JP2650914B2 - Process abnormality diagnosis device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process abnormality diagnosis apparatus for diagnosing a block including a transition of a plant in order to contribute to an improvement in reliability and operation rate of a plant. It is.

[Conventional technology]

FIG. 4 is a process connection diagram showing a conventional process abnormality diagnosis apparatus disclosed in, for example, JP-A-55-4614.
In the figure, 1 is a plant that outputs a process signal to be diagnosed, 2 is a process signal indicating a process state quantity in the plant 1, and 3 is a standard such as a factory setting value or a plant design value required for process diagnosis. Is a reference value signal for each process state quantity from the plant database 3, and 5 is a process abnormality diagnosis for comparing the process signal 2 with the reference value signal 4 stored in the plant database 3. An abnormality diagnosis processing unit 6 for performing a determination process is a message indicating a diagnosis result.

Next, the operation will be described. The process signal 2 of each process state of the plant 1 is sent to the abnormality diagnosis processing unit 5, where it is compared with the reference value signal 4 of the corresponding process state quantity stored in the plant database 3, and the deviation is set in advance. Abnormality is detected based on whether or not it falls within each of the threshold values thus set, and abnormality diagnosis processing is performed using a relational database of abnormal events and cause candidates described in a CCT or IF-THEN type branch processing rule. The diagnosis result is output and displayed as a message 6. Here, the reference value and the threshold value of the process state quantity are fixed values set in advance. The abnormality detection index is, for example, in the form of | X−Xref | <ε... Normal | X−Xref |> ε. Here, X is a process state quantity which is a process signal from the plant 1, Xref is a process state quantity reference value, and ε is a threshold value. In the CCT method, this index gives the status of the primary node. In the method based on the IF-THEN type rule, this index gives the success or failure of a fact indicating an abnormal event, and a diagnosis is made by searching for the cause by a rule of a format referring to the fact.

[Problems to be solved by the invention]

Since the conventional process abnormality diagnosis apparatus is configured as described above, the reference value of the process state quantity and the threshold value for abnormality determination are fixed, so that abnormality diagnosis can be performed only in the steady state of the process. It is limited to the vicinity, and there is a problem that diagnosis cannot be performed when the state of the process fluctuates greatly such as when the load of the plant 1 fluctuates. In addition, although a method of predicting the reference value of the process state quantity using a simulator of the process dynamic characteristic is also conceivable, generally the dynamics of the actual process are very complicated, and it is difficult to estimate the state quantity purely by numerical calculation processing. Since it requires an enormous amount of calculation and time, it is extremely difficult to realize it in real time, and there has been a problem that the diagnostic apparatus becomes very expensive.

The present invention has been made in order to solve the above-described problems, and has been developed using a process-based macro model obtained by converting a knowledge obtained from a physical law of a target process or a process dynamic characteristic analysis into a knowledge base. It is an object of the present invention to obtain a process abnormality diagnostic device which estimates a reference value and a threshold value of the internal state quantity in real time, and can perform abnormality diagnosis based on the estimated values in the case of a change in the process state.

[Means for solving the problem]

The process abnormality diagnosis apparatus according to the present invention estimates an internal state quantity from a current environment value of the process when it is assumed that the process is currently normal, and sets a threshold indicating a normal range of the process to a current value of the process. A process abnormality is estimated by estimating from the internal state quantity and comparing the difference between the current internal state quantity of the process and the internal state quantity when the process is assumed to be normal with the threshold value. It is.

[Action]

The inference processing unit according to the present invention uses a physical model of process dynamics stored in a frame-type database, a physical rule stored in a rule-type database, and a rule obtained by dynamic characteristic analysis to determine a current process state quantity during normal operation. Predict the internal state quantity of the process, compare the predicted value with the current process state quantity, and compare the difference one after another with a threshold for abnormality determination determined according to the process state, and Perform dynamic status anomaly diagnosis.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, reference numeral 7 denotes a frame-type database in which a physical model of process dynamics, which is various attributes (physical property values, relational expressions, etc.) appearing in the physical structure and physical laws of a target process, is stored. This is a rule-type database that stores procedural knowledge obtained by analyzing dynamic characteristics such as physical laws (how to use relational expressions, etc.), empirical rules, and heuristics in a search format, and can describe rules in binary logic and fuzzy logic. These are provided in the plant database 3. Reference numeral 9 denotes a frame type database (input means) for managing the process state quantity. Reference numeral 10 denotes a frame type database 7, 9 as a working memory, and an inference processing unit (inference means, abnormalities) for performing inference processing based on the rule type database 8. Diagnostic means)
It is.

Next, the operation will be described. First, when a process signal 2 indicating a process state quantity is sent from the plant 1,
This is stored in the frame type database 9. The process signal 2 includes a signal serving as a boundary condition (boundary value) and a signal indicating an internal state of the process. First, the abnormality diagnosis processing unit 5 refers to the boundary value sent as the process signal 2 and the frame type database 7 to predict the internal state of the process based on the procedural knowledge stored in the rule type database 8. I do. By comparing this predicted value with the internal state which is the process signal data from the plant 1, it is determined that the process is abnormal. Here, taking the drum type steam generator as an example, the boundary value corresponds to the feed water flow rate, the feed water pressure and the outlet steam pressure, and the internal state corresponds to the temperature distribution and density distribution inside the steam generator. However, the physical quantity distribution itself cannot be detected, and the liquid level and the like correspond to the internal state. Hereinafter, a method of estimating the internal state of the process and a method of determining an abnormality will be described.

First, the physical structure of the target process is described in a frame format. The target process is broken down into several components from its physical mechanism. Components are not only those that are in line with the substance of the process (pumps, valves, pipes, etc.), but also those that are considered more appropriate to be decomposed as elements from an engineering point of view, for example, nodes used in dynamic characteristic simulation.・ Junkions are also included. The frame of each component describes its physical attributes and connection relationships between the components. Further, among the relational expressions relating to mass balance and heat balance, those relating to the structure of the expression (such as input / output relation between variables) are also described in the frame. Most of the knowledge described in these frame formats is engineering well-known. On the other hand, know-how knowledge such as how to use these frame-type knowledge is described in the rule-type database 8.

Next, a procedure for predicting the internal state of the process is shown in FIG. First, in step 1ST, the boundary value of the target process is set in the frame database 9. Next, in step 2ST, the current internal state distribution is provisionally initialized from the given boundary value and the history so far. For this purpose, knowledge obtained from dynamic characteristic analysis and process operation experience is used as a rule in the rule type database 8. The type of rule is a fuzzy logic type, for example, "The primary inlet temperature is high, the secondary inlet flow is low, ...
・ If this is the case, the secondary outlet temperature will be considerably high. A lot of things like that are collected. With these rules, a rough provisional estimate of the unobservable quantity in the process internal state is obtained.

Next, using this provisional estimated value, a mass heat balance calculation between the components is performed in step 3ST. That is, an observable state quantity and a physical property value are calculated by giving a process boundary value and a predicted value of the internal state quantity. At this time, if the variable corresponding to each slot of the frame type database 7 storing the physical structure of the process is known, the variable is determined if the variable is unknown, and if unknown, the value is repeatedly obtained to obtain the target physical quantity. calculate.

Next, in step 4ST, constraint conditions regarding the process are calculated from the calculated various physical property values, and it is determined whether or not the constraints are satisfied. Taking a drum-type steam generator as an example, assuming the liquid phase volume (water holding amount) from each boundary condition, calculate the volumes of the two-phase portion and the gas phase portion, respectively, and the sum of these is the volume of the entire drum. That is a constraint. Whether or not the constraint condition is satisfied is determined based on whether or not the equation as the condition is satisfied within a certain allowable range.

If the condition is satisfied, it is assumed that the state quantity has been correctly estimated. If the condition is not satisfied, the predicted value of the internal state quantity is corrected in step 5ST. In this case, inference based on fuzzy logic rules is used. For example, "If the total volume estimate is slightly larger than the entire drum,
Slightly reduce liquid phase estimates. " The process returns to step 3ST using the corrected value and repeats this.

As described above, the target process internal state quantity is estimated. Next, the abnormality determination method will be described. Basically,
The determination is made by comparing the internal state quantity of the process estimated by the above method with the observed value (actually measured value). If the state of the process is fluctuating, the threshold value is adjusted according to the degree. The observed process amount is X, the estimated value of X is Xest,
Assuming that the threshold is ε, | X−Xest | <ε, and ε is adaptively changed. For example, a fuzzy logic-type rule such as “ε = A · ΔY + f (Z) +...” If “the time variation rate of the process amount Y is considerably large, the value of the process amount Z is small,. Determine ε based on the format. Here, the time variation rate of the process amount is estimated by a difference operation from the history of the sampled process data.

That is, in the abnormality diagnosis at the time of a process change according to the present invention, as shown in FIG. 3, a process state quantity in a normal state in response to an operation input is estimated (step 11ST), and in parallel with this, the current process state is determined. Next, the threshold value ε for abnormality determination is estimated according to the procedure of the flow chart of FIG. 2 (step 12).
ST), it is determined whether the difference between the current process state quantity and the normal process state quantity estimated in step 11ST is larger than the threshold value obtained in step 12ST (step 13ST),
If it is smaller, it is determined that the process state quantity is normal (step 14ST), and if it is larger, it is determined that it is abnormal (step 14ST).
15ST), end each process.

As described above, in the process abnormality diagnosis device, the process internal state is efficiently predicted based on the dynamic characteristic analysis and the process operation experience, and the threshold value for abnormality determination is appropriately determined (adjusted) in accordance with the process state. Therefore, appropriate abnormality diagnosis can be performed even when the process fluctuates.

In the above embodiment, the case where the present invention is applied to process diagnosis has been described. However, the present invention may be used for process monitoring and control, and the same effects as in the above embodiment can be obtained.

〔The invention's effect〕

As described above, according to the present invention, the internal state quantity when the process is assumed to be currently normal is estimated from the current environment value of the process, and the threshold value indicating the normal range of the process is set to the current value of the process. Configured to diagnose the process abnormality by comparing the difference between the current internal state amount of the process and the internal state amount when the process is assumed to be normal with the threshold value Therefore, even if the state of the process fluctuates greatly, there is an effect that the abnormality of the process can be accurately diagnosed.

[Brief description of the drawings]

FIG. 1 is a block connection diagram showing a process abnormality diagnosis apparatus according to an embodiment of the present invention, FIG. 2 is a flowchart showing a flow of a process internal state estimation process, and FIG. 3 is a flowchart showing a process abnormality diagnosis flow. FIG. 4 is a block diagram showing a conventional process abnormality diagnosis apparatus. 1 is a plant, 2 is a process signal, 3 is a plant database, 7 is a frame type database, 8 is a rule type database, and 10 is an inference processing unit. In the drawings, the same reference numerals indicate the same or corresponding parts.

 ──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-61-51208 (JP, A) JP-A-57-86908 (JP, A) JP-A-64-21510 (JP, A)

Claims (1)

(57) [Claims] 1. A plant database for storing knowledge relating to the physical characteristics of a process in a plant, input means for inputting environmental values and internal state quantities of a process, and referring to the knowledge stored in the plant database. And estimating the internal state quantity assuming that the process is currently normal from the environment value input by the input means, and setting a threshold indicating the normal range of the process to the internal state quantity input by the input means. Estimating means for estimating the difference between the internal state quantity input by the input means and the internal state quantity estimated by the estimating means with the threshold, and diagnosing abnormality in the process. Process abnormality diagnosis device equipped with