JPH02236199A - Diagnosing of process - Google Patents

Abstract

PURPOSE:To reflect an intention regarding a designer's target of a process and also to enable a diagnosis of an abnormality by constituting a process operation assisting device of a module for selecting a process flow model, a module for calculating an abnormality and a module for identifying elements of the abnormality causes, by quickly catching the abnormality of constituent parts of an industrial process. CONSTITUTION:The first function calculates functional targets of each process constituent element such as a flow rate, an amount of storage, a generating amount and an input/output balance of a mass and an energy, for instance, by utilizing process data being fed from an inputting module 2. The second function is a function converting a difference between a calculated functional index and a present target range, which indicates a deviation extent between the two, to an abnormality, and is performed for each constituent element by feeding a calculating parameters of the abnormality in a data base 6 of a process flow model, to a linear calculational function. The third function has a function to calculate a difference between the abnormality of each constituent element and the abnormality previously calculated, and to output the difference to a module 4 for identifying elements of the abnormality causes.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a process diagnostic device that supports process operation by quickly identifying abnormalities in constituent components in a process of an industrial plant.

(Prior Art) Many devices have been developed using various methods for plant diagnosis of industrial process plants. In particular, with recent advances in computer technology, various diagnostic devices with multiple diagnostic purposes have been developed. It has become possible.

Among these methods, H
MULTI proposed by Orten Lind
There is a Level Flow Model.

H. LINO, “The llse or
Flow Models for De-sign
of Plant Operating Proc.
edures"Riso-H-2341, Pape
r presented at: IWG/NPPCI
Specialists Meeting of
Procedure and System+s
for Assistino an Operat
orduring Nosal and Ano
malous Nuclear PowerPla
nt Operation Situations,
December 5-7.1979, Garchi
ng, F. R, GermanyH. L.I.
ND, ” Use or Flo
w Models for Automated
Plant Diaonosis” .!n:R
asmussen, J., and Rouse
, H., B. (Eds.), “Hu.
man Detection and Diagno-
“SiS or system failures”.
Plenum Press, New York, 19
81. This ttnd's Multi Level Flow M
In odel, a process is decomposed by functions such as generation source, transportation, and storage, and these decomposed components are multilayered according to the degree of aggregation as a process model. Diagnosis is based on a method that calculates the balance between the input/output and existing amount of each functional component, and identifies the imbalanced function as the source of the abnormality.

By the way, the most common abnormality diagnosis method is a method of identifying which of abnormal events registered in advance in a database has occurred based on a pattern of process data values. Although this method can quickly identify abnormal events that were taken into consideration when designing the abnormality diagnostic device, it is not possible to diagnose abnormal events that were not considered when designing the diagnostic device.

Multi Level Flow Hodel above
The advantage of this method is that it is possible to identify the origin of an unknown phenomenon beyond its function.

(Problem to be solved by the invention) Multi Level Flow by Lind
The method using the Model can deal with unknown abnormal events. However, because they do not model the intent of the process plant designer, they cannot accommodate information about what the entire process must accomplish at each point in time. In other words, even if a process does not achieve its purpose as an industrial plant or is in a dangerous state, if the flow among each functional component is balanced, no abnormality will be detected.

Similar difficulties are found in devices that use a model that simulates a process and perform diagnosis using differences between this model and data from the actual process. When diagnosing processes in artifacts such as industrial plants, it may be impossible to identify abnormalities using only a model of the target process, and the goals that the process should achieve may also reflect the designer's intentions. It is necessary to incorporate it into the model.

This invention was made in consideration of the above-mentioned circumstances, reflects the intention of the designer regarding the goal of the process,
Another object of the present invention is to provide a process diagnostic device that can diagnose abnormalities caused by events unknown to designers.

[Structure of the invention]

(Means for Solving Problem M) This invention provides a process flow model database in which process flow models created for each process situation and abnormal calculation parameters regarding the performance of each component of the process are stored in correspondence. and based on process data,
A process flow model selection module that selects an appropriate process flow model from the above process flow model database, calculates a performance index regarding the performance of each process component using process data, and calculates the difference between this performance index and the target value as described above. An abnormality calculation module that converts into an abnormality degree using abnormality calculation parameters and calculates an abnormality degree change from this abnormality degree, and a flow connection between the process components determined to be abnormal in the above process flow model in the upstream direction of the flow connection. The present invention is characterized by comprising an abnormality cause element identification module that calculates the reliability of the abnormality cause from the abnormality degree change m and identifies the process component in which the cause of the abnormality exists.

(Operation) Therefore, according to the process diagnosis design according to the present invention, the method using the model of the target process and the
Since the designer's intention regarding the purpose of the process is stored in the process flow model database in the form of constant calculation parameters, it is possible to accurately diagnose events that would not be judged as abnormal using a method that uses a process model. can.

In addition, since the process flow model is used to diagnose process components by tracing them in the upstream direction of flow connections, abnormalities caused by unknown events that were not taken into account when designing the process diagnostic device can also be diagnosed when making process operation decisions. Diagnosis can be performed.

"Example" Embodiments of the present invention will be described below based on the drawings.

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

The process diagnostic device 1 includes an input module 2, an abnormality degree calculation module 3, an abnormality cause element identification module 4, a display module 5, a process flow model database 6, and a process flow model selection module 7. The input module 2 inputs process signals from the processes of the industrial plant and sends them to the abnormality degree calculation module 3 and the process flow model selection module 7.
Outputs the process signal to

The process 7-model database 6 stores a process flow model created for each process situation, a process state represented by process data, and anomaly calculation parameters related to the performance of each process component in association with each other. .

Here, the process flow model is a model in which a plant process is divided into a plurality of plant components according to function and configuration, and connections between these plant components are expressed as a flow.

For example, when diagnosing a turbine bleed check valve erroneous closing accident in a boiling water nuclear power plant, a process flow model as shown in FIG. 2 is used. This process flow model is a model at rated operation. In this process model, the parts surrounded by square frames are process components, solid arrows indicate energy flows, and dash-dotted arrows indicate maintenance capacity flows.

In addition, if this turbine bleed check valve error lock-up accident occurs, the route from the turbine to the feed water heater to the condenser will be
Since the turbine oil check valve is shut off due to incorrect closing, the amount of steam required to rotate the turbine increases and the electrical output of the generator increases.

In addition, the above-mentioned abnormality degree calculation parameter refers to the performance index obtained for each plant component in the responsibility calculation module 3 as described later, but when converting this performance index into an abnormality degree using a linear function, This is the parameter for this linear loss.

This abnormality degree calculation parameter reflects the designer's intention regarding the goal of the process.

For example, the above turbine bleed check valve error! The abnormality degree calculation parameters used at the time of the Ivi accident are set based on the assumption that the goal of the plant is to supply rated power.

? On the other hand, the process flow model selection module 7 selects an appropriate process flow model at that time from the process flow model database 6 based on the process data input from the input module 2, and selects an appropriate process flow model at that time from the process flow model database 6 for each process component in the process flow model. The abnormality degree measurement parameter is outputted to the abnormality degree calculation module 3, and the process 7 row connection data in the process flow model is outputted to the abnormality cause element identification module 4. Note that these two modules 3 and 4 have a process flow model database 6.
Instead of sending all of the data therein, only a command for switching the settings of the data stored in the two modules 3.4 may be output.

Now, as shown in FIG. 3, the abnormality calculation module 3 has three main functions (A), (B). (C).

First machine! (A) is to use the process data input from the input module 2 to calculate a performance index representing the function of each process component at that point in time. Here, are the performance indicators related to quality, solidity, energy, etc.? , storage capacity, input/output balance, etc. Further, a plurality of performance indicators may be obtained for one process component.

For example, in the event of a turbine bleed check valve erroneous closing accident as shown in Figure 2, this first function (A) is capable of determining the sending end output as a performance index of supplied power, and the ratio of the station load to the generating end power as a performance index of power distribution. , generating end power as a performance index for conversion to electric power, turbine generator heat consumption rate (efficiency with respect to transferred energy at turbine inlet) as a performance index for conversion to rotational energy, and condenser exhaust as a performance index for steam exhaust heat. Oil enthalpy is required as a performance index for maintaining hot mouth and energy conversion. These performance indicators may directly use process data detected by sensors in the plant process, but in most cases they are calculated using calculation formulas from detected process data.

Next, the second function (B) converts the difference between the two (the difference between the performance index and the target range), which indicates how much the calculated performance index deviates from the target range of the current state, into an abnormality degree. It is a function. This conversion to abnormality is performed for each process component by inputting the abnormality degree calculation parameters in the process flow model database 6 into a linear function. This degree of abnormality is expressed as a number from 0 to 100, with 0 being normal and
It is calculated so that 100 is a definite abnormality.

When there are multiple performance indicators for one process component, multiple degrees of abnormality are calculated, and the highest degree of abnormality among them is taken as the degree of abnormality of that process component.

Finally, the third function (C) is a function that calculates the difference (amount of change in abnormality degree) between the normality degree of each process component and the previous abnormality degree, and outputs it to the abnormality cause element fixing module 4. . Process components having performance indicators that were determined to be abnormal by the abnormality degree calculation module 3 at the time of the turbine bleed check valve erroneous closing accident in a boiling water nuclear power plant are indicated by * in FIG.

The abnormality cause element identification module 4 inputs the abnormality degree change of the process component from the contribution degree meter module 3,
The system traces the process components determined to be abnormal in the process flow model in the upstream direction of the flow connection, calculates the degree of abnormality confidence from the amount of change in the degree of abnormality, and identifies the process component in which the cause of the abnormality exists. be.

Specifically, (1) below. (2). (The three processes are carried out. In other words, (1) 10% of the amount of change in the degree of abnormality input from the degree of abnormality calculation module 3 is renormalized into the degree of certainty of the cause of the abnormality of the process component itself, and 90% is determined to be the cause of the abnormality. It is transmitted to a certain upstream process component.If there are multiple abnormalities in the upstream process components, the maintenance function 70- is weighted three times as much as the flow of energy or mass and is transmitted in proportion to the degree of abnormality. <<2>> The same processing as in (1) above is also performed for the abnormality degree change amount transmitted from the process component. <<3>> If there is no abnormality in the upstream process component, the abnormality degree change is All of the quantities are renormalized into the anomaly cause certainty of that process component.

In the abnormality cause element identification module 4, to calculate the abnormality cause certainty, the abnormality degree change amount is set to ΔD, and the previous abnormality cause certainty is set to C. Then, the new abnormality cause certainty C
is executed in the smell. As a result of this processing, the process component with the highest certainty of the cause of the abnormality is identified as the cause of the abnormality, and is output to the display module 5. This display module 5 displays the identification results to the plant supervisor. For example, in the case of a turbine bleed check valve erroneous closing accident, a process component related to maintaining energy conversion is identified and displayed as the cause of the abnormality.

According to the above embodiment, since the designer's intention and diagram regarding the purpose of the process are stored in the process flow model database 6 in the form of abnormality degree calculation parameters, events that cannot be determined as abnormal using a method using a process model are It is also possible to make an accurate diagnosis. In addition, since the process flow model is used to trace the flow connections between process components in the upstream direction and calculate the certainty of the cause of anomaly for each process component, unknown unknowns that were not considered when designing the process diagnostic device can be calculated. Diagnosis necessary for process operation can also be performed for abnormalities caused by events.

As the process component causing the abnormality is quickly identified and displayed as described above, plant supervisors can stop using the process component causing the abnormality or adopt an alternative method. It is possible to quickly obtain important information for immediate processing, such as switching process goals.

In addition, in the above embodiment, since the degree of abnormality change is used for diagnosis and the reliability of the cause of the abnormality is determined as described above, even if an abnormality appears in other process components that are not determined to be abnormal, at an early stage of the occurrence of an abnormality, , Since the certainty of the cause of the abnormality is low for other process components, they are not determined to be abnormal.

In addition, although the abnormality is transmitted not only to the downstream side of the process flow model but also to the upstream side in the late stages of the occurrence of the abnormality, the reliability of the cause of the abnormality diagnosed up to that point does not decrease, so accurate abnormality diagnosis can be performed.

Furthermore, since the anomaly cause certainty was calculated using the formula (2), the overall anomaly cause certainty can be determined at once, excluding a total of 18 errors, regardless of the order of processing for which process components are calculated.

〔Effect of the invention〕

As described above, according to the process diagnosis arrangement according to the present invention, a process flow model database storing process flow models and abnormality calculation parameters;
A process flow model selection module that selects an appropriate process flow model from this process flow model database, and a duty calculation that calculates performance indicators, calculates abnormality traces using the above abnormality calculation parameters, and calculates the amount of change in abnormality degree. Abnormality cause element identification that calculates the certainty of the cause of anomaly by tracing the flow connection between the module and the process component determined to be abnormal in the process flow model, and identifies the process component in which the abnormality has occurred. Since it has a module, it is possible to reflect the designer's intention regarding the goal of the process, and also to diagnose abnormalities caused by events unknown to the designer.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of a process diagnostic device according to the present invention, FIG. 2 is a diagram showing an example of a process flow model stored in the process flow model database of FIG. 1, and FIG. 2 is a flowchart showing the functions of the abnormality degree calculation module of FIG. 1; DESCRIPTION OF SYMBOLS 11... Process diagnosis device, 3... Abnormality calculation module, 4... Abnormality cause element identification module, 6...
- Process 7 ports - model database, 7... process flow model selection module. Applicant's agent Hisashi Hatano Figure A

Claims (1)

[Claims] Process flow models created for each process situation,
and a process flow model database in which abnormal calculation parameters related to the performance of each component of the process are stored correspondingly, and a process flow model selection module that selects an appropriate process flow model from the process flow model database based on the process data. , calculate a performance index regarding the performance of each process component using process data, convert the difference between this performance index and the target value into an abnormality degree using the above abnormality calculation parameters, and calculate the abnormality degree change from this abnormality degree. The abnormality degree calculation module calculates the abnormality cause certainty by tracing the flow connection between the process components determined to be abnormal in the above process flow model from the above abnormality degree change amount, and calculates the abnormality cause certainty factor from the above abnormality degree change amount. A process diagnostic device comprising: an abnormality-causing element identification module that identifies existing process components.