JP3187246B2 - Plant diagnostic system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an abnormality of various instruments constituting a chemical plant, a steel plant, etc. More specifically, the present invention relates to a plant diagnostic system that can diagnose various abnormalities accurately by adjusting a diagnostic rule according to a process state.

[0002]

2. Description of the Related Art In recent years, plants used in all fields, such as steel and chemicals, are automated by a process control system using a computer, and humans (operators) are located in a central control room, where production management, operation management, The system focuses on operations such as safety management and facility management.

By the way, in such a plant,
There are various causes of plant abnormalities, such as instrument failures of sensors installed in the plant, abnormalities in the process itself, erroneous operations by the operator, and failures in process equipment or control equipment. It is desirable to be able to diagnose. Conventionally, model-based inference (MBR) has been introduced as one that can deal with failures that cannot be predicted or diagnose unknown objects (Van Vliet, E., et al.).
C. The NH3expert; `` A Model-Based Diagnostic Expert
System for an Ammonia Cooling Plant, made with MARV
IN, '' Second Working Paper of the MBRproject, Leide
n; Werkgroep Veiligheid, University of Leiden, 1991.
1).

Further, a system has been proposed that generalizes essential diagnostic work based on the behavior of a chemical process and facilitates system construction (Tomita, G .; Abnormal diagnosis of chemical plants led by empirical knowledge). System Development, Transactions of Chemical Engineering, Vol.15, No.2, pp.258-267, 1989.2).

[0005]

However, each of these systems uses a diagnostic rule based on signals from a group of instruments for abnormality diagnosis of a plant, and describes the diagnostic rule. However, there are problems such as the fact that it is difficult to apply the method to a plant in which there are many types of objects to be diagnosed, such as that it takes a long time to develop the system due to its limitations in terms of performance.

[0006] The present invention has been made in view of such a point, it is possible to determine a diagnostic rule (range) of instrument abnormality and process abnormality and to easily construct these diagnostic rules.
It is another object of the present invention to provide a plant diagnostic system that can flexibly respond to a change in a plant.

[0007]

SUMMARY OF THE INVENTION The present invention for achieving the above object is a plant diagnosis system for diagnosing an abnormality of an instrument or a process constituting a plant, and inputs information from a plant to be diagnosed. And a table that defines a plant model in which the plant to be diagnosed is modeled and an abnormality diagnosis rule to be applied according to the operation state of the plant. Using this table, diagnosis is performed according to the operation state of the specified plant. Inference means for extracting an abnormality diagnosis rule applied to a target plant, and diagnosis means for receiving information output from the plant model and diagnosing the plant according to the abnormality diagnosis rule specified by the inference processing means are provided. This is a plant diagnostic system characterized by the following.

[0008]

[Function] Causes of an abnormal plant include an instrument failure, an abnormality of the process itself, and the like. The diagnostic means is provided with a plurality of diagnostic means so that diagnosis based on instrument failures and process abnormalities can be optimized, and an abnormality diagnostic rule applied to these diagnostic means depends on the operating state of the plant. It is described in the table in advance.

The inference means extracts the optimum abnormality diagnosis rule according to the operation state of the designated plant using the table, and applies the abnormality diagnosis rule to the diagnosis means. The diagnosing means includes, for example, an instrument failure diagnosing means convenient for diagnosing an abnormality caused by an instrument failure and a process abnormality diagnosing means convenient for diagnosing a plant abnormality. According to the diagnosis rules , the optimum diagnosis means is selected, processes information from the plant model, and diagnoses the plant.

[0010]

An embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a conceptual diagram of the configuration of a system according to the present invention. In the figure, reference numeral 1 denotes a plant database collected from a plant to be diagnosed, which is constructed from information from equipment such as flow meters, valves, pumps, heat exchangers, distillation towers, pipes, etc., which constitute the plant. I have. As information from the equipment to be diagnosed, for example, if the equipment is a flow meter, a thermometer, or the like, it is a process variable such as a flow signal, a temperature signal, or attached to these devices. Vibration sensor, acceleration sensor, thickness sensor, acoustic emission (A
E) The signal is a device-specific signal from a sensor or the like.

Reference numeral 2 denotes a plant model in which a plant to be diagnosed is modeled. Applicable plant data is input from the plant database 1 and information necessary for plant diagnosis is output. The plant model 2 is constructed (constructed) by connecting a plurality of device models 21, 22,... That model actual devices such as valves and pumps via connecting means 23 so as to correspond to a plant to be diagnosed. ) Has been. Reference numeral 3 denotes a means for constructing the plant model 2.
As the plant model construction means, a known graphic editor or the like technique is used, and a plurality of device models and device connection means are appropriately used, and these are connected by, for example, a wiring operation so that a plant model can be constructed. It has become.

Each of the device models 21, 22,... Has process variables (variables for which measured values are set), instance variables (device-specific internal variables), and state variables (state-specific Variable to be set) can be set. In addition, it is possible to register (describe) an abnormal constraint condition for detecting an abnormality of the device, a normal constraint condition that is satisfied when the device is operating normally, and an inference method for specifying a diagnosis propagation method for a process error. It has become.

Reference numeral 4 denotes a user interface unit.
And a keyboard, and by operating these, the plant model 2 can be constructed via the plant construction means 3,
The operation status of the plant and the start of the plant diagnosis can be inputted, and the diagnosis result and each database can be monitored. Reference numeral 5 denotes inference means having a table 50 defining an optimum abnormality diagnosis rule (range) according to the operation state of the plant, and using this table to infer an abnormality diagnosis rule to be applied to a plant to be diagnosed. Is configured. Anomaly diagnosis rules refer to the entire table
ing. The operating state of the plant given to the inference means 5 corresponds to the operating state of the actual plant. For example, the operating state is specified by the operator from the user interface unit 4 or given from the control device controlling the actual plant. May be used.

Numeral 6 denotes a diagnostic unit which receives information output from the plant model 2 and receives an abnormality designated by the inference unit 5.
It has a function of diagnosing the plant according to the diagnostic rules . The diagnostic means 6 includes an instrument failure diagnostic means 61 which is convenient for diagnosing a failure of an instrument (a meter such as a flow meter or a thermometer) constituting the plant, and a diagnostic instrument 6 for diagnosing a process abnormality. Convenient process abnormality diagnosis means 62
Are provided.

FIG. 2 shows a table 50 used by the inference means 5.
It is a figure showing an example of. The horizontal axis of the table indicates the plant operation state (meaning the state of the plant operation and is characterized by the control target value of the device group. For example, there are a startup operation state, a steady operation state, and a shut down operation state. ) Is defined in a tree structure, and each device is defined on the vertical axis. For this reason, general
The horizontal axis shows the operating state of the runt, and the general
Because the vertical items are on the vertical axis, this table
It is possible to cope with such a plant. The contents defined in the upper state are merged with lower information priority and can be used in the lower state. Then, for each state, a state variable group used for each device, a valid abnormality constraint
A condition , its monitoring (evaluation) cycle, a valid normal constraint condition, a valid abnormality diagnosis method, and the like are registered (described). Therefore, conditions and values necessary for diagnosis are set for each device according to the operation state of the plant.

The operation of the apparatus having the above-mentioned structure will be described below. First, prior to the diagnosis, a plant model 2 simulating the actual plant to be diagnosed is constructed by the plant construction means 3. This can be easily established by connecting the ports of the respective devices by device connection means and defining the process status and the like. It is also possible to add a new device definition or device information.

When the operation of the plant starts, process data is collected from the actual plant, and the corresponding process data is given to the plant model 2. Inference means 5, from the user interface unit 4 or the control device or the like, and information about the operation state of the real plant, reasoning
A start instruction is given, and an inference process is executed according to the following procedure. (A) Extract abnormal constraint conditions set for each device in the operation state of the specified plant according to the table 50 (extract all constraint conditions obtained by merging the contents defined in the upper state with lower information priority). . Abnormal constraint
The item is one of the items described in the table. (B) Evaluate each abnormality constraint condition of each device according to a specified monitoring cycle. As a result of the evaluation, if the condition is satisfied, the process proceeds to the next step (c). (C) Diagnosis of an instrument failure is performed by the instrument failure diagnosis means 61 with respect to the established abnormal constraint condition. The instrument failure diagnosis means 61 verifies the reliability of information (process variable value) handled by each device model in the plant model 2. When inconsistency is detected, the instrument including the process variable fails or is abnormal. Judge. (D) When an instrument abnormality is detected in the step (c), the fact is notified to an operator or the like via the user interface unit 4. Thereafter, when the inference processing is executed, the process returns to the step (b). (C)
In the step, all the variable values are verified, and if no abnormality is detected, the process proceeds to the next step. (E) Diagnosis of a process abnormality is performed by the process abnormality diagnosing means 62 with respect to the abnormality constraint condition established in the step (b). The process abnormality diagnosing unit 62 diagnoses the process abnormality by propagating between the devices connected to each other via the device connection unit in accordance with the inference method definition given from the inference unit 5. (F) Notifying the operator or the like of the diagnosis result in the step (e) via the user interface unit 4. (G) To continue the inference, the process returns to the step (b). The above-mentioned steps (a) to (g)
This is the procedure by which the conditions are evaluated and processed. How to evaluate
Is the device process variable, instance variable, state
Whether the value of the variable is normal or abnormal (is within the range of normal values)
Or not).

FIG. 3 shows an example of the instrument failure diagnosis means 7.
It is a functional block diagram. In this figure, the process variables
The take-out means 71Inference start conditionIs satisfied,
The process variable group referenced by the start condition is
Take out of file 2. The abnormality detecting means 72 is the inference means 5
Given from table 50 by these instructionsAbnormal diagnosis rules
(Abnormal constraints, normal constraints)
Evaluate the process variables, etc. extracted from Dell 2 and establish
The abnormal constraint condition is transmitted to the abnormal instrument determining means 73.

The abnormal instrument judging means 73 receives the transmitted abnormal constraint condition, and judges an instrument failure by the following procedure. (A) Take out the abnormality constraint condition transmitted from the abnormality detection means. (B) Detecting process variables related to the extracted abnormal constraint condition from the connection relationship of the devices. That is, the process variable of another device via the port referred to by the abnormality constraint condition is extracted. (C) With respect to the normal constraints of all the devices, the normal constraints referring to the process variables extracted in the step (b) are taken out directly or through a port. (D) Among the extracted normal constraints, the conditions that include all the variable groups referred to by the abnormal constraints are excluded.
This is the process referenced in the retrieved constraints.
All variables are unreliable and must be added to judgment
It means. (E) All the extracted normal constraint condition groups are evaluated, and if all the conditions are satisfied, it is determined that all the instrument groups are normal. If all the conditions are not satisfied, the process variables referenced by the group of conditions are extracted. (F) The device defining the process variable is specified and determined as an abnormal instrument. The above procedure is the instrument failure diagnosis procedure.
This is an example in which inconsistencies are detected when the column verifies.

FIG. 4 is a functional block diagram showing an example of the process diagnosis means 8. In this figure, the abnormality detection means 81 receives an inference start command from the user interface unit 4, evaluates the abnormality restriction conditions of the device, and transmits the established abnormality restriction conditions to the process abnormality inference means 82.
The process abnormality inference means 82 searches for an abnormality diagnosis method (inference method definition) corresponding to the transmitted abnormality constraint condition. Then, in accordance with the searched abnormality diagnosis method (inference method definition), diagnosis is performed by propagating between devices connected to the port. Anomaly diagnosis triggers anomaly constraints
And start the diagnosis, and the inference method
You. The method of chaining diagnostics in this way is diagnostic propagation
Is the way.

Here, the abnormality constraint condition transmitted to the process abnormality inference means 82 and the abnormality inference method to be searched are described as follows. Head:-body 1, body 2, body 3, ..., body N. The abnormal constraint condition is such that the head is an identifier (tag) and the body is an evaluation expression. When the inference is activated, the evaluation expression is evaluated. If the expression is satisfied, it means that an abnormality represented by the tag has occurred.

For example, if the abnormal constraint is a violationTAG:
-When expressed by {evalForm}, evaluation expression e enclosed by {}
When valForm is satisfied, it means that the condition of the abnormal state tag violationTAG is satisfied. Evaluation formula evalForm
The when standing, for example, process variable is greater than the upper limit value
It is time. An evaluation formula is used to determine whether the value of a process variable is normal.
An expression used to determine whether an error has occurred.
This is an expression for comparing small relations. Also, for example, when the normal constraint condition is represented by constraintTAG:-{evalForm},
When the evaluation expression evalForm is established, the normal state tag constraint
It means that the TAG condition is satisfied.

The anomaly inference method has an identification name (tag) at the head and a logical product of a plurality of elements at the body. That is, head: represented by -cond1, cond2, ..., condN, and the element cond of the main body is a tag or an evaluation expression. Here, if an abnormality occurs, the main body of the abnormality inference method having a head that matches the tag is evaluated. When an element of the main body is represented by a tag, an abnormal inference method using the corresponding tag as a head is selected, and the main body is evaluated recursively.
A tag may be activated from another inference method in its own device, or may be activated from another device through a port.

In order for the tag of the main body to propagate to the inference method of another device, it is expressed by tag @ port name. FIG. 5 is an operation conceptual diagram showing how the inference by the inference means 5 is chained. Here, a simple plant is assumed in which device A and device B are connected via output ports outA and inB.

Now, regarding the device A, the evaluation expression evalFormA1 of the anomaly constraint condition violTagA1 surrounded by a frame is evaluated,
It is assumed that the condition is satisfied. In this case, the abnormal inference method violTagA1 is activated, and its body is evaluated. In the evaluation of the body, first, the evaluation expression evalFormA2 is evaluated, and when it is satisfied, the abnormal inference method tagA1 is searched. Here, when tagA1 is searched, its body is evaluated.
If tagA1 does not exist as a result of the search, it is determined that the abnormal inference method tagA1 does not hold, and another abnormal inference method having violTagA1 in the head is searched. If it does not exist, it is determined that violTagA1 has not been established, and the inference of the cause of the abnormality has failed.

In FIG. 5, the abnormal inference method tagA1
Is searched, and the evaluation expression {evalFormA3} of its body is evaluated first, and then tagB2 {out2 is evaluated. This traces device B connected to output port outA of device A, and searches for an abnormal inference method in which the connected port is specified. Here, the abnormal inference method tag @ inB is searched. First, the anomaly inference method defined first is selected, and the inference is chained. If the evaluation expression does not hold in the middle, or
If no tag is found, the next abnormal inference method tag ＠ i
An evaluation of nB is performed. As described above, the search for the cause of the abnormality is performed by chaining the abnormality inference method in the device via the port.

FIG. 6 is a diagram for explaining a mechanism for referring to various data and attribute information handled by the device models 21, 22,. Here, as shown in the figure, the three pipes connected by a T-shaped pipe (TwoOutputPipe) P1
It is assumed that a plant is composed of flow meters FM1, FM2, and FM3, and a reference value (process amount ) of the flow meter FM1 is referred to.

Each device (flow meter FlowMeter, pipe Two)
Each port definition of OutputPipe is as follows. FlowMeter <variable> indicatedValue: Number <port> inlet flowRate: = indicatedValue outlet flowRate: = indicatedValue twoPointRoutletRport
utlet2 outlet1 flowRate: = flowRate @ inlet-flowRate @ out
let2 outlet2 flowRate: = flowRate @ inlet-flowRate @ out
let1 In the flowmeter device definition FlowMeter, the inlet of the input part and the flow of the outlet of the output part
The mechanism for referring to the value of the wRate attribute is described so as to return the indicated value indicatedValue. Also, the device definition TwoOutput for the T-shaped pipe
In tPipe, flowRa of inlet of input part
As a mechanism to refer to the value of the te attribute, the output part ou
It is described that the sum of the flowRate attribute values of “tlet1” and “outlet2” is calculated.
In the description of “outlet2”, the other output value is subtracted from the value of “inlet”.

In this way, the ou of the flow meter FM1 can be described.
To refer to the flowRate attribute value of “tlet”, first, the device connected to the “outlet” port of the flow meter FM1 and the port thereof are extracted. This is the same as the inlet port of the T-shaped pipe P1. Therefore,
In order to refer to the flowRate attribute value of the port, a mechanism for referring to the value of the T-shaped pipe P1 is applied.
At that time, the ports “outlet1” and “outlet” are further added.
2 refers to the flowRate attribute of the flow meter FM2 and the flowRate of the inlet port of the FM3 and FM3.
Go see the value of the attribute. In this way, the value reference is performed recursively to derive the final value. In deriving this value, if the port is not connected or the value cannot be referred to, the value cannot be referred.

As described above, by configuring so as to be able to refer to the attribute of the information of each device transmitted to the coupling portion of the device to be coupled, the data (value) handled by each device model
Can be constructed with good modularity. Also, there is an effect that it is sufficient to simply change the connection relationship of the devices even when the configuration of the plant is changed.

[0031]

As described above in detail, according to the present invention, the rules for abnormality diagnosis are selected according to the operating conditions of the plant, and the most appropriate rules are applied for diagnosis of various process abnormalities. As a result, accurate plant diagnosis can be performed. In the plant model, by connecting the device models via device connection means (ports), a plant to be diagnosed can be constructed with good modularity. Thus, it is possible to flexibly respond to a change in the configuration of the plant and the like.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram of the configuration of a system according to the present invention.

FIG. 2 is a diagram illustrating an example of a table used by an inference unit.

FIG. 3 is a functional block diagram illustrating an example of an instrument failure diagnosis unit.

FIG. 4 is a functional block diagram illustrating an example of a process diagnosis unit.

FIG. 5 is an operation conceptual diagram showing how inference by an inference unit is chained;

FIG. 6 is a diagram for explaining a mechanism for referring to various data and attribute information handled by a device model.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Plant database 2 Plant model 21 and 22 Device model 3 Plant model construction means 4 User interface part 5 Inference means 6 Diagnosis means 61 Instrument failure diagnosis means 62 Process abnormality diagnosis means

──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Yasuo Hirata 2-9-1, Nakamachi, Musashino-shi, Tokyo Yokogawa Electric Co., Ltd. Examiner Toru Hongo (56) References JP-A-7-174617 (JP, A JP-A-1-245332 (JP, A) JP-A-63-4433 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01M 19/00

Claims (5)

(57) [Claims] 1. A plant diagnosis system for diagnosing an abnormality of an instrument or a process constituting a plant, wherein information from the plant to be diagnosed is input, and a plant model which models the plant to be diagnosed is provided. An inference means for extracting an abnormality diagnosis rule to be applied to a plant to be diagnosed according to the operation state of a specified plant using the table and having a table in which abnormality diagnosis rules to be applied according to the operation state of the plant are described. When receives information output from the plant model, the estimated
A diagnostic means for diagnosing the plant in accordance with the abnormality diagnosis rule specified by the logical means . 2. A plant model is constructed by connecting a plurality of device models via device connection means. Each device model includes a process variable, a device-specific internal variable, an abnormality constraint condition for detecting an abnormality of the device, Normal constraint conditions that are satisfied when the device is operating normally,
2. The plant diagnosis system according to claim 1, wherein an inference method for specifying a diagnosis propagation method for a process abnormality can be described. 3. The diagnostic means includes an instrument failure diagnostic means which is convenient due to an instrument failure and a process abnormality diagnostic means which is convenient for diagnosis of a plant abnormality. A process variable retrieval unit that retrieves the process variable group referred to by the abnormal constraint condition from the plant model, and an abnormality diagnosis that is given from a table based on an instruction from the inference unit
An abnormality detecting means for evaluating a process variable extracted from a plant model according to a rule , and an abnormal instrument judging means for diagnosing an instrument failure in response to an abnormal constraint condition established by the abnormality detecting means. 1 plant diagnostic system. 4. The diagnosis means includes an instrument failure diagnosis means which is convenient due to a failure of the instrument and a process abnormality diagnosis means which is convenient for diagnosis of a plant abnormality. Anomaly detection means for receiving an inference start command to evaluate anomaly constraint conditions of a device; receiving an anomaly constraint condition established by the anomaly detection means, searching for an inference method definition corresponding to the anomaly constraint condition; 2. A plant diagnosis system according to claim 1, further comprising a process abnormality inference means for performing diagnosis by propagating between devices connected to the ports according to the following. 5. A plant model is constructed by connecting a plurality of device models via device connection means, and each device model is configured such that its connection portion can refer to an attribute of information transmitted to a connection destination. The plant diagnostic system according to claim 1, wherein: