JPH0477616A - Abnormality diagnosing apparatus - Google Patents

Abstract

PURPOSE:To perform the high-speed analysis of abnormality or to perform the analysis of abnormality characterized by a high speed and excellent explanation capability by combining the technologies such as neural networks and knowledge processing systems. CONSTITUTION:An event discriminating device 30 identifies the kind of an abnormal event which is generated in a plant by using the time-series data which are received from the plant 10 through a data receiving device 20. A neural network receives the data with respect to the amounts of the main states from the plant and outputs event codes. A diagnosis controlling device 40 reads the generated event code out of the event discriminating device 30, retrieves diagnostic strategy data 50 and determines the diagnostic strategy. A diagnosis processing device 60 reads the diagnostic strategy from the diagnosis controlling device 40 and performs the diagnostic processing by using diagnostic knowledge bases 70. When the strategy is inputted from the diagnosis controlling device 40, a corresponding diagnosis processing part is started. After the result of the diagnosis is obtained, the response is made by using an output device 80 and an input device 90.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a diagnostic device, and particularly to a device for diagnosing the cause of an abnormality occurring in a plant or large-scale equipment.

[Conventional technology]

Conventionally, for example, regarding neural networks, there is an invention disclosed in Japanese Patent Laid-Open No. 1-116869. In addition, regarding the application of neural networks to identify abnormal events in atomic couplants, NIS P.I.
pages 851 to 856 (Proc, of
5PIE Vo Q, 1095Ppplicati
ons of Artificial Intelli
gence VII (1989) 851-856).

[Problem to be solved by the invention]

The above conventional technology uses a neural network or
It is about its applications, and does not discuss conventional knowledge processing systems such as expert systems, or the combination of neural networks with other technologies.

Processing using neural networks is generally fast, while knowledge processing systems are characterized by their excellent ability to explain the obtained results.

An object of the present invention is to provide a high-speed abnormality diagnosis device or a high-speed abnormality diagnosis device with excellent explanation ability by combining techniques such as a neural network and a knowledge processing system.

[Means to solve the problem]

In order to achieve the above object, the present invention uses a neural network to identify an abnormal event occurring in a plant or the like, and based on the identification result.

Determine strategies for diagnosis using a diagnostic processing device using knowledge processing technology, such as limiting the scope of diagnosis.

[Operation] Effective diagnostic strategies such as limiting the range of knowledge used in one diagnostic process based on the event identification results by the neural network can speed up the diagnostic process.

Furthermore, if diagnosis is performed using knowledge processing technology, explanations to questions about the results can be provided in the same way as in conventional knowledge processing systems.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an apparatus according to an embodiment. In the figure, 10 is a plant, 20 is a data acquisition device, and 30 is an event discrimination device using a neural network. 40 is a diagnostic control device which uses 1 diagnostic strategy determination data 50. Further, 60 is a diagnostic processing device.

A diagnostic knowledge base 70 is used. 80 is an output device, 9
0 is an input device. An example in which the present invention is applied to an atomic couplant will be described below.

Next, the operation of the embodiment will be explained. Event discrimination device 3
0 identifies the type of abnormal event occurring in the plant using time-series data imported from the plant 10 via the data import device 20. Figure 2 shows the configuration of the neural network. As shown in the figure, the neural network used in this example has an output layer.

It consists of three layers: an intermediate layer and an input layer. Time-series data of major state variables such as reactor pressure and reactor water level are used as input to the input layer. On the other hand, the output of the output layer corresponds to the code of the event. That is, each unit outputs 0 or 1 (or a value close to it). For this neural network, plant data or
Using data from the simulator, weighting factors are determined in advance so that the event code corresponds to the corresponding event. A normal error backpropagation algorithm or the like is used to learn this weighting coefficient.

The neural network takes in data about key state quantities from the plant and outputs event codes.

FIG. 3 is a chart showing the contents of the diagnostic strategy determination data 50. In this data, a range of system equipment to be diagnosed is given in correspondence with the event code.

FIG. 4 is a flowchart showing an overview of processing in the diagnostic control device 40. This device reads the generated event code from the event discrimination device 30 (step 1001). Next, a diagnostic strategy is determined by searching the diagnostic strategy data 50 (step 1002) based on the event code (step 1003).

Here, as a diagnostic strategy, we have shown the limitation of the range of system equipment to be diagnosed, but there are other strategies as well.
This includes switching the diagnosis method, such as using en rules or using knowledge about the functional configuration of system equipment.

FIG. 5 is a flowchart showing the processing in the diagnostic processing device 60. In this device, the diagnostic strategy from the diagnostic control device 4o is read (step 2001), and the diagnostic knowledge base 70
Step 2002) to activate relevant knowledge in
Diagnosis processing is executed (step 2003). Here, if a strategy regarding a diagnosis method is received from the diagnosis control device 40, a diagnosis processing unit corresponding to the strategy is activated. If nothing is entered, diagnosis using the 1f-then rule is performed.

After obtaining the diagnosis result, the output device 80 and the input device 9
By using 0, it is possible to answer questions about the validity of diagnosis results, plant status, etc., like in a normal knowledge processing system. Further, it is also possible to manually change the diagnostic strategy using the input/output device.

According to the device of the present embodiment, by determining a diagnostic strategy using the event discrimination results using a neural network, it is possible to efficiently perform a diagnosis using a diagnostic processing device using knowledge processing. This makes it possible to shorten the time it takes to obtain diagnostic results compared to normal systems. Furthermore, regarding the diagnosis results, it is possible to logically explain the knowledge used, the reasoning process, etc., as in a normal knowledge processing system.

Note that in the device of this embodiment, all processing in the diagnostic processing device is performed by knowledge processing, but some or all of the diagnostic processing may be performed by a method other than knowledge processing such as a model comparison method. The present invention is similarly applicable to these cases. Further, in that case, it is possible to similarly realize faster diagnostic processing. When using various diagnostic methods,
Decide which method to use as a diagnostic strategy.

〔Effect of the invention〕

According to the present invention, it is possible to realize a diagnostic device that quickly identifies the cause when an abnormality occurs in a plant or large-scale equipment.

Therefore, the use of a diagnostic device such as the present invention has a significant effect on improving economic efficiency and safety.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention, FIG. 2 is an explanatory diagram showing the configuration of a neural network, and FIG.
The figure is an explanatory diagram showing the contents of data, and FIGS. 4 and 5 are flowcharts showing an overview of the processing. 10...Plant, 20...Data import device,
30...Event discrimination device, 40...Diagnostic control device, 5
゜...Diagnostic strategy determination data, 60...Diagnostic processing device, 7o...Diagnostic knowledge base, 80...Output device,
9゜...Input device. M1 Figure 2 Event Court Town Show! 11〒-ノ84 Itokan Qi 7 弗弗弗弗图

Claims (1)

[Scope of Claims] 1. Means for capturing measurement data from a diagnostic target, means for determining an occurring event using a neural network, means for determining a diagnostic strategy based on the occurring event, and a diagnostic method. An abnormality diagnosis device comprising: means for executing processing. 2. Taking in measurement data from a diagnostic target, determining occurrence events from the measurement data using a neural network, and determining a diagnostic strategy based on the results of this determination;
An abnormality diagnosis device characterized by executing diagnostic processing based on the strategy. 3. An abnormality diagnosis apparatus according to claim 1 or 2, wherein the diagnosis process is based on knowledge processing using a knowledge base. 4. The abnormality diagnostic device according to claim 1 or 2, wherein the diagnostic strategy is related to the range of the diagnostic target or the diagnostic method.