs an accepted hypothesis as a failure factor.

[0005]

The fault diagnosis system based on the conventional model is configured as described above, and obtains the observation information before the occurrence of the failure and the observation information after the occurrence of the failure (after the completion of the failure phenomenon) obtained from the system. Diagnosis processing is performed as the input data. For this reason, the processing range in the hypothesis generation process and the simulation verification process is wide, which may be disadvantageous in terms of diagnostic accuracy, and the processing time tends to be long.

The present invention has been made to solve the above problems, and an object of the present invention is to provide a failure diagnosis system capable of improving diagnosis accuracy and reducing processing time.

[0007]

According to a first aspect of the present invention, there is provided a failure diagnosis system for each element constituting a system to be diagnosed.
Describes the set of multiple states assumed to be possible for the element, the set of transitions indicating the state change between each of the above states, and the conditions necessary for each of the above transitions to express the causal relationship of the state change of the element A transition network comprising a set of transition conditions to be performed, inputting observation information before the occurrence of a fault obtained from the system, and extracting a corresponding state as an initial state from the state set of each transition network, A means for inputting observation information after the occurrence of a fault obtained from the system and extracting the corresponding state from the state set of the transition network as a final state, which is intermediate between the before and after the fault, at least Input the observation information obtained at a predetermined time slice where an abnormality is recognized in some elements, and set the corresponding state as the abnormality detection state, Means for extracting from the set, starting from the extracted abnormality detection state, sequentially tracing the transitions, transition conditions, and states backward (in a direction opposite to the causal direction) to determine a range within which the extracted initial state can reach the limit. Means for searching, means for extracting as a disturbance hypothesis transition a transition which does not occur unless set from outside, among transitions belonging to the searched range, for each of the extracted disturbance hypothesis transitions, Starting, the transition, the transition condition, and the state are sequentially followed in a forward direction (the causal direction),
Means for verifying whether or not the final state can be reached via the abnormality detection state with all the elements from which the observation information has been obtained, and a disturbance hypothesis transition when the verification information can be reached as a failure factor Output means.

According to a second aspect of the present invention, there is provided a failure diagnosis system according to the first aspect, wherein a set of states of a transition network that can be extracted as an abnormality detection state or a final state for each element is selected in advance. The extraction of these states by the extracting means is performed within the range of the selected state.

A fault diagnosis system according to a third aspect of the present invention provides a fault diagnosis system according to the first or second aspect, in which a set of transitions of the transition network which requires an external setting for generation of each element is selected in advance. In addition, the selected transition in the range searched by the search means is extracted as a disturbance hypothesis transition.

According to a fourth aspect of the present invention, in the fault diagnosis system according to any one of the first to third aspects, the verification means is configured such that, for each extracted disturbance hypothesis transition, starting from the disturbance hypothesis transition, Means for performing a simulation for sequentially changing conditions and states in a forward direction (the direction of the above-mentioned causal factors) to sequentially change the current state of the element; means for obtaining a set of states that can be the current state in the course of the simulation; The set of states is compared with the anomaly detection state at the intermediate time section and the final state, and all elements for which observation information is obtained can reach the final state via the anomaly detection state at the intermediate time section. Means for judging whether or not it is the case.

According to a fifth aspect of the present invention, in the failure diagnosis system according to the fourth aspect, in the course of the simulation, since the element related to the sensor should be operating but not operating, the sensor is assumed to be malfunctioning. If the inconsistency with the final state cannot be resolved except by performing the above, the sensor is provided with a means for setting a hypothesis of the sensor failure state.

According to a sixth aspect of the present invention, in the failure diagnosis system according to any one of the first to fifth aspects, when there are a plurality of extracted abnormality detection states, a disturbance hypothesis extracted based on each of the abnormality detection states is provided. A set of transitions is obtained, and verification is performed only on disturbance hypothesis transitions obtained from the product of the set of disturbance hypothesis transitions for each abnormality detection state.

[0013]

In the failure diagnosis system according to the first aspect,
In particular, the abnormality detection state is extracted from the observation information obtained in the time section in the middle of the failure phenomenon, and the search processing that traces backward from this abnormality detection state, the transition of disturbance hypothesis transition, and the verification processing that traces forward is performed. Since the search is performed, the range of the search is reduced as compared with the conventional case, so that the number of disturbance hypothesis transitions is also reduced, thereby improving diagnosis accuracy and shortening the processing time.

Further, in the failure diagnosis system according to the second aspect, candidates which can be in the abnormality detection state or the final state are selected in advance, so that the processing time in these extraction means is reduced, and the extraction of an unsuitable state is performed. Is also prevented.

Further, in the failure diagnosis system according to the third aspect, since candidates that can be disturbance hypothesis transitions are selected in advance, the processing time of the extraction means is reduced, and the extraction of inappropriate transitions is prevented. .

Further, in the failure diagnosis system according to the fourth aspect, each time the current state of the element is moved forward by one step by simulation, a set of possible states is recalculated, and a comparison is made between the abnormality detection state and the final state. Perform verification.

Further, in the failure diagnosis system according to the fifth aspect, in a predetermined case, by setting a hypothesis of a sensor failure state, the simulation proceeds to reach the final state.

Further, in the failure diagnosis system according to claim 6, when the number of generated failures may be limited to one,
The number of disturbance hypothesis transitions to be verified can be reduced to improve diagnostic accuracy and reduce processing time.

[0019]

【Example】

Embodiment 1 FIG. FIG. 1 is a diagram showing an overall schematic configuration of a failure diagnosis system according to Embodiment 1 of the present invention. In the figure,
Reference numeral 1 denotes observation information obtained from a system to be diagnosed, which is, for example, observation data before a failure such as a “closed” state of a circuit breaker provided in the system, for example, after a failure such as a “after operation” state of a relay (failure). There is observation data obtained after the phenomenon is completed) and observation data obtained in a time section between before the occurrence of the failure and after the occurrence of the failure.

Reference numeral 2 designates a model of the component of the system to be diagnosed by focusing on the functions of the components constituting the system to be diagnosed (corresponding to the various circuit breakers and relays described above, other transformers, buses, and the like). Although the contents will be described in detail with reference to FIG. 2 described later, a transition network is created for each element, and specifically, is stored in advance in, for example, a ROM.

Reference numeral 3 denotes a diagnostic algorithm which is a main part of the present system, and details of its function will be described later with reference to a flowchart of FIG. 3, but a hypothesis generator 31 for extracting a disturbance hypothesis transition based on the observation information 1; A hypothesis verification unit 32 that verifies the hypothesis and obtains a diagnostic solution. In particular,
It is composed of a ROM storing a software program for executing each algorithm, a RAM for reading and writing necessary for an arithmetic processing process, and a CPU for executing arithmetic processing.

4 is a hypothesis verifying unit 32 of the diagnostic algorithm 3
Is a diagnostic solution output that outputs a failure location obtained as a result of the diagnosis, specifically, a display device such as a CRT or a printer. In addition, although not shown, an input device such as a keyboard for instructing various operations is provided.

Next, the details of the transition network will be described with reference to FIG. The procedure for creating this transition network, which is equivalent to the system modeling itself, is not the main subject of the failure diagnosis system of the present invention, even though it is a premise of the present failure diagnosis system. FIG. 2A shows a transition network of the equipment when the protection range of the relay is considered as a unit of equipment, FIG. 2B shows a transition network of an open / closed state of a circuit breaker, and FIG. This is an example of such a transition network.

The transition network describes a set of a plurality of states assumed to be possible by the element, a set of transitions indicating a state change between the states, and a condition necessary for each transition to occur, and describes a state of the element. And a set of transition conditions expressing the causal relationship of the change.

In FIG. 2A, the code is surrounded by a circle or an oval in the state of the element. For example, E0 indicates "normal" and E1 indicates "power failure". E2 to E9 can be interposed between the states E0 and E1, and transitions are indicated by arrows connecting adjacent states, for example, E0 and E2. That is, it indicates a state change from E0 to E2.
Next, a transition condition is drawn from a small circle that intersects with the arrow indicating the transition and is displayed as a square. In the above example, the state E0
Is a condition necessary for transition to the state E2.

As shown in the lower section of the figure, in the transition network of the equipment, there are roughly four types of states: "normal", "power failure", "accident", and "accident propagation". The accident state of the equipment is a state in which an accident has occurred in the equipment. The accident propagation state is a state in which no accident has occurred in the facility but is affected by an accident in a nearby facility. If for some reason only the accident equipment cannot be separated from the sound system by tripping the circuit breaker, the equipment on the power supply side of the accident equipment becomes the accident transmission equipment. The number of stages in the accident propagation state is the number of facilities having impedance between the facility and the accident facility. The accident state or the accident propagation state becomes a power failure state when the circuit breaker is tripped and separated from the healthy system. Therefore, the transition condition to the power failure state is the same for transition from all states other than the power failure.

FIG. 2B shows a transition network of a circuit breaker as an element, which is similar to FIG. 2A. If any of the tripping relays that trip the closed circuit breaker is present, the circuit breaker
The state transits to the “trip signal reception” state. Thereafter, if the operation of the circuit breaker is observed, the state transits to the C2 "open" state, and if not, the state transits to the C3 "failure" state.

FIG. 3C shows a transition network of a relay. The shape is substantially the same as FIG. The condition for transitioning to the R1 "operation condition satisfaction" state is described in a different form for each type of relay.

By the way, the above-mentioned transition condition is specifically expressed by a specific state of one or a plurality of other elements and a conditional expression or the like in which these states are combined by a logical expression of AND or OR. . For example, as a transition condition for changing the state of the circuit breaker as the element 1 from C0 “closed” to C1 “trip signal reception”, one of the relay A as the element 2 and the relay B as the element 3 (OR ) Is R2
The condition "after operation" is set.

As described above, in FIG. 2, the transition networks of the individual elements are individually displayed, but the respective elements are connected in a mutual causal relationship via the above-described transition conditions. That is, by a set of transition networks for all the elements constituting the system to be diagnosed and a set of transition conditions relating these respective transition networks,
The system is modeled.

As will be described in the following description, for example, in the state of FIG. 2A, transition to E2 "short circuit accident" and E6 "ground fault accident" with a black square in the middle of the solid arrow. Is a disturbance hypothesis transition described later. This means that the transition to this state, that is, the transition from E0 “normal” to E2, E6 is
This indicates that these transitions do not occur unless set from outside, and these are analyzed in advance for each element as candidates of disturbance hypothesis transitions from their functions and are selected in advance.

A sensor failure state indicated by a double circle indicates that a sensor-related element should be operated but is not operating in the course of a simulation described later, so that the sensor is assumed to be malfunctioning. This is a state that is subject to a hypothesis that causes the occurrence of inconsistency with the final state when it cannot be resolved. C3 “failure” (FIG. 2B) is a state related to the circuit breaker, but is set here according to the sensor.

Next, the operation of the diagnostic algorithm executed using each of the observation information and the transition network described above will be described with reference to the flowchart of FIG. First, observation information is input (S1). As observation information,
First, there is conventional observation data obtained from a system before and after a failure occurs. The state before the occurrence of the failure is a state before the existence of the failure is recognized, and corresponds to the initial state. The term “after failure” refers to a state in which the presence of the failure has been recognized and the state has reached a steady state, and corresponds to the final state. An example of the setting of the initial state and the final state will be described with reference to the relay transition network of FIG. The initial state of the relay is R0 if not locked.
It is "operable", and if locked, it is R4 "operation lock". The final state is R2 "after operation" if the operation of the relay is observed, and R0 "operable" or R3 "failure" if the operation is not observed. Thus, the information that can be observed is determined by the characteristics of the system.

In the failure diagnosis system according to the present invention, in addition to the above, an abnormality is recognized in at least some of the elements between the before and after the occurrence of the failure, in other words, at least a part of the elements is abnormal. For an element, use observation data obtained at a time section where it is recognized that the state has changed from the initial state of the element. This is referred to as an abnormality detection state. To briefly show an example of the abnormality detection state, if a relay operation is observed immediately after a breaker trip, the relay operation becomes an abnormality detection state.

The states corresponding to the above initial state, final state, and abnormality detection state are extracted from the state set of the transition network of each element (S2 to S4). By the way, among the states that can exist in each element, there may be a state in which the above-described final state or the abnormality detection state cannot be obtained from the functional characteristics of each element. Therefore, a state that can be a final state or an abnormality detection state is selected in advance for each element, this is recorded on the transition network, and the extraction processing (S
3, S4) is performed within the range of the selected state. As a result, the calculation time required for the extraction process is reduced, and the extraction error is eliminated.

The above extraction processing is performed for all elements (S5). It should be noted here that, especially in the final state and the abnormality detection state, not all elements can be obtained depending on the characteristics of the system and its components, the details of the failure, the conditions at the time of failure diagnosis, etc. It is not. If the number of unacceptable elements is large, the accuracy of diagnosis is naturally affected. However, in this step (S5), once the extraction of each state of the available elements is completed, the process proceeds to the next hypothesis generation step.

The generation of the hypothesis is performed in steps S6 and S7 in FIG.
The specific contents will be described with reference to FIG. In FIG. 4, for simplification of description, only two elements, a bus A (element 1) and a relay A (element 2), are taken. In the figure, the display of the transition network is the same as in the case of FIG.
_1.1, s _2.1 the sign of the..., And the transition t _1.3, t _2.1 ··
And the transition conditions are _denoted by g _1.3 , g _2.1 . The contents of each state are shown in the figure.

Here, the concepts of the current state and the allowable state set are newly introduced. The current state is a state that the element takes at a point in the process of proceeding from transition to state. After the current state is set to the initial state, a transition is sequentially performed, and a set of states existing on a path where the final state can be the current state is defined as an allowable state set. FIG.
The state in the range surrounded by the dotted line frame corresponds to this. For example, in the transition network of relay A, if s
_{When the} transition from _2.2 “Satisfaction of operating conditions” to s _2.4 “Failure”,
State _s2.4 is excluded from the allowable state set because the state cannot transition to _s2.3 "after operation" which is the final state.

The operation of generating a hypothesis of the present subject will be described. Now, the observation information obtained from the system to be modeled in Figure 4, the state s _2.3 relay A is extracted as an abnormal detection state. This state _s2.3 is also the final state. The main feature of the present invention lies in the point that the extraction of the abnormality detection state is added. However, as the description regarding this will be described later, here, first, the generation of a hypothesis that is a main part of the diagnosis algorithm will be described.

Based on the above assumptions, starting from the extracted abnormality detection state (s _{2.3 of} relay A), transitions, transition conditions, and states are sequentially shifted backward (in the direction opposite to the direction of the causal relation relating to the state change). The search is made for a range in which the initial state (S2 in FIG. 3) extracted previously can reach the limit. In FIG. 4, the searched range is hatched. Among these, the transition condition g _2.1 for establishing the transition t _2.1 in the relay A is that the bus A is in the state s _1.3 (g
_2.1 = s _1.3 ), so that the above-mentioned backward progress reaches the state s _1.3 of the bus A via the transition condition g _2.1 of the relay A. Then, over the transition network of the bus A further proceeds backwards, stop at its initial state s _1.1. Therefore, the hatched portion on the transition network of the bus A also belongs to this search range. Although omitted in FIG. 4, g
_{Starting with 2.2} , other elements are marked with a diagonal line.
Although an example in which only one transition condition is traced is shown here, there may be a case where several transition conditions are traced for marking.

Next, the search range (shown by hatching in FIG. 4)
In the transitions that belong to the
The transitions that do not occur are extracted. Here, the bus A is in the state
s_{1. 1}State "s" from "normal"_1.3Transition t resulting in “short circuit accident”
_1.3Corresponds to this, so this is called the disturbance hypothesis transition
Extract (S6 in FIG. 3). Transition that can be a disturbance hypothesis transition
Transition, that is, a transition that does not occur without external setting
These disturbance hypothesis transitions
Transitions that can be candidates are selected in advance for each element.
This is recorded on the transition network and extracted
(S6 in FIG. 3) within the range of the selected transition.
By doing so, the processing time is reduced and there is no
You.

When there are a plurality of transitions that are candidates in the search range, all of the plurality of transitions are extracted as disturbance hypothesis transitions. The above extraction processing is performed for all abnormality detection states (S7), and the generation of the hypothesis ends.

Next, a description will be given of a process flow for verifying a hypothesis, that is, verifying whether or not the disturbance hypothesis transition extracted as described above can be adopted as a diagnostic solution. The verification process is performed for each extracted disturbance hypothesis transition. Therefore, starting from the first disturbance hypothesis transition, a simulation is executed in which the transition, the transition condition, and the state are sequentially traced forward (the direction of the causal relation relating to the state change) to sequentially change the current state of the element. If this part is illustrated in FIG. 4 earlier, the disturbance hypothesis transition t
_The state s _1.3 generated by _1.3 becomes the current state first, and then a possible forward transition is determined, and the current state is shifted to the state generated by the transition.

The simulation proceeds for each simultaneously occurring transition. This one simulation process can be divided into (1) execution of a transition and (2) determination of the next transition to occur. (1) In executing the transition, the transition is caused to change the current state, and the allowable state set described later is corrected.
Record in the history of transitions that have occurred. (2) In determining the next transition, the transition condition expressed by the state of another element is reviewed, and the transition condition that satisfies at that time is determined. A transition satisfying the transition condition is a candidate for the next transition.

If there is one candidate, the transition is determined to be the next transition. When there is a plurality of candidates and they compete with each other, they are determined in the following manner. That is, based on the performance of the system obtained by the results, analysis, and the like, for the transition in which the above-mentioned conflict of each element is assumed, a degree at which the transition occurs quickly (here, for convenience, this degree is referred to as a timer index) is set. Keep it. When there is a conflict, the transition with the largest timer index is selected, and the transition is regarded as the next transition. If there are a plurality of transitions with the maximum timer index, all of them are determined to be the next transition.

The above-mentioned permitted state set is modified in the following manner. That is, first, in the transition network of the element where the transition has occurred, a state that can be reached only from the state before the transition has occurred is excluded from the allowable state set up to that time. When a transition condition that is unlikely to be satisfied comes out as a result of reviewing the transition condition, a state that can be reached only by a transition according to the transition condition is similarly excluded from the allowable state set.

As a result of repeating the above processing, it is verified whether or not the abnormality detection state and the final state occurring via the abnormality detection state are both included in the allowable state set for all elements for which the observation information has been obtained. If any element is not included,
Even if the simulation is continued as it is, the current state of the element does not become the final state. This is a contradiction, the hypothesis is not satisfied (NO in S9), and the hypothesis is rejected. If the current state is the final state in all elements, the phenomenon obtained based on the hypothesis matches the fact obtained from the observation information. Therefore, the hypothesis is established (YES in S9), and the hypothesis is regarded as a diagnostic solution. In other cases, that is, when the final state of all elements is included in the allowable state set and there is an element whose current state is not the final state, the validity of the hypothesis cannot be determined (YET in S9). Continue the simulation.

The above-described verification is performed for all disturbance hypothesis transitions (S12), and a diagnostic solution obtained by the hypothesis establishment is output (S13), and the failure diagnosis processing is terminated.

In the course of the simulation, if the element relating to the sensor should operate but is not operating, and the inconsistency with the final state cannot be resolved except by assuming that the sensor has failed, This failure state is set as a hypothesis from the sensor failure states described with reference to FIG. 2, and the process proceeds. Of course, in this case, the hypothetically set sensor failure forms a part of the diagnostic solution.

Embodiment 2 FIG. Next, in the present invention, an operation and an effect in the case of inputting observation information at an intermediate time section before and after the occurrence of a failure and using it for diagnosis as an abnormality detection state will be described as a second embodiment of the present invention. A description will be given in comparison with a case where only the initial state and the final state are used.

Hereinafter, description will be made based on an example shown in FIG.
The upper part of the figure shows a transition network that models a system, in which four elements, element 1 to element 4, exist.
Each element can take two states represented by numbers “0” and “1”. In the figure, the contents of the respective symbols are shown at the right end of the figure. Among them, those with a black square in the middle of the solid arrow (transition) are candidate transitions that can be extracted as disturbance hypothesis transitions. A dotted line indicates a transition condition together with A (AND) or O (OR) displayed at the intersection.

The lower part of the figure is an example of the obtained observation information of each element, which is indicated by one of the state numbers “0”, “1” or “unknown”.

First, consider a conventional case in which a final state is extracted from observation information obtained after the occurrence of a failure, and a hypothesis is generated by following the transition network backward from the final state. As shown in the table at the bottom of the figure, the observation information that is the final state is obtained from the element 2 and the element 4, so that the state is traced backward from the state “1” of both elements. Here, the disturbance hypothesis transition between the element 1 and the element 3 is extracted based on the state “1” of the element 4, and therefore, the hypothesis verification is performed on the two disturbance hypothesis transitions at a later stage of the diagnosis algorithm. In this case, a diagnostic solution will be pursued.

On the other hand, according to the present invention, an abnormality detection state is extracted from the observation information obtained at the intermediate point in time, and a hypothesis is generated based on the state. The observation information obtained at this intermediate point is as shown in the table at the bottom of the figure, from which the state “1” of the element 4 is extracted as the abnormality detection state. When the transition network is traced backward from the abnormality detection state of the element 4, the disturbance hypothesis transition of both the element 1 and the element 3 can be extracted as in the conventional case. Only one disturbance hypothesis transition is extracted. That is, here, the observation information that the element 2 takes the state “0” at the same intermediate point is obtained. As a result, even if the element 3 transitions to the state “1”, the AND condition with the state of the element 2 is not satisfied, and the element 4 cannot transition to the state “1”.

As described above, according to the present invention, the disturbance hypothesis transition is extracted in a more limited manner by newly using the observation information at the intermediate point, and the diagnostic accuracy is improved by that amount. The processing time for hypothesis verification is also reduced.

Embodiment 3 FIG. FIG. 6 illustrates the operation and effect of the present invention in a case different from FIG. However,
Hypothesis generation (extraction of disturbance hypothesis transition) in the third embodiment
Is different from the first and second embodiments. That is, in the failure diagnosis, in the above-described embodiment, the possibility of multiple failures is considered. Therefore, when there are a plurality of extracted abnormality detection states, the disturbance hypothesis transition extracted based on each abnormality detection state is determined. All the unions of the sets (sum method) are subject to hypothesis testing. On the other hand, in the third embodiment, the product of the set of disturbance hypothesis transitions extracted based on each abnormality detection state (product method) is subjected to hypothesis verification, and there is no problem even if the number of faults is limited to one. It can be applied to the diagnosis of

Referring to FIG. 6, first, considering the conventional case, the state "1" of the element 5 is extracted as the final state, and the transition network is traced backward from this state to obtain three elements 1, 2, and 3. Is extracted. On the other hand, when the observation information at the intermediate time according to the present invention is used, the abnormality detection state is extracted by the elements 4 and 5. Of these, from element 4 to element 3
Is extracted, and from element 5, elements 1, 2,
3, three disturbance hypothesis transitions are extracted. Therefore, the disturbance hypothesis transition finally set by the intersection is conventionally three, whereas in the present invention it is one (element 3), which is the same effect as described in the second embodiment. is there.

[0058]

As described above, in the failure diagnosis system according to the first aspect of the present invention, the abnormality detection state is extracted from the observation information obtained in the time section during the failure phenomenon. From the starting point, a search is performed in a backward direction, a disturbance hypothesis transition is extracted, and a verification process is performed in a forward direction.Therefore, the range of the search is reduced as compared with the conventional case, and therefore, the disturbance hypothesis transition is performed. Is also reduced, thereby improving the diagnostic accuracy and shortening the processing time.

Further, in the failure diagnosis system according to the second aspect, since the candidates which can be in the abnormality detection state or the final state are selected in advance, the processing time in these extraction means is reduced, and the extraction of the inappropriate state is performed. Is also prevented.

Further, in the fault diagnosis system according to the third aspect, since candidates that can be disturbance hypothesis transitions are selected in advance, the processing time in the extraction means is reduced, and the extraction of inappropriate transitions is prevented. .

Further, in the failure diagnosis system according to the fourth aspect, each time the current state of the element is advanced one step forward by simulation, a set of possible states is obtained again, and the comparison with the abnormality detection state and the final state is performed. Since the verification is performed, the hypothesis verification processing is performed more reliably and accurately.

Further, in the failure diagnosis system according to the fifth aspect, in a predetermined case, the simulation is advanced by setting a hypothesis of the sensor failure state, so that a diagnostic solution can be obtained reliably.

In the failure diagnosis system according to claim 6, when the number of generated failures may be limited to one,
The number of disturbance hypothesis transitions to be verified can be reduced to improve diagnostic accuracy and reduce processing time.

[Brief description of the drawings]

FIG. 1 is a diagram showing an overall schematic configuration of a failure diagnosis system according to a first embodiment of the present invention.

FIG. 2 is a diagram for explaining the contents of a transition network.

FIG. 3 is a flowchart showing processing contents of a diagnosis algorithm of the present invention.

FIG. 4 is a diagram for explaining a process of generating a hypothesis;

FIG. 5 is a diagram for explaining a hypothesis generation process according to the second embodiment of the present invention in comparison with a conventional case.

FIG. 6 is a diagram for explaining a hypothesis generation process according to a third embodiment of the present invention in comparison with a conventional case.

[Explanation of symbols]

1. Observation information, 2. Transition network, 3. Diagnostic algorithm, 31 Hypothesis generation unit, 32 Hypothesis verification unit, 4. Diagnostic solution output.

Continuation of the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G01R 31/08 G01R 31/00

Claims (6)

(57) [Claims] For each element constituting a system to be diagnosed, a set of a plurality of states assumed to be taken by the element, a set of transitions indicating a state change between the respective states, and the respective transitions occur A transition network consisting of a set of transition conditions that describes the conditions necessary to express the causal relationship of the state changes of the elements, and inputs observation information before the occurrence of a fault obtained from the above system,
Means for extracting a state corresponding to this as an initial state from the state set of each transition network, inputting observation information after a fault occurrence obtained from the system,
Means for extracting a state corresponding to this as a final state from the state set of the transition network, which is intermediate between before the occurrence of the failure and after the occurrence of the failure, and in a predetermined time section in which an abnormality is recognized in at least some elements. Means for inputting the obtained observation information and extracting a corresponding state as an abnormality detection state from the state set of each transition network, starting from the extracted abnormality detection state, and setting the transition, transition condition and state Sequentially backward (opposite to the above causal direction)
Means for searching for a range within which the extracted initial state can reach the limit, means for extracting transitions that do not occur unless set from the outside as disturbance hypothesis transitions among transitions belonging to the searched range, For each of the extracted disturbance hypothesis transitions, the transition, the transition condition, and the state are sequentially traced forward (the causal direction) starting from the disturbance hypothesis transition, and the abnormality is determined for all elements for which the observation information is obtained. A failure diagnosis system comprising: means for verifying whether the final state can be reached via a detection state; and means for outputting, as a failure factor, a disturbance hypothesis transition when the state can be reached by the verification. 2. A set of states of the transition network that can be extracted as an abnormality detection state or a final state for each element is selected in advance, and the extraction of these states by the extraction means is performed by the above-mentioned selection. The fault diagnosis system according to claim 1, wherein the fault diagnosis is performed within a range of the state. 3. A set of transitions of the transition network, which needs to be set from outside for its generation for each element, is selected in advance, and the selected one of the selected sets in the range searched by the search means is selected. 3. The fault diagnosis system according to claim 1, wherein the transition is extracted as a disturbance hypothesis transition. 4. The verification means, for each of the extracted disturbance hypothesis transitions, sequentially traces the transitions, transition conditions and states in a forward direction (the causal direction) starting from the disturbance hypothesis transitions, and sequentially determines the current state of the element. Means for performing a simulation to change, means for obtaining a set of states that can be the current state in the course of the simulation, and comparing the set of states that can be the current state with the abnormality detection state and the final state at an intermediate time section 4. A means for judging whether or not all elements from which observation information has been obtained can reach a final state via an abnormality detection state in an intermediate time section. The failure diagnosis system according to any one of the above. 5. In the course of the simulation, since an element related to the sensor should be operating but not operating,
5. The sensor according to claim 4, further comprising means for setting a hypothesis of the sensor failure state when the inconsistency with the final state cannot be resolved except by assuming that the sensor has failed.
The failure diagnosis system as described. 6. When there are a plurality of extracted abnormality detection states, a set of disturbance hypothesis transitions extracted based on each abnormality detection state is obtained, and a product of a set of disturbance hypothesis transitions for each abnormality detection state is obtained. 6. The fault diagnosis system according to claim 1, wherein the verification is performed only for the disturbance hypothesis transition obtained from the above.