JPH0410032A - Expert system - Google Patents

Abstract

PURPOSE:To obtain a control system or a monitor system having the extremely high reliability by providing an inference interpolating mechanism which owns a simple deciding function to perform the backward back-up when no inference result is obtained. CONSTITUTION:An inference interpolating mechanism 18 is provided in addition to an inference mechanism 17 and a knowledge base 16. The mechanism 18 owns a simple deciding function which performs the backward back-up if no inference result is obtained. Furthermore a control means 20 is added to monitor the time passed after execution of an inference and then uses the deciding result of the mechanism 18 when the passed time reaches a prescribed level. As a result, the inference result or the second best result can always be supplied within a prescribed time when an operator performs an operation by reference to the inference result of an expert system or when a controller performs the automatic control of a plant based on the inference result. Thus it is possible to construct a control system or a monitor system which has the extremely high reliability.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to improvement of an expert system used for plant operation support and automatic control.

(Conventional technology) In recent years, with the development of artificial intelligence, so-called expert systems, which use computers to realize conclusions determined by the intuition and insight of experts, are being incorporated into plant operation support and automatic control. Became.

Such expert systems, which are used for plant operation support and automatic control, collect data representing process conditions from operating plants in real time, and combine this data with a knowledge base. Inference is made by the inference mechanism in combination with There is.

(Problems to be Solved by the Invention) Real-time performance is strongly required in an expert system used in such a plant monitoring system. In particular, in an expert system that uses inference results directly to control a plant, it is important to execute inference in accordance with changes in the plant's state and to reliably obtain inference results in as short a time as possible.

On the other hand, due to the inherent characteristics of expert systems, when it comes to inferences in situations where the knowledge base is incomplete (lack of knowledge), no results can be obtained, or even in situations where inference results can be obtained, inference operations cannot be performed. Due to lack of time, results may not be obtained within the specified inference time.

For this reason, with conventional expert systems, there is a concern that even if you try to use them in a plant monitoring and control system, you may not be able to obtain inference results in time for some reason. cannot be used for

In other words, when considering the promotion of unmanned plants, intelligent operation, and ultra-automation of plant operations, in which inference results obtained in real time are used to automatically control and operate without human intervention, it is necessary to fully adopt expert systems. If a situation arises in which the inference result cannot be obtained, the control or instructions necessary at that time will not be performed at all, leading to an uncontrolled situation that is extremely dangerous and can lead to serious problems. This could lead to consequences.

Therefore, the purpose of this invention is to use a separately prepared simple judgment mechanism to obtain the next best result when an inference cannot be obtained using the original knowledge base and inference mechanism in an expert system. The purpose of the present invention is to provide an expert system that avoids situations where there is no inference result.

[Configuration of the Invention (Means for Solving the Problem) In order to achieve the above object, the present invention is configured as follows. That is, in an expert system that has a knowledge base and an inference mechanism, and uses the knowledge base to perform inference on input data by the inference mechanism to obtain a result, the inference mechanism and the knowledge base In addition, it is equipped with an inference interpolation mechanism that has a simple judgment function that provides logistical support when no inference result is obtained, and monitors the elapsed time after the inference is performed, and when the elapsed time reaches a predetermined time, the inference is The present invention is characterized by comprising a management means that uses the determination result of the interpolation mechanism.

(Function) In addition to the above-mentioned inference mechanism and knowledge base, this expert system with such a configuration is equipped with an inference interpolation mechanism that has a low-level judgment function that provides logistical support when an inference result cannot be obtained. ,When an inference operation is performed, if an inference result cannot be obtained within a fixed time even if the original inference mechanism infers using the knowledge base, the result obtained by the inference interpolation mechanism is adopted and the expert The next best P of the system is transmitted to the output target, for example, a plant monitoring and control system, and the plant is operated intelligently.

Therefore, according to the present invention, when an operator performs operation by referring to the inference results of the expert system, or when a control device automatically controls the plant based on the inference results, the expert system does not automatically control the plant within a predetermined time. Since inferred results or suboptimal results can always be provided, plant operations can be carried out with confidence under any plant condition, and extremely reliable control or monitoring systems can be constructed. , can provide expert systems that are ideal for monitoring and control.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 2 is a block diagram showing an embodiment of a plant monitoring and control system using an expert system according to the present invention. Data representing the state (process data) is imported, and this imported data is displayed on a display device (not shown), etc., and the operator 22
The plant 1 is controlled based on the operation details information of the plant 1 that is inputted by the user, etc.

Reference numeral 4 denotes an expert system, which is functionally independent from the monitoring and control device 3, and which takes in data (process data) representing the state of the plant 1 and performs inference. The result can be presented to the operator 22 via the aforementioned display device (not shown), or the monitoring/control device 3 can send a control signal to the plant 1 based on the inference result. It has become.

FIG. 3 is a block diagram showing the configuration of an electronic computer equipped with the expert system 4 shown in FIG. 2. In FIG. 3, 5 is a transmission control device, and this transmission control device 5 takes in data (process data) representing the state of the plant 1 shown in FIG. 2 and outputs it.

The keyboard 10 is an input device for inputting data, commands, etc. as a man-machine interface.
It is provided for registering and editing the knowledge base in the expert system 4 and for setting parameters related to process data processing. The clock module 6 correctly manages the inference execution cycle of the expert system 4 and outputs the time used for the start timing of the expert system 4.

The auxiliary storage device 8 stores a knowledge base, various data used by the expert system 4, inference result data obtained by inference performed by the expert system 4, and the like.

The arithmetic control device 7 performs inference based on the process data taken in by the transmission control device 5 and the knowledge base stored in the auxiliary storage device 8, and obtains an inference result. In addition, the arithmetic and control device 7 outputs the data of the inference result to the CRT display device 9 and the monitoring/control device 3 shown in FIG.

The printer 11 prints the contents displayed on the CRT display device 9 based on the drive command signal output from the arithmetic and control device 7.

FIG. 1 is a block diagram functionally showing the configuration of an electronic computer equipped with the expert system 4 shown in FIG. 3, including the flow of data inside the computer.

The process data processing unit 13, inference engine 17, inference interpolation mechanism 18, and inference result output unit 20 shown in FIG.
This is a concrete example of the functions provided by the arithmetic and control unit 7 shown in FIG. The auxiliary storage device 8 includes the process data definition file 12, process data file 14, interpolation mechanism process data file 15, knowledge base 16, and interpolation mechanism database 19.

In the process data definition file 12, various information such as registration of process data to be collected, process data collection cycle, contents of processing of process original data, etc. are registered.

The process data processing unit 13 creates a current value file, a time series file, a model calculation file, etc. based on the various information registered in the process data definition file 12, and the inference engine 17 creates each of these created files. , performs inference based on various information output from the process data file 14 and information output from the knowledge base 17 , and passes the inference result to the inference result output unit 20 .

If the inference engine 17 does not obtain an inference result within a predetermined time after starting inference, it outputs information indicating that there is no inference result.

In the interpolation mechanism process data file 15, data used by the inference interpolation function 18 is extracted from the data stored in the process data file 14, and is registered as is or after being processed.

The inference interpolation mechanism 18 in the expert system 4 is
If the inference interpolation process is executed based on the information output from the interpolation function process data file 15 or the various information row interpolation function database 19, and the inference result ``no'' signal is received from the inference engine 17. , the next best result is passed to the inference result output unit 20.

The inference interpolation mechanism 18 is a low-level judgment function compared to the inference mechanism. It is something that outputs a result that is uniquely determined according to the input data from among the input data as the next best solution, although it is not the best, or something that produces a judgment result by performing simple calculations or inferences. be.

The inference result output unit 20 obtains the results from the inference engine 17 and the inference interpolation function 18 and outputs them. Normally, the inference results received from the inference engine 17 are output to the monitoring/control device 3, but if If an inference result cannot be obtained from the inference engine 17 within a predetermined time, the next best result output from the inference interpolation function 18 is passed to the monitoring/control device 3.

In this system having such a configuration, raw process data is collected from the plant 1, the expert system 4 makes inferences based on this data, and the results are provided to the monitoring/control device 3. The monitoring/control device 3 imports raw process data from the plant 1 (data representing the state of the plant 1: process data), displays this imported data on a display device (not shown), etc., and displays it to the operator. 22 etc., the plant 1 is controlled based on the operation details information of the plant 1 input by the expert system 4, and the inference result received from the expert system 4 is presented to the operator 22 via the above-mentioned display device etc. Alternatively, based on the inference result, a control signal is sent to the plant 1 to perform control.

Here, the inference results output by expert system 4 are:
This is the original inference result output by the inference engine 17 or the second-best result output by the inference interpolation mechanism 18.

Whether the inference result output by the expert system 4 is the original inference result or the suboptimal result output by the inference interpolation mechanism 18 depends on whether the inference engine 17 outputs the inference result within a predetermined time. If the inference engine 1 outputs an inference result within a predetermined time, the original inference result is output from the expert system 4, and if the inference result is not output within a predetermined time, the expert system 4 outputs the next best result. The result will be output.

The expert system 4 has a configuration as shown in FIG. 3, and its functional configuration is as shown in FIG. The arithmetic control unit 7 includes a process data processing unit I3, an inference engine 17,
It has an inference interpolation mechanism 18 and an inference result output unit 20, and the auxiliary storage device 8 includes a process data definition file 12, a process data file 14, an interpolation mechanism process data file 15, a knowledge base 16, and an interpolation mechanism database 19. have.

Various information such as income is registered in the process data definition file 12, and the process data processing unit 13 processes the current value file, time series file, etc. based on the various information registered in the process data definition file 12. Create model calculation files, etc., perform inference on each of these created files based on various information output from the process data file 14 and information output from the knowledge base 17, and output the inference results. Pass it to Department 20.

Furthermore, for the interpolation mechanism process data file 15, data used by the inference interpolator fig is extracted from the data stored in the process data file 14,
It may be registered as is, or it may be processed and registered. The inference interpolation mechanism 18 executes inference interpolation processing based on various information output from the interpolation function process data file 15 and information output from the interpolation function database 19, and receives a signal indicating ``no inference result'' from the inference engine 17. If so, the next best result is passed to the inference result output unit 2o.

The inference result output unit 20 receives the inference result from the inference engine 17 and outputs the inference result to the monitoring/control device 3. If an inference result cannot be obtained from the inference engine 17 within a predetermined time, the next best result output from the inference interpolation mechanism 18 will be passed to the monitoring/control device 3.

As a result, the original inference result can be obtained within a predetermined time and passed to the monitoring/control device 3, thereby avoiding confusion such as the conventional no-instruction/no-control state due to inference results not being obtained. can.

The inference process of the expert system 4 having the above configuration will now be explained.

First, the entire flow of the inference function is shown in the fourth function flow diagram.
This will be explained with reference to the figures and FIG.

This system is characterized by performing two inferences: inference by the original inference engine and inference by the inference interpolation mechanism, but the following depends on whether or not these are performed simultaneously.

First Inference Flow> This is a method in which when inference by the original inference engine fails, inference is moved to inference by an inference interpolation mechanism.

That is, as shown in FIG. 4, the expert system 4 first performs the original inference execution (st2) after collecting process data (stl), and when the inference result is obtained within a specified time, it starts the inference interpolation mechanism. 18 outputs the inference result without operating (st3, 5t4). If an inference result cannot be obtained within a specified time, the inference interpolation mechanism 18 is operated to obtain and output the next best result (s
t3. st5,5ty).

In the second inference method, the inference interpolation mechanism 18 does not operate in a normal inference execution state, so the computer load is not heavy.

Second Inference Flow> This is a method in which inference by the original inference engine and inference by the inference interpolation mechanism are performed in parallel.

In this method, as shown in FIG. 5, the expert system 4 performs the original inference function and inference interpolation function simultaneously and in parallel after collecting process data (STll).
2.5t13) If an inference result is obtained by the original inference function within the specified time, the inference result by the original inference function shall be adopted.5t14゜5t15) If,
If the inference result by the original inference function cannot be obtained within the specified time, the next best result of the inference interpolation mechanism 18 is adopted (
stl4,5tlB).

In this inference method, the inference interpolation mechanism 18 is always operated in parallel with the inference engine 17, so if the inference result by the original inference function is not obtained within a specified time, the inference interpolation mechanism 18 This has the advantage that the next best result can be used.

Next, the algorithm of the inference interpolation mechanism 18 will be explained.

First interpolation method> This method uses a simple inference method that uses a subset of the knowledge base 16 used in the original inference engine 17. This method extracts important and basic knowledge from the contents of the knowledge base 16,
An interpolation function database 19 shown in FIG. 1 is constructed.

According to this method, even complex inferences can be simplified and inference results can be reliably obtained in a short time. Moreover, since the contents of the interpolation function database 19 are created based on the knowledge base 16, the development efficiency of the interpolation mechanism is high.

Second Interpolation Method> This method is based on a conventional calculation method, and constitutes the inference interpolation mechanism 18.

One example is a method of correcting the previous inference result using the current plant state as a condition, based on the previous inference result. The range to be corrected is set as a function of the plant state at the time of the previous inference and the current plant state.

Another example is a method of solving a problem by performing model calculations based on mathematical methods such as mathematical programming. In this case, a mathematical model of the plant is created, the system is identified, the original process data is input, the plant state is assumed, and arithmetic processing is performed. According to this method, by employing a completely different algorithm, which is not an inference method, as the method of the inference interpolation mechanism 18, it becomes possible to deal with a wider range of plant states.

As described in detail above, according to the present invention, when an operator performs operation by referring to the inference results of an expert system, or when a control device automatically controls a plant based on the inference results, the expert system Since it is possible to always supply inferred results or suboptimal results, plant operations can be carried out with peace of mind under any plant condition, creating an extremely reliable control or monitoring system. can.

As explained above, in the present invention, the expert system has a lower level of judgment ability than the inference mechanism in advance as a back-up support for the inference engine, but it is certain that the result of judgment that is valuable as the next best option is In addition to providing an inference interpolation mechanism that can generate a prediction, a time management means is also provided, and when the inference engine does not produce a judgment result within a predetermined time, the judgment result of this inference interpolation mechanism is used. Therefore, in a wide range of plant situations, Since it is possible to always output stable results, it increases the reliability of monitoring and control systems that utilize expert systems, and furthermore makes it easier to expand the scope of application.

In particular, when a conventional calculation method for plant management is adopted for the inference interpolation mechanism, it becomes a combination of AI (artificial intelligence) technology and conventional technology, and it is possible to interpolate the shortcomings of each and take advantage of the strengths, and also to control the expert system. When applying more advanced technology to automatic control, such as implementing automatic plant operation by applying it to a system, there is a logistics support function, so it is necessary to start from the initial stage of acquiring knowledge based on the actual plant. operation becomes possible, and the startup time of the monitoring and control system is shortened.

It should be noted that the present invention is not limited to the embodiments described above and shown in the drawings, but can be implemented with appropriate modifications within the scope without changing the gist thereof.

[Effects of the Invention] As detailed above, according to the present invention, when an operator performs operation by referring to the inference result of an expert system, or when a plant is automatically controlled by a control device based on the inference result, The expert system can always provide inference results or suboptimal results within a predetermined time, making it possible to safely operate the plant under any plant condition, resulting in an extremely reliable system. , a control system or a monitoring system can be constructed, and an expert system optimal for monitoring and control can be provided.

2 is a block diagram functionally showing the configuration of the electronic computer in which the expert system 4 is installed, including the flow of data inside the computer; FIG.
The figure is a block diagram showing an embodiment of a plant monitoring and control system using the expert system according to the present invention, and FIG. 3 is a block diagram showing the configuration of an electronic computer (computer) equipped with the expert system 4 in FIG. 19... Database for interpolation mechanism, 20... Inference result output unit.

This is a diagram.

1...Plant, 3...Monitoring/control device, 4...
Expert system, 5... Transmission control device, 6...
- Timing module, 7... Arithmetic control device, 8... Auxiliary storage device, 9... CRT display device, 10... Keyboard, 11... Printer, 12... Process data definition file, 14 ...Process data file, 15...Process data file for interpolation mechanism, 16.
...Knowledge base, 17...Inference engine, 18...
Inference Interpolation Mechanism, Applicant's Representative Patent Attorney Takehiko Suzue

Claims (1)

[Scope of Claims] An expert system having a knowledge base and an inference mechanism, wherein the inference mechanism performs inference on input data using the knowledge base and obtains a result, the inference mechanism comprising: In addition to the knowledge base, it is equipped with an inference interpolation mechanism that has a simple judgment function that provides logistical support when no inference result is obtained, and also monitors the elapsed time after the inference is performed until the elapsed time reaches a predetermined time. and a management means for using the determination result of the inference interpolation mechanism when the inference interpolation mechanism is reached.