Adaptive knowledge inference method and system

• US 5,208,898 A
• Filed: 05/30/1989
• Issued: 05/04/1993
• Est. Priority Date: 06/08/1988
• Status: Expired due to Fees

• Current Assignee: Hitachi Control Systems Incorporated, Hitachi, Ltd.
• Original Assignee: Hitachi Control Systems Incorporated, Hitachi, Ltd.
• Inventors: Otokita, Tohru, Funabashi, Motohisa
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Downs, Robert W.
• Application Number:
  US07/359,023
• Time in Patent Office: 1,435 Days
• Field of Search: 364/513, 364/152, 364/507, 395/900, 395/61
• US Class Current: 706/45
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: G06N 7/04 Physical realisation
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

A knowledge processing system in which a grade representing a degree at which an event is satisfied or unsatisfied is obtained depending on a condition part represented in a form of a logical arithmetic expression including an expression of a fuzzy logic and on a grade of satisfaction of the condition part generates rules on assumption of unmeasurable events associated with states of the external field of the system so as to add the rules to the rules related to the unmeasurable events. Forecast of a value of a measurable event is more correctly achieved depending on grades of satisfaction of the unmeasurable events and grades of satisfaction of measurable events. A warning may also be issued according to the forecasted value.

Claims

1. A knowledge processing system in which each rule is represented in a form of a pair having a condition part and an action part so as to conduct an inference, the system comprising:

inference engine means for executing the inference;

object data storage means connected with said inference engine means for holding object data to be accessible therefrom;

rule storage means connected with said inference engine means for holding rules each including a pair of condition part related to an event to be measured and an action part specifying a grade representing a degree to which the event is satisfied based on the object data from said object data storage means and the inference to be executed by said inference engine means;

procedure storage means connected with said inference engine means for holding entities of the inference to be executed by said inference engine means based on a result supplied from said rule storage means, said inference engine means determining said condition part of one of said rules held in said rule storage means based on said object data held in said object data storage means to execute one of said entities; and

estimation rule storage means connected with said inference engine means for holding a plurality of estimation rules each including a pair of a condition part related to an event to be measured and an action part related to an event not to be measured used by an estimating means for estimating an estimation value associated with a non-measured predetermined event whereby the inference may be executed by using the estimation value as a condition part related to said non-measured predetermined event, to execute an entity of an inference based on a result supplied from said estimation rule storage means.

2. The knowledge processing system according to claim 1 further comprising forecast rule storage means connected with said inference engine means for holding a forecast rule used by a forecast means for estimating a forecast value related to an event to be measured depending on said object data in said object data storage means and said value in said estimation rule storage means.

3. The knowledge processing system according to claim 2 further comprising a warning rule used by a warning means for issuing a warning, the warning rule including an executing procedure for another action part which issues the warning and a condition part having a condition satisfied when said forecast value in said forecast rule storage means is outside a predetermined range held in a rule table and determined by said inference engine means.

4. The knowledge processing system according to claim 1 wherein said estimation rule storage means includes rule means including an action part related to an event to be measured and a condition part related to a comparison of an event to be measured and an event not to be measured to obtain a grade representing a degree of satisfaction of the event to be measured.

5. The knowledge processing system according to claim 2 wherein said forecast means produces said forecast value of the event to be measured depending on a result in said estimation rule storage means and on a result produced from said rule group means.

6. An adaptive knowledge inference method in a knowledge processing system comprising the steps of:

holding in an object data storage connected with an inference engine object data to be accessible therefrom;

holding in a rule storage connected with said inference engine a plurality of rules each including a pair of a condition part related to an event to be measured and an action part specifying a grade representing a degree to which the event is satisfied based on the object data from said object data storage and an inference to be executed by said inference engine;

holding in a procedure storage connected with said inference engine entities of an inference to be executed by said inference engine based on a result supplied from said rule storage said inference engine determining said condition part of one of said rules held in said rule storage based on said object data held in said object data storage in order to execute one of said entities,

said condition part represented with logical arithmetic expressions including those of fuzzy logic, said action part supplies

a conclusion related to grades representing degrees to which a plurality of events are respectively satisfied or unsatisfied and/or issues an instruction to execute a plurality of procedure processings such that knowledge associated with an object problem is represented and is described by use of the plurality of rules;

representing a described rule of the plurality of rules, as having a condition part and an action part including an event related to a non-measurable event and another described rule of the plurality of rules, as having a condition part including said non-measurable event and an action part including an event related to a measured event;

inferring by the inference engine a grade representing a degree at which an event corresponding to an unmeasurable state is satisfied according to the action part of at least one of the described rules of the plurality of rules based on grades respectively representing degrees at which events corresponding to measurable states are respectively satisfied when drawing a final conclusion, by establishing correspondences between states of an external field of the system and events, in a case where the states of the external field of the system for which correspondences are established with respect to the events comprise knowledge;

processing the inference by repeatedly applying the rules depending on the grades of satisfaction of the events; and

,

drawing the final conclusion by use of a result of the inference.

7. The adaptive knowledge inference method in a knowledge processing system according to claim 6 further comprising the steps of:

drawing the final conclusion by establishing correspondences between measured values of states and events in an external field of the system;

permissibly drawing a wrong final conclusion depending on the states of the external field of the system;

setting an assumed event after drawing said wrong final conclusion for adding rules associated with the assumed event and inferring a grade of satisfaction of the assumed event from grades of satisfaction of events determined in association with measurable states of the external field of the system by use of at least a portion of described rules;

thereby drawing the final conclusion by use of a result of the inference.

8. The adaptive knowledge inference method in a knowledge processing system according to claim 6 further comprising the steps of:

including causality knowledge associated with time in said rules and a forecast of a grade of satisfaction of an event in a portion of the final conclusion to be drawn;

preparing a forecast rule describing the causality knowledge with respect to time for said unmeasurable event corresponding to said unmeasurable state and/or for the assumed event inferring present values of grade of satisfaction of the unmeasurable event and the assumed event based on grades of satisfaction of events determined by the measurable states up to a present point of time and by use of at least a portion of the described rules;

inferring and forecasting the grade of satisfaction of the assumed event based on a result of the inferring present values and the forecasted rule; and

,

thereby obtaining a forecasted value of a grade of satisfaction of a final conclusion event by use of a result of the forecast and the inferring present values and the grades of satisfaction of the event determined by the measurable state up to the present point of time.

9. The adaptive knowledge inference method in a knowledge processing system according to claim 6 wherein for at least a portion of events of which grades of satisfaction can be attained through an adaptation of the rules, quantitative values are estimated for states of the external field of the system based on the grades of satisfaction of events, thereby furthering inference processing by use of a result of the estimation.

10. An adaptive knowledge inference method in a knowledge processing system according to claim 6 wherein in a case where an inference is effected for a grade of an unmeasurable event or an assumed event or where a quantitative value is obtained for a state of an external field of the system corresponding to said event, if the inferred grade of satisfaction or the quantitative value is beyond a predetermined range of values or is different from a predetermined value, a warning or a notification is issued to a system user or/and a subsequent inference is continuously executed by use of a predetermined default value.

11. The adaptive knowledge inference method in a knowledge processing system according to claim 6 wherein said entities of the procedure storage are described in rule knowledge and supplied to the knowledge processing system.

12. The adaptive knowledge inference method in the knowledge processing system according to claim 6 further comprising the steps of:

analyzing said grade of said non-measurable event;

estimating a quantitative value obtained for a state of said external field to produce a result;

issuing a warning to the knowledge processing system if said analyzing and estimating produces a result beyond a predetermined range of values; and

,

continuously executing subsequent inferring until said result is produced within said range.

13. A knowledge processing system in which each rule is represented in a form of a pair having a condition part and an action part so as to conduct an inference, comprising:

inference engine means for executing the inference;

object data storage means connected with said inference engine means for holding object data to be accessible therefrom;

rule storage means connected with said inference engine means for holding rules each including a pair of a condition part related to an event to be measured and an action part specifying a grade representing a degree to which the event is satisfied based on the object data from said object data storage means and the inference to be executed by said inference engine means;

procedure storage means connected with said inference engine means for holding entities of said processing to be executed by said inference engine means based on a result in said rule storage means;

said inference engine means receiving sensor data from sensor means connected thereto, based on the processing described in said rule storage means so as to update a corresponding part of the object data held in said object data storage means said inference engine means determining said condition part of one of said rules held in said rule storage means based on said object data held in said object data storage means to execute one of said entities; and

,

estimation rule storage means connected with said inference engine means for holding a plurality of estimation rules each including a pair of a condition part related to an event to be measured and an action part related to an event not to be measured used by an estimating means for estimating an estimation value associated with a non-measured predetermined event;

wherein said inference engine means executes the estimation rule based on said updated object data.

14. The knowledge processing system according to claim 13 further comprising forecast rule storage means connected to said inference engine means for holding a forecast rule used for estimating a forecast value related to an event to be measured depending on said object data in said object data storage means and said value in said estimation rule storage means.

15. The knowledge processing system according to claim 14 further comprising a warning rule including an executing procedure for another action part for issuing a warning and a condition part having a condition satisfied when said forecast value in said forecast rule storage means is outside a predetermined range held in a rule table and determined by said inference engine means.

16. The knowledge processing system according to claim 13 wherein said estimation rule storage means includes rule means including an action part related to an event to be measured and a condition part related to a comparison of an event to be measured and an event not to be measured to obtain in a rule implementation means a grade representing a degree of satisfaction of the event to be measured.