《智能控制智能控制智能控制 (13).pdf》由会员分享,可在线阅读,更多相关《智能控制智能控制智能控制 (13).pdf(12页珍藏版)》请在得力文库 - 分享文档赚钱的网站上搜索。
1、Design Principle of Expert Control Systems0102030405Design Principles for Expert ControllersMultiplicity in model description Dexterousness in on-line processingFlexibility in control strategiesHierarchy in decision-making mechanismsReal-time inReasoning&decision-makingMultiplicity in Model Descript
2、ionThe principle of multiplicity in model description deals with applying multiple description forms of models for the controlled object and controller.The main description forms include:Analytical model01Discrete event-based model02Fuzzy model03Rule-base model 04Model-based model05Multiplicity in M
3、odel DescriptionIt is more convenient to establish the input and output fuzzy sets as well as their fuzzy relation.CFuzzy modelInvolves differential equations,difference equations,transfer functions,etc.AAnalytical modelThe model is used for a discrete system and has found even more applications in
4、analysis and design of complex systems.BDiscrete event-based modelMultiplicity in Model DescriptionThe basic form of a production rule is(IF).(THEN).DRule-base modelThis approach can reduce on-line computation by off-line pre-computation to produce simplified models individually matched to the task.
5、EModel-base modelMultiplicity in Model Description The general model of an expert controller can be represented as follows:U=f(E,K,I,G).Its basic form is:IF E AND K THEN(IF I THEN U)f is intelligent operator;E=e1,e2,.,em is the input set of the controller;K=k1,k2,.,kn is the set of experience data a
6、nd facts in knowledge base;I=I1,I2,.,Ip is the output set of inference engine;U=u1,u2,.,uq is the output set of the controller.Dexterousness in On-line ProcessingIn information utilization,the feature information should be extracted and utilized.In information storage,the feature information that is
7、 meaningful for making control decisions should be memorized.In information processing,the numerical computation and symbol operation should be combined.Dexterousness in on-line processingFlexibility in Control StrategiesThe control strategies and control parameters can be flexibly modified through
8、on-line information acquired.The adaptive strategy for handing abnormal phenomena should be designed in the expert controller.Flexibility in control strategiesHierarchy in Decision-making Mechanisms The core of intelligent control is to anthropomorphize the human intelligence.The design of intellige
9、nt controller should follow the principle of hierarchy.Hierarchy in decision-making mechanisms Real-time in Reasoning and Decision-makingThe size of knowledge base should not be too large,and the inference mechanism should be as simple as possible.Reasoning and decision-making in real-timeKeeping th
10、e structural featuresof the expert control system,butwith a smaller size of knowledgebaseandsimplerinferencemechanism.Combining with the techniqueof expert systems to increasethe decision-making level ofthe original controller,e.g.PID.Methods for Implementing the Expert ControllersTwo main methods f
11、or implementing the expert controllersSummary Design Principles for Expert Controllers Multiplicity in model description Dexterousness in on-line processing Flexibility in control strategies Hierarchy in decision-making mechanisms Real-time in Reasoning&decision-making Methods for Implementing the Expert Controllers