JPH03161801A - Abnormality diagnostic device for fuzzy inference - Google Patents

Abstract

PURPOSE:To easily set up a judgement condition by estimating the cause of abnormality from a varied phenomenon (abnormal phenomenon) by fuzzy inference at the time of detecting the variation of a plant and indicating its counterplan processing. CONSTITUTION:A fuzzy rule forming part forms a fuzzy rule based upon relational data between an abnormal phenomenon and an abnormal cause and stores the formed fuzzy rule in a fuzzy rule storing part, and at the time of operating the plant 1, a diagnostic data base forming function part 2 fetches various original data from the plant 1 by a data input part, applies prescribed static processing to original data by a historical data forming part 203, extracts the original data by a prescribed timing, and writes the extracted data in a historical data storing part 204 in accordance with time sequence to form historical data. When an abnormarity deciding part 205 decides the generation of abnormality, an abnormality cause inference part and a counterplan processing inference part 4 execute fuzzy inference based upon the historical data. Consequently, an accurate judgement condition can easily be set up.

Description

[Detailed Description of the Invention] A. Field of the Invention The present invention relates to an abnormality diagnosis device that monitors the status of a plant and, when an abnormality is detected, determines the cause of the abnormality and countermeasures.

2 B. Summary of the Invention The present invention performs statistical processing on various types of raw data imported from Plan 1 to generate historical data, and when an abnormality is determined from the original data, the cause of the abnormality and countermeasures are determined from the Hiss 1 logical data. The fuzzy rules obtained through fuzzy inference and used in the inference are generated from human-generated relationship data between abnormal phenomena and causes of abnormalities, and enable accurate routine diagnosis through simple setting operations.

C. Conventional technology Modern plants have become larger, more complex, more continuous, and faster, and while productivity and product quality have improved significantly, plant operation requires advanced maintenance technology. be done.

3. When operating a plant, plan!・If a minor failure occurs in the equipment that makes up the system, fluctuations will occur in operation. If appropriate measures are not taken against breakdowns, there is a risk of quality defects, complete shutdown, and even secondary disasters. For this reason,
Early detection of plant operational fluctuations and causes of failure is important.

In the past, plant abnormality diagnosis relied on the comprehensive judgment of veteran plant operators, but in response to the increasing sophistication of maintenance technology, abnormality diagnosis systems using AI (artificial intelligence) have been developed.

D. Problem to be Solved by the Invention Generally, abnormality diagnosis systems can only diagnose abnormalities within the range of pre-given preconditions, and if the state of the plant deviates even slightly from the preconditions, abnormality diagnosis cannot be performed. There is a possibility that it is not.

Furthermore, there is a risk of misdiagnosis due to the influence of noise contained in the original data from the plant.

For this reason, in conventional abnormality diagnosis systems, preconditions must be set in detail and over a wide range, and in order to ensure the reliability of the original data, measures such as using a large amount of raw data as the subject of judgment are taken. However, there was a problem in that the setting of judgment conditions became extremely complicated.

In view of these problems, it is an object of the present invention to provide a plant abnormality diagnosis device that allows easy setting of judgment conditions and has excellent reliability.

E. Means for Solving the Problems and Effects In the abnormality diagnosis apparatus according to the present invention, an operator inputs data relating to an abnormal phenomenon and an abnormal cause in advance using the phenomenon cause relation data manual section. The fuzzy rule generation section generates fuzzy rules based on this relational data and stores them in the fuzzy rule storage section.

During plant operation, the data input unit takes in various types of raw data from the plant, and the historical data generation unit performs predetermined statistical processing on the raw data. For example, the raw data is extracted at a predetermined timing, and historical Historical data is generated by performing processing such as writing to a data storage unit.

On the other hand, the abnormality determination section monitors the presence or absence of an abnormality based on the original data. When the abnormality determination unit determines that an abnormality has occurred, the abnormality cause inference unit and countermeasure inference unit perform fuzzy inference based on this historical data, determine the cause of the abnormality and countermeasures, and display the inference results on the display unit. .

Abnormal cause inference department and countermeasure #l! When the position inference section operates, it is also possible to extract a necessary portion of historical data depending on the type of abnormality that has occurred, and perform inference using that data.

F. Example F.1 System overview Embodiments of the present invention will be described below with reference to the drawings.

The abnormality diagnosis system according to the present embodiment analyzes the mechanism of abnormality occurrence at the designer level during planning and design of Bran I,
The relationship between causes of abnormalities and abnormal phenomena, the operational know-how of experienced operators, and measures to be taken against abnormal situations-7 are stored in a computer as a knowledge base, and the process status is measured online to detect fluctuations in the plant. When this occurs, the cause of the abnormality is estimated from the fluctuation phenomenon (abnormal phenomenon) using fuzzy inference, and countermeasures are also indicated.

FIG. 1 shows an overview of an abnormality diagnosis system according to an embodiment of the present invention.

As shown in this figure, this system is comprised of a diagnostic database generation function section 2, a knowledge database construction function section 3, an abnormality cause/countermeasure inference function section 4, and a man-machine interface 5.

F. 2. Explanation of each part F. 2. 1 Diagnostic database generation function unit The diagnostic database generation function unit 2 imports various information (original data) generated in the plant 1-8 online, and uses this raw data to facilitate abnormality diagnosis. It generates a database (diagnostic database).

FIG. 2 shows the configuration of the diagnostic database generation function section 2. As shown in FIG.

The input control unit 201 takes in the state of the process in the plant 1 (see FIG. 1) online in real time, and generates the original data file 202.

The historical data generation unit 203 reads the original data from the original data file 202 at necessary timing according to the type of the original data, and stores the read data in the historical data file 204 in chronological order.

9 The abnormality determination unit 205 monitors process state fluctuations and abnormal conditions according to the abnormality determination data stored in the abnormality determination mask file 206. When an abnormality is detected in the diagnosis target, the abnormality detection notification is sent to the diagnostic database generation unit.

The diagnostic database generation unit 207 starts operating upon receiving the abnormality detection notification, reads data necessary for diagnosis from the historical data file, and generates the diagnostic database 208. This diagnostic database 208 is
It is handed over to the abnormality cause and countermeasure inference unit 5.

F. 2. 2. Knowledge Database Construction Functional Unit The knowledge database construction functional unit 4 creates a knowledge database for diagnosing abnormalities by fuzzy inference, and can be realized by 10 as part of the man-machine function.

FIG. 3 shows the structure of the knowledge database.

As shown in this figure, the knowledge database 301 is composed of inference rules (rule base) and affiliation functions (membership functions).

This embodiment generally deals with inference rules of the following format.

ifX is A and Y is B then
Z is C where X, Y are input variables, and 2 is an output variable. Bit type data or numerical type data can be freely handled as variables x, y, and Z.

For example, bit type data includes alarm contacts, timer up status, CB (circuit breaker) status, etc., and numerical type data includes process status (current, oil temperature, various levels, etc.), timer 11 value, etc. There is.

A, B, and C in the above inference rules are variables XY,
This is a fuzzy label indicating the degree of Z. Various fuzzy labels are used depending on the variable.

For example, when expressing deviation, NB (negative and large),
NS (negative and small), ZE (approximately zero), PS (positive and small), and PB (positive and large) are used.

Also, when expressing the ON/OFF state, ON, OFF
is used.

When expressing the size, ■S (very small), MS (small), MM (medium), ML (large), vr, (very large) are used.

When expressing time, VS (very short time), s'rl2 (short time), M (medium), LT (long time), VT (
used for a very long time).

When expressing high or low, VL (very low), ML (low),
MM (medium), Ml-1 (high), VH (very high) are used.

FIG. 4 shows the general form of the membership function.

Generally, the membership function has a function value m of 1 at point Xp.
．． 0, and the function value m becomes m1 at the point Xt.

F. 2. 3. Abnormality causes/countermeasures inference function department No. 5
The figure shows the configuration of the abnormality cause and countermeasure inference function section.

The abnormality cause/countermeasure inference unit 401 determines the cause of the abnormality and countermeasures by performing fuzzy inference using the diagnostic database 208 and the knowledge database 301 that have been filed at the time of abnormality detection I3. The cause of the abnormality and the countermeasures are stored in the inference result file 402.

Fuzzy inference can be performed using existing techniques such as the maximum-minimum method.

FIG. 6 shows a fuzzy inference method using the maximum-minimum method.

Assume that the following fuzzy relationship holds true.

R 1: u X=A+ Y=B+ then Z=
C1R2: if X=At Y=B2 then Z
=Ct It is assumed that the input variables X and Y are X+''/ shown in FIG.

14 When inferring using the maximum-minimum method, first find the degree of truncation for each fuzzy rule. In other words, for rule Rl, variables X and Y in the condition part and membership function A
A function value is determined from I (x) and L (y), and the minimum value is taken from among the determined values and set as the standing degree ω1. Similarly, for rule R2, the degree of establishment ω is determined.

Next, for each rule Rl, R2, the membership function C,,C, of the conclusion part and the degree of establishment ω. Find a trapezoidal function from ω. That is, for the rule R1, a trapezoidal function (shaded area) is obtained by taking the smaller value of the membership function C1 and the standing degree ω. Similarly, for rule R2, a trapezoidal function is obtained.

Then, a composite function is obtained by taking the maximum value of the trapezoidal function of each fuzzy rule, and the center of gravity G of this composite function is determined and used as an inference value. In other words, the degree to which Z holds true is approximately G.

F. 2. 4 Man/Machine Interface R Nobu: The man/machine interface function section 6 has a CRT and an input device, and has functions for constructing abnormality determination files and knowledge databases, and provides inference results such as abnormality causes and countermeasures to the operator. It has notification functions, etc.

Regarding abnormality determination file construction, it has functions for defining diagnostic conditions and historical data collection.

The definition of diagnostic conditions defines how changes in the original data are understood. The historical data collection 16 definition defines information for transferring original data to the historical file, which is necessary for inferring the cause of an abnormality and countermeasures.

Regarding knowledge database construction, we need to define each variable, that is, define the names of input variables, output variables, intermediate variables, etc., generate and define membership function names (fuzzy labels) and function curves, and generate and define inference rules.・There is a function to perform definitions, etc.

Regarding the notification of inference results, there is a function to arrange the order of inference results and a translation processing function.

Sorting the inference results is a process in which when multiple causes are estimated through inference, they are sorted in descending order of probability. The translation process is a process of translating the inference results into the language that the operator uses on a daily basis.

F. 3 Application Example The operation of the system of this embodiment will be specifically explained below using an application example to a simple model plant.

F. 3.1 Model plant Figure 7 shows a model plant.

This plant is a bomb (P) driven by constant water level control.
) 601, a solenoid valve (Vl) 602 provided downstream of the pump 601, a tank (B) 603, and a tank 60
It consists of a motor-operated valve (MVI) 604 provided on the downstream side of No. 3, and a flow meter 605 that measures the amount of water supplied.

In this configuration, the water from the water supply source A is temporarily transferred to the tank 6.
03 and supplies the necessary amount of water to destination C using an electric valve 604.

The values that can be measured with this plant are tank water level (LI),
These are the valve opening degree (ZI) and the water flow rate (FI).

F. 3. 2. Construction of a knowledge database In order to diagnose abnormalities in this plant, it is first necessary to construct a knowledge database. In order to build a knowledge database, the operator creates an abnormal phenomenon/cause matrix.

FIG. 8 shows an abnormal phenomenon/cause matrix.

In this figure, the fuzzy labels are:

A: Very likely to occur B: Occurs to a considerable degree C: Possible to occur 19 D゛ Rarely likely to occur E. Rarely to occur It is unknown whether or not this will occur. Possible causes of this abnormality include pump failure. By listing the causes of each abnormality, predicting whether an abnormal phenomenon (change in measured value) will occur when the abnormality occurs, and defining the degree of occurrence as A-E. be made.

FIG. 9 shows membership functions related to likelihood. Fuzzy labels A to E are attached to each member synoptic function.

FIG. 10 shows membership functions related to water level or flow rate. Each membership function is then given a fuzzy label.

20 PB; Sudden. ．． J:. Increase PS: Increase ZE: No change NS: Decrease NB: Sudden decrease Based on the above matrix and membership function,
A knowledge database is constructed.

The generated fuzzy rule is as follows.

However, this rule applies when process fluctuations in water level and flow rate match. (Left below) 21 RJ: jf F2=PB then Il1=E,
H3=E, H4=C, H5=E. II8=BR2
: if F2=PS then tll=E, lI
3=E, H4=B. H5=D, H6=ER3:
ir P2=NS then Hl=B, I{3=B
, }14=B, H5=B, }16=ER4: i
rF2=NB then tll=B, tl3=B,
tl4=c, lI5=B, I16=ER5: i
f F3=PB then H5=B, H6=ER6
: ir F3=PS then lI5=A, H6
=ER7:F3=NS then tl5=E, l
I6=BR8: if F3=NB then H5=
E, I+6=A FIG. 11 shows the inference of the cause of the abnormality in the model plant.

As shown in this figure, if there is no disturbance and the water level (F2) suddenly becomes .

122 H]=0.75 H3=0.75 H4-0 75 H5=0.85 H6=0 10 Translate this into everyday language, The CRT has the following A message is displayed.

■ There is a very high possibility that the motor-operated valve is malfunctioning in the open state. ■ pump failure, Solenoid valve (open) failure, water level gauge Although it may be malfunctioning, ■ Electric valves rarely fail in idle conditions.

G. Effects of the Invention As explained above, the abnormality diagnosis device according to the present invention generates fuzzy rules based on the relationship data between a given abnormal phenomenon and the cause of the abnormality, and generates fuzzy rules using this fuzzy rule. Based on the inference, the cause of the abnormality and countermeasures are determined and displayed.

Therefore, even if the preconditions for abnormality diagnosis are detailed and wide-ranging and there is a large amount of original data to be judged, it is possible to easily set accurate judgment conditions.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an overview of an embodiment according to the present invention, FIG. 2 is a block diagram showing the configuration of a diagnostic database generation function section, FIG. 3 is a block diagram showing the configuration of a knowledge database, and FIG. 4 is an explanatory diagram showing the membership function,
Fig. 5 is a block diagram showing the structure of the abnormality cause/countermeasure inference function section, Fig. 246 is an explanatory diagram showing fuzzy inference using the maximum-minimum method, and Fig. 7 is an outline of model plan 1 in an application example. Figure 8 is an explanatory diagram showing abnormal phenomena, original prisoner 71, and wax. Figure 9 is an explanatory diagram showing membership functions related to likelihood of occurrence. Figure 10 is an explanatory diagram showing membership functions related to water level and flow rate. Figure, Figure 11 is model blanc i
It is an explanatory diagram showing inference of the cause of the abnormality in . 3... Diagnosis database generation function section, 4... Knowledge database construction function section, 5... Abnormality cause and countermeasure inference function section, 6... Man-machine interface function section. 25 Overview of the embodiment Figure 3 Structure of knowledge database Figure 4 Membership function IO Figure 5 Abnormal cause/countermeasure inference function unit 208, Qー○ @+l Tsukuda 4a2

Claims (1)

[Claims] (1) A raw data input unit that imports various types of raw data from the plant, a historical data generation unit that performs statistical processing on the raw data to generate historical data, and a historical data storage unit that stores this historical data. an anomaly determination unit that determines abnormalities from original data according to established standards; a phenomenon cause relationship data input unit that receives relationship data between an abnormal phenomenon and an abnormal cause; and a fuzzy rule generation unit that generates fuzzy rules from this relationship data. and a fuzzy rule storage unit that stores the generated fuzzy rules; and an abnormality cause and countermeasure inference that infers the cause of the abnormality and countermeasures from historical data according to the fuzzy rules when an abnormality is detected by the abnormality determination unit. Department and
An abnormality diagnosis device using fuzzy inference, comprising a display section that displays the inferred cause of the abnormality and countermeasures.