JPH06161760A - System diagnostic device - Google Patents

Abstract

PURPOSE:To decide a decision without variance and speedily take a diagnosis by mechanizing the diagnosis which is taken by experts. CONSTITUTION:This device is equipped with a signal input part 1 which observes a state quantity in a system, a signal processing part 2 which processes information sent from the signal input part 1 into vector format, a case data base 4 which is stored with case vectors for collation, a collation and registration part 3 which collates an input vector generated by the signal processing part 2 with case vectors in the case data base 4 and registers it as a new case vector, and a deformed vector generation part 6 which generates a deformed vector as a hypothesis by using the case vectors in the case vector data base. Further, the device is equipped with a comparison part 7 which compares the generated deformed vector with the input vector processed by the signal processing part 2 and writes the result in a decision description part 5 and a control part 8 which controls the collation and registration part 3 and deformed vector generation part 6 according to the result written in the decision description part 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system diagnostic device for diagnosing a cause of failure of a large-scale and complicated system such as a computer system or an electronic device for controlling a plant.

[0002]

2. Description of the Related Art Conventionally, a system diagnostic device for diagnosing a cause of failure of a large-scale and complicated system has been developed by using an abnormality detection sensor provided in elements constituting the system, for example, circuits, electric motors, piping, structures, It was done based on the signal. These signals are displayed on the display device of the system, etc., but what kind of character the abnormality is from the displayed signal,
For example, it is often the case that a specialist has a sole judgment as to whether there is a serious failure somewhere in the device and whether the operation should be stopped immediately or whether the operation can be continued as it is due to a minor failure.

However, in recent years, generally, the demand for the reliability of such a large-scale system tends to become strict. For example, a power system digital relay can be said to be a large-scale electronic device, but in recent years, the power system is required to have higher and higher reliability, and the demand for a relay system for protecting the power system is further increased. It will be tough.

While this relay system is required to have the ability to react to a system abnormality without delay, unnecessary operation must be avoided as much as possible so as not to impair the operating rate of the power system. Further, the digital relay is also required to have the ability to monitor its own abnormality as an electronic device.

[0005] In such a situation where further reliability is demanded for a large-scale system, relying on an expert who has a discrepancy in judgment depending on the degree of skill and the situation will lead to various diagnoses in the future. It is likely to cause problems. Also, expert diagnosis tends to be delayed. Therefore, there is an increasing need to mechanize the system diagnosis performed by such experts.

[0006]

By the way, in order to diagnose the system, it is necessary to observe a group of data regarding the state in the system and estimate the cause of the failure from these data. Therefore, in order to mechanize the diagnosis by the expert, it is necessary to know what kind of information processing the expert is performing from the above-mentioned group of data in the diagnosis. Therefore, as a result of analyzing the information processing performed by the observer, it was found that the following judgments have a large weight in the diagnosis of many systems.

From the operational experience, the correspondence between the observation data indicating the state of the system and the failure cause is grasped, and when a certain new observation data is input, the failure cause is estimated by estimating from the correspondence relationship.

This information processing is already known because an expert system for estimating the cause of a failure has been developed, and in order to accurately diagnose a failure by this, a causal relationship between the observation data and the cause of the failure is known. Must be fully understood. In reality, there are many systems where this causal relationship has not been fully elucidated, and with this method alone,
It is not possible to construct a practical system diagnostic device.
Moreover, when the causal relationship is not sufficiently clarified, it is realistic to use the concept of estimating the cause of failure from the comparison with past cases.

The present invention has been made to meet the above-mentioned demands, and its purpose is to make the determination uniform by mechanizing the diagnosis made by an expert.
It is an object of the present invention to provide a system diagnostic device that can make a quick diagnosis.

[0010]

In order to achieve the above object, the present invention provides a signal input means consisting of a sensor group for observing a state quantity in a system and a vector of information sent from the sensor group of the input means. A signal processing unit that processes into a format, a case database in which case vectors for matching are stored, and an input vector generated by this signal processing unit is collated with a case vector in the case database,
Collation / registration means for registering as a new case vector in the case database, deformation vector generation means for generating a deformation vector as a hypothesis using the case vector in the case database, and the deformation vector generation means. Comparing means for comparing the deformation vector with the input vector processed by the signal processing means and writing the result in the judgment describing means; and the collation / registration unit based on the result written in the judgment describing means, And a control means for controlling the deformation vector generation unit to display and output the case vector that matches the input vector or the deformation vector when it is determined that there is a deformation vector.

In addition to the above configuration, if the control unit determines that there is no deformation vector that matches the input vector, the mismatching element of the case vector closest to the input vector between the collation / registration unit and the expert. Is to provide a man-machine interface for questioning and answering whether or not a mismatch is allowed.

[0012]

In the system diagnosing device having such a configuration, when the observation amount indicating the system state is input from the signal input means to the signal processing means, the input vector is processed from this observation amount, and this input vector is The similarity between the past case vector stored in the case database and the allowable deformation vector generated by the deformation vector generation means is evaluated by comparing with the comparison means, so that the main part of intelligent information processing is mechanized. As a result, it is possible to exert the diagnostic ability at high speed and without variation in the determination result.

Further, in the collation / registration section,
From the answer obtained through the interface, for the mismatch element of the case vector closest to the input vector, if mismatch is allowed, register the input vector as a variant of the closest case vector in the case database,
When the disagreement is not allowed, the input vector is registered as a new case in the case database, so that it is possible to construct a completely new case and its accompanying information.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an example of a configuration in which the system diagnostic device according to the present invention is applied to a fault diagnosis of a digital relay of a power system. In FIG. 1, reference numeral 1 denotes a signal input section including sensors 1a, 1b, 1c, ... For detecting electric signals from an analog input circuit of a digital relay, an A / D converter, an abnormality detection circuit provided in a computing element, and the like, 2 is a signal processing unit for processing the information sent from each sensor 1a, 1b, 1c, ... Of the signal input unit 1 into a vector format, 3
Compares the input vector generated by the signal processing unit 2 with the case vector stored in the case database 4,
The collation / registration unit registers a new case vector in the case database 4.

Reference numeral 5 is a judgment description portion in which the collation result in the collation / registration unit 3 is described, 6 is a deformation vector generation unit for generating a deformation vector as a hypothesis using the case vector in the case database 4, 7 Is a comparison unit that compares the deformation vector generated by the deformation vector generation unit 6 with the input vector.

Further, 8 is a control unit for controlling the collation / registration unit 3 and the deformation generation unit 6 according to the determination result described in the determination description unit 5, and 9 is a collation / registration unit controlled by the control unit 8. 3 is a display unit that displays the collation result of No. 3.

On the other hand, 10 is a man-machine interface for question and answer between the collation / registration unit 3 and an expert, and 11 is an expert who estimates the cause of failure through the man-machine interface 10. This is a failure cause estimation support unit that supports Next, the operation of the system diagnostic device configured as described above will be described.

Now, various sensors 1a of the signal input section 1,
1b, 1c, ... Receive electric signals from the abnormality detection circuit of the digital relay, etc., and when the detection signals are sent to the signal processing unit 2, the signal processing unit 2 appropriately scales these signals and The following vector is generated by binarizing "1" and "0" according to the property. [A1, a2, a3, ...] Signal 1, signal 2, signal 3 ...

The vector elements a1, a2, a3, ...
Generally, it may be a continuous analog quantity or may be discretized into “1” and “0” depending on the nature of the signal, but here, all are “1” and “0”. It is binarized.

The input vector thus generated in the signal processing unit 2 is sent to the collation / registration unit 3 and collated with the case vector stored in the case database 4 collected in the past. In this case, the case database 4 is stored in FIG.
As shown in, the pattern of the elements of the case vector observed in the past (here, since the elements are binarized, it will be referred to as a bit patterner) and the cause of failure and the like are stored as its auxiliary information. Further, for each case vector, information about a permissible deformation vector that can be regarded as the same type as the case is also stored. Here, the operation after a certain input vector is read by the collation / registration unit 3 will be described with reference to the flowcharts shown in FIGS.

First, in FIG. 3, when the input vector is read into the collation / registration unit 3 (step S1), the collation / registration unit 3 checks whether or not there is a case vector completely matching the bit pattern of the input vector. (Step S
2). Then, if there is a vector that completely matches the bit pattern, the matching / registration unit 3 describes the result in the determination description unit 5 as a message (step S3). For example, the message is “Match with case vector No. 6”. The determination description unit 5 is realized by a storage device similar to a working memory of a normal expert system.

If there is no input vector that matches the bit pattern, the collation / registration unit 3 writes a message such as "no matching case vector" in the determination description unit 5 (step S4), and The number of the case vector closest to the input vector (the one with the smallest number of mismatch bits) is obtained in order to be used in.

On the other hand, the control unit 8 corresponds to a normal expert system inference engine and rule group, and the control unit 8 reads the message described in the determination description unit 5 (step S5). The control unit 8 determines whether or not there is a case vector in which the message read from the determination description unit 5 completely matches the input vector (step S6). If there is a matching vector, the number of the case vector or The cause of the case is displayed on the display unit 9 (step S7). If there is no case vector that completely matches the input vector, the deformation vector generation unit 9 is activated (step S8).

In FIG. 4, the deformation vector generating unit 6 refers to the information of the allowable deformation vector, checks whether the allowable deformation is registered for the case vector closest to the input vector, and if it is registered, the allowable deformation vector is registered. A permissible deformation vector is generated from the information of (step S
9).

In this case, the permissible change vector is generated in this embodiment by a method paying attention to invertible bits as shown in FIG. In addition, a certain input vector is "1" or "0" obtained by binarizing the signal from the sensor as described above.
However, if the "1" and "0" signals are not of equal importance for all bits, and if a particular bit makes little sense for a fault condition, There is also.

The knowledge about such deformation varies depending on the case vector, and according to the experience of the operator of the target system, what kind of deformation is allowed for a certain case vector beforehand is shown in FIG. Must be registered as. In this case, the bit number of the invertible bit may be stored as the allowable deformation information. In the example of FIG. 6, the invertible bits are (3, 5, 7). There are several combinations of invertible bits for one case, for example five variants are registered for case 1 of FIG.

As shown in FIG. 6, when a case vector A does not match the input vector but is determined to be close, the deformation vector generation unit 6 determines the allowable deformation information of FIG. An allowable deformation vector A ′ of A is generated. This permissible deformation vector A'is considered to be of the same kind as the case vector A, although the details are different, and this deformation vector A'is sent to the comparison unit 7 and compared with the input vector (step S10). In this comparison unit 7,
If the input vector matches the permissible deformation vector A ', the fault state corresponding to the input vector is set to the case vector A
It is determined that the failure state is close to the failure state, and the result is written in the determination description unit 5 as a message (step S1).
1).

On the other hand, the control unit 8 reads the message from the determination description unit 5 (step S12), determines whether or not there is an allowable deformation vector that matches the input vector (step S13), and the applicable allowable deformation vector is If there is, the deformation vector generation unit 8 is stopped (step S14), and the case is displayed on the display unit 9 (step S15). Also,
If there is no permissible deformation vector, it is determined whether or not there is another permissible deformation vector.
The registration unit 3 is activated to register the input level (step S17).

The collation / registration unit 3 registers the input level as one of the allowable deformation vectors of the case vector closest to it or as a completely new case vector in FIG. First, in order to check whether or not the former can be registered, it is checked whether or not the mismatch can be allowed for each mismatched bit of the input vector and the closest case vector. A rule-based system or the like can be used for this determination, but in this embodiment,
The expert is asked a question through the machine interface unit 10 (step S18). It is determined whether or not all the mismatch bits are acceptable (step S19), and if acceptable, the input level can be registered as a modification vector of the case vector closest to it (step S20). If all the mismatch bits are unacceptable, it is registered as a completely new case vector (step S21).

When the input vector is registered as a completely new case vector, the cause of failure cannot be analogized due to the similarity with the past case vector. Therefore, the cause of failure is newly estimated, and this is shown in FIG. It is necessary to record such additional information. The cause of failure is estimated by asking an expert through the man-machine interface 10 (step S22). At this time, the failure cause estimation support unit 11 is also used. This is realized, for example, by an expert system that incorporates empirical rules for estimating what kind of failure occurs when the input level has what kind of bit pattern. actually,
No existing system is fully satisfactory with this technology. If such a system can be perfected, the judgment function based on the collation with the past case vector described so far becomes unnecessary. The failure cause estimation support unit 11 only provides information that is useful for experts within the scope of current technology. When the cause of failure or the like is obtained in this way, this is registered as attached information (step S23).

In the above embodiment, the fault diagnosis of the digital relay of the electric power system is intended, but the other applications are different only in the observation data and can be applied in the same manner as described above.

[0033]

As described above, according to the present invention, the input vector consisting of the observed amount indicating the state of the system is evaluated by the similarity with the past case vector and its allowable deformation vector. Since the main part of the intelligent information processing performed in the system diagnosis has been mechanized, it is possible to provide a system diagnosis device which has a high speed and a diagnostic ability without variation in the determination result.

[Brief description of drawings]

FIG. 1 is a block circuit diagram showing an embodiment of a system diagnostic device according to the present invention.

FIG. 2 is a diagram for explaining a storage format of case vectors in the case database of the embodiment.

FIG. 3 is a view showing a flowchart for explaining the operation of the embodiment.

FIG. 4 is a diagram showing a flowchart following FIG. 3;

FIG. 5 is a diagram showing a flowchart following FIG. 4;

FIG. 6 is a diagram for explaining a method of generating an allowable deformation vector in the embodiment.

[Explanation of symbols]

1 ... Signal input unit, 2 ... Signal processing unit, 3 ... Verification / registration unit,
4 ... Case database, 5 ... Judgment description section, 6 ... Deformation vector generation section, 7 ... Comparison section, 8 ... Control section, 9 ... Display section, 1
0 ... Man-machine interface, 11 ... Failure cause estimation support unit.

Claims (3)

[Claims] 1. A signal input means composed of a sensor group for observing a state quantity in the system, a signal processing section for processing the information sent from the sensor group of the input means into a vector format, and a case vector for collation. And a case database that is stored, collate the input vector generated by this signal processing unit with the case vector in the case database,
Collation / registration means for registering as a new case vector in the case database, deformation vector generation means for generating a deformation vector as a hypothesis using the case vector in the case database, and the deformation vector generation means. Comparing means for comparing the deformation vector with the input vector processed by the signal processing means and writing the result in the judgment description means; and the collation / registration unit based on the description content written in the judgment description means. And a control means for controlling the deformation vector generation unit and displaying and outputting the case vector that matches the input vector or the deformation vector when it is determined that there is a deformation vector. 2. A signal input means composed of a sensor group for observing a state quantity in the system, a signal processing section for processing the information sent from the sensor group of the input means into a vector format, and a case vector for collation. And a case database that is stored, collate the input vector generated by this signal processing unit with the case vector in the case database,
Collation / registration means for registering as a new case vector in the case database, deformation vector generation means for generating a deformation vector as a hypothesis using the case vector in the case database, and the deformation vector generation means. Comparing means for comparing the deformation vector with the input vector processed by the signal processing means and writing the result in the judgment description means; and the collation / registration unit based on the description content written in the judgment description means. Controlling the deformation vector generating unit and the case, if there is a case vector matching the input vector, or if there is a deformation vector control means for displaying and outputting the case, if there is no deformation vector matching the input vector by this control means If judged, the case vector closest to the input vector is mismatched between the collation / registration section and the expert. The elements, system diagnostic apparatus characterized by comprising a man-machine interface for Q & whether mismatch is tolerated. 3. The matching / registration unit, based on the answer obtained through the man-machine interface, with respect to the mismatch element of the case vector closest to the input vector, if mismatch is allowed, the input vector is closest to the input vector. A variant of the above is registered in the case database, and when the disagreement is not allowed, the input vector is registered as a new case in the case database.
The system diagnostic device according to 1.