JPS61156461A - Method and device for obtaining knowledge - Google Patents

Abstract

PURPOSE:To sample truly control (judgment) knowledge which is difficult to make a form and documentate by giving the place to judge by simulation, etc., to a specialist (controller), making the specialist exchange and judge the information actually. CONSTITUTION:When knowledge is obtained, as the first means to sample the knowledge concerning the object field of a control system an and the judging knowledge in the system, respective functions of an obtaining means 104 of the judging rule, a memory area 105 of the judging rule and a judging rule model 106 are used. The knowledge sampled at first means is arranged, a logical inconsistency and a functional error are detected, as the second means to correct it, respective functions of a structure means 107 of judging knowledge and a knowledge model 109 are used. A control work is executed independently from a knowledge base control system, the flow of the control and the flow of the information are sampled related to the process of the control work, and as a means to record, an obtaining means 10 of the knowledge concerning the judging work, a memory area 102 of the work knowledge and a work model 103 are used.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a knowledge acquisition method and disguise, and in particular, a knowledge acquisition method suitable for faithfully acquiring judgment work knowledge for various and difficult to formalize control systems. METHODS AND APPARATUS.

[Background of the invention]

Conventionally, knowledge acquisition technology in fields where knowledge engineering has been applied or tested has been based on system designers gathering knowledge from conversations with experts in the control target or from literature, and adding logical inconsistencies and functional problems to the content. It was sufficient to check for errors and make corrections. For example, in the medical field and the new material development field, where knowledge engineering efforts are relatively advanced, it is relatively easy to have a dialogue between field experts and system designers, and various types of knowledge can be directly acquired from the dialogue. can.

However, in a complex air traffic control system, the above-mentioned dialogue becomes difficult, and it is difficult to acquire various types of knowledge through so-called dialogue. One of the reasons for that.

This is because the knowledge and decision-making knowledge about air traffic control objects possessed by air traffic controllers is generally based on personal experience and is not formalized or documented (or difficult to do so). Therefore, in order to spread the use of knowledge engineering to various fields, it was necessary to build knowledge related to control target fields and judgments for acquiring judgment knowledge, that is, business knowledge (business models). (Araya et al. “Expert system for traffic rescheduling” National Institute of Electrical Engineers of Japan Conference #1454
゜S58. Nishiyo et al. [Support system for rule acquisition in traffic rescheduling] Institute of Electrical Engineers of Japan National Conference #1343. (See S59) [Object of the Invention] The purpose of the present invention is to solve such conventional problems and to improve the ability of air traffic control experts (
When faithfully acquiring various types of knowledge from air traffic controllers), Formalization 9 Knowledge that allows you to collect valuable control (judgment) knowledge in documentation III! An object of the present invention is to provide a method and apparatus for obtaining IF.

[1 Requirement of the Invention] To achieve the above object, 1 knowledge of the present invention! ! 1lffl acquisition method and device include a collection means for collecting knowledge and judgment knowledge regarding the control target of the control system, and organizing the knowledge collected by the collection means, detecting and correcting logical contradictions and functional errors. A knowledge acquisition device comprising a modifying means and a storage means for storing knowledge modified by the modifying means as a knowledge base.

When acquiring the above judgment knowledge, the controllers of the multiple subsystems of the air traffic control system are made to perform pretend or actual air traffic operations independent of the processing of the air traffic control system, and the control (starting f's) during the execution is performed. and the flow of information (i.e., source and content) is collected in relation to the process of air traffic control operations, and the collected knowledge is compared with the knowledge base of the storage means to correct the knowledge base. .

(Example of the invention) Embodiments of the present invention will be described below with reference to the drawings.

FIG. 3 is a configuration diagram of a knowledge acquisition function showing the positioning and characteristics of the present invention.

As shown by the dashed line in Figure 3, each function (104
to 109) are included in the conventional acquisition technique, and the upper functions (101 to 103) are insufficient in the conventional technique, that is, the functions of the present invention.

When acquiring knowledge, the first means for acquiring knowledge regarding the target field of the control system and judgment knowledge in that system is the judgment rule acquisition means 1042.
Judgment rule storage area 105, judgment rule model 106
A second means for organizing the knowledge collected in the first means, detecting logical contradictions and functional errors, and correcting them using each of the functions described above is a judgment knowledge construction means 10.
7. Judgment knowledge (knowledge base) 108 is used as a means for editing the various kinds of knowledge obtained by the second means using each function of the knowledge model 109 and storing it in the knowledge storage area as the knowledge base of the control system. .

In addition, the control operations are executed independently of the knowledge-based control system, and the control (startup) flow (i.e.

As a means for collecting and recording the flow of information (i.e., source and content) in relation to the process of air traffic control operations, there is a means for acquiring knowledge regarding judgment operations, which is a function of the present invention. 101. Business knowledge storage area 102. A business model 103 is used. Note that the contents of each function will be described later using specific examples.

FIG. 1 is a functional configuration diagram of the first knowledge acquisition subsystem in a distributed control system showing an embodiment of the present invention, and FIG. 2 is a configuration diagram between each knowledge and each subsystem.

In FIGS. 1 and 2, 10 is the control target system, 1
0i is the i-th control target subsystem.

11, lli is a control target simulator for producing various situations of the control target that require judgment by the control system, and 20i is a storage area that stores various knowledge regarding the control target subsystem 10i and its control (judgment) method. and,
Control device subsystem with judgment function, 23-1”n
, 23 +, 63-1 ” n , 63 i + 6
3T-1"n, 63Ti is a transmission control unit,
30i is a terminal bag VFi (display, keyboard) for the system designer's interaction with the air traffic control subsystem 20i and the knowledge acquisition system, 30Ti is an audio interaction device, 33-1 to n, 331 are actual communication terminals used by the air traffic controller. Control of the control target or the performance of the simulator,
and a control mode changeover switch that allows the control mode to be changed from the dialog terminal device 30i, such as control of an actual control target by the control target subsystem +04 or control of a simulator, etc.; 351 is a knowledge acquisition subsystem; In order to store the collected judgment knowledge in the control (judgment) knowledge storage area of the control device subsystem 20i, a knowledge base construction unit 501 is constructed based on commands from the dialogue terminal device 301. Terminal equipment for controller interaction (display, keyboard),
50Ti is a device for voice interaction with other air traffic controllers, that is, a telephone handset, and 55i is a display device that displays various information necessary for air traffic control decisions and provides air traffic controllers with a "place of failure". .60i and 60Ti are terminal devices for air traffic controllers to interact with!
50i, the handset 50T i , and the display device 55i to monitor the activation flow (source and its content) and information flow (source and its content) in the process of executing the judgment,
61i is a monitoring/collection unit that collects judgment knowledge; 61i is a timer part that records the startup flow and information flow time; 62
i is processed into data for judgment and sent to the monitoring/collection unit 60, 60
Data processing units 65i and 65Ti are storage units for storing the judgment knowledge etc. collected by the monitoring and collection units 60 and 60T, and 67T is a transmission control unit 23 and 63 of each knowledge acquisition subsystem. .63T, transmission lines for signals and audio signals (between multiple controllers and control subsystems), respectively, 80-1-n is the control mode changeover switch 331 in Figure 1, and the control target subsystem. It includes everything except the LO 4° control target simulator Lli, the transmission control units 231° 63i, 63Ti, and the transmission line 67.67T, and is a knowledge acquisition subsystem.

The main contents handled by the system designer using the display and keyboard of the dialogue terminal devices 30i and 30Ti are as follows.

(1) Switching the control mode using the control mode changeover switch 33i.

(1:) Starting and terminating the control target simulator lit.

(iii) Monitoring of judgment content in each control mode.

(iv) Judgment knowledge collected from the starting flow of the judgment process by the controller and the flow of information, that is, the storage units 65i, 65
Replaying the contents of Ti.

(V) Modification of the contents of the storage section 65i.

(vl) Modification of the control knowledge within the control device subsystem 20i based on the contents of the storage unit 65i.

As a display example on the display device 55i, FIG. 4 shows timetable rescheduling control in a single train operation management system. The display contents are as follows from top to bottom.

(1) Display of number line usage timetables in designated order (55a).

(11) Display of train operation schedule (55b).

(iii) Display of vehicle or crew operating schedule (5)
5c, 55d).

(1v) Display of other related information (55e).

However, in the diagram, 1 to 5 are station numbers, 81 to S4 are stations, a
2-a4. b12. clol-c104 indicates a train.

Each of the above-mentioned displays 55a to 55d is in a diamond format having a common horizontal time axis, and 55c and 55d are in a horizontal bar representation.

In addition, when the air traffic controller operates the above display contents and ranges, the controller uses the keyboard of the interactive terminal bag [50i].The contents (a) to (d) will be described below.

(a) Input to the next judgment event in the judgment process.

(1) Event name (11) Target of judgment (related train name, point line, etc.).

(iii) Evaluation scale of judgment and its weighting (loss to each train passenger concerned).

(1■) Constraint conditions for judgment and their weighting (vehicle inspection plan, availability of backup crew, etc.).

(b) Content specification in the judgment process Whether the next process is abnormality detection, situational judgment, or search for an alternative.
The controller will specify whether it is a judgment decision or not.

(c) Request for (display) judgment information (1) Request for display of primary information (raw information such as the train schedule at the stage of rescheduling, abnormal phenomena on the schedule, etc.).

(ii) Data processing unit 62 for secondary information (train schedules, predictions from the current situation, predictions for change inputs, etc.)
Request for display of information generated by i.

(iii) Requests for related information to the system (train depot location status, inspection plans for related vehicles from that day to the next day, seat reservation status for related trains, designated transit passenger status, operational schedules for various crew members, etc.).

(1v) Dialogue with another controller (exchanging information with controllers for each district designation, FLE Vehicle Directive, General Passenger Ministerial Ordinance, etc.).

(d) Input of values related to judgment, input of decisions (evaluation value of judgment, decided operation rescheduling room (change procedure)).

The monitoring/collection unit 60i is connected to the terminal bag M3゜i, 50i for interaction.
, according to instructions from the timer part 61i.

A processing flow as shown in FIG. 5 is executed. Note that (1) to (1v) below are functional contents.

(1) Control of the judgment process of air traffic controllers (dispatchers).

(11) Decision sequence of dispatcher (individual decision events in each decision process (for example, sequence control of train departure order decision).

(iii) The controller (dispatcher) uses a terminal device for interaction 50t
And display device! When sending and receiving instructions and information with another dispatcher using the 55i, the flow of activation for instructions, the flow of information during sending and receiving, etc. are organized in the order of occurrence and stored in the storage units 65i and 65Ti. Store.

(1v) The judgment events, request information, and dialog information for the activation factor being judged are organized according to their frequency of occurrence, and stored in the storage units 65i and 65Ti in the same manner as in (iii) above.

The activated monitoring/collection unit 60i performs a mode analysis based on the contents of the instruction and learns! Steps 200° 210-21 to determine whether the mode is 1IN1ft acquisition mode, reproduction of knowledge content, formalization of knowledge, or modification of knowledge within the system.
4) If neither of these is the case, process the abnormality and end the process (step 219).

In the case of knowledge acquisition mode, control mode changeover switch 33
In response to the control task selected in step i, the judgment process and judgment sequence for the dispatcher is displayed on the display device 55i, and the dispatcher monitors and collects various information and judgment status for that "situ". are stored in the storage units 65i and 65Ti, and are stored in a predetermined format of air traffic control operations or actual air traffic control.

Alternatively, the display device 15 may be
The judgment process and judgment sequence, which are the display contents of 5i, are updated one after another, and the collection is repeated to acquire the judgment knowledge of the dispatcher (steps 220 to 233).

In the case of knowledge content reproduction and knowledge formalization, functions such as knowledge content reproduction function and knowledge formalization correction are operated, respectively.
Execute (steps 215, 216).

In the case of modifying the internal knowledge R (knowledge base) of the control device subsystem 20i, the monitored and collected knowledge is stored in the storage units 65i, 65.
The original content is updated by transferring it from Ti to the control (judgment) knowledge storage area of the control Mffi subsystem 201 and storing it (steps 217 and 218).

Next, the effects of the train operation management system described as an embodiment of the present invention in the case of timetable rescheduling will be described below.

(i) In the operation rescheduling work of the train operation management system,
Since the dispatcher executes the process while looking at the content displayed on the display device 55i (where decisions are made for timetable rescheduling), the dispatcher can execute the process with almost the same judgment as that made by a single person.

(11) Since the dispatcher monitors and collects the activation flow and information flow in the judgment process performed in (1) above, by analyzing it, individual judgments that have not been formalized or documented in the past can be made. Knowledge 1: Judgment knowledge such as knowledge about judgment sequences can be faithfully acquired.

(iii) The extent of the relationship between the decision-activating factors that have occurred by reproducing and analyzing the decision knowledge in (1:) above and the information necessary for making a series of individual decision events.

In addition, it is possible to know the relationship between information related to judgments, such as temporal and spatial ranges.

(:V) Information is collected according to the judgment process and judgment sequence between the dispatchers (dialogue by each terminal device and telephone conversation) exchanged by function and subject, so that other subsystems respond to individual judgment events. You can see the relationship with the directive.

〔Effect of the invention〕

As explained above, according to the present invention, when faithfully acquiring various types of knowledge from air traffic control experts (air traffic controllers) in order to apply knowledge engineering to air traffic control systems, the above experts (air traffic controllers) By providing an opportunity to make judgments through simulations, etc., and having them actually exchange information and make decisions, it is possible to faithfully collect control (judgment) knowledge that is difficult to formalize or document.

[Brief explanation of drawings]

Fig. 1 is a functional block diagram of the first knowledge acquisition subsystem showing an embodiment of the present invention, Fig. 2 is a block diagram of each knowledge/subsystem, and Fig. 3 shows the positioning and characteristics of the present invention. FIG. 4 is a display example in the case of timetable rescheduling in the train operation management system, and FIG. 5 is a processing flowchart of the monitoring/collection unit 60i. 10: Control target system, jOi: i-th control target subsystem, 11. lli: Control target simulator, 201: Control device subsystem, 23-1-n, 23
i, 63-1-n, 63 i, 63T-1-n,
63T i :Transmission control unit, 30i. 501: Dialogue terminal device, 30Ti, 50Ti: Dialogue device, 33-1 to n, 33i: Control mode changeover switch,
35i: Knowledge base construction unit, 551: Display device, 55a
~55e: Display example, 60i, 60Ti: vL viewing/collection section, 61i: 3+ now part, 621: Data processing section, b
5 i + 65 T: : Storage section, 67.67T
: Transmission path between multiple controllers and control subsystems, 80-1 to n: Knowledge acquisition subsystem, lot, 10
4: yfl obtaining means, 102° 105: Storage area, 103
: Business model, 106: Judgment rule model, 107: *
Construction means, 108: Judgment knowledge, 109: Knowledge model, 11
0: Controller. 111 Niji system designer. Patent applicant: Hitachi, Ltd. Patent attorney Masatoshi Isomura 7゛1゜Figure 3: Cow diagram

Claims (4)

[Claims] (1) A collection means for collecting knowledge and judgment knowledge regarding the control target of the control system, a correction means for organizing the knowledge collected by the collection means, detecting logical contradictions and functional errors, and correcting them; In a knowledge acquisition device having a storage means for storing the knowledge modified by the modification means as a knowledge base, when acquiring the judgment knowledge, the controller of the plurality of subsystems of the control system is given a pretense that is independent of the processing of the control system. , or perform actual air traffic control operations.
A knowledge acquisition method characterized by collecting the control and information flow being executed in association with the process of control operations, comparing the collected knowledge with the knowledge base, and correcting the knowledge base. (2) Collect the source and content of the interaction between the air traffic controller and another air traffic controller using voice or code in association with the process of air traffic control operations, and compare the collected knowledge with the knowledge base. 2. The knowledge acquisition method according to claim 1, wherein the knowledge base is modified based on the knowledge base. (3) Comparison with the knowledge base is performed by reproducing the collected knowledge, that is, the control and information flow, the content and source of interaction information, in synchronization with the process of air traffic control operations, and based on the content, the knowledge base is The knowledge acquisition method according to claim 1 or claim 2, characterized in that the knowledge acquisition method comprises: modifying the knowledge acquisition method; (4) a collection means for collecting knowledge and judgment knowledge regarding the control target of the control system; a correction means for organizing the knowledge collected by the collection means and detecting and correcting logical contradictions and functional errors; A knowledge acquisition device having a storage means for storing the knowledge corrected by the correction means as a knowledge base, comprising: a simulation generation means for generating a fake air traffic control operation for an air traffic controller in order to acquire the judgment knowledge; control mode switching means for switching to real or sham control, and monitoring/gathering means for monitoring control and information flow in the process of air traffic control operations and interaction information with another air traffic controller to collect knowledge of operations; A knowledge acquisition device characterized by comprising a display means that displays various information on air traffic control and provides an opportunity for an air traffic controller to make a decision.