《学位论文-—模糊控制的理论与发展概述.doc》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
模糊控制考核论文模糊控制的理论与发展概述摘要模糊控制理论是以模糊数学为基础,用语言规则表示方法和先进的计算机技术,由模糊推理进行决策的一种高级控制策。模糊控制作为以模糊集合论、模糊语言变量及模糊逻辑推理为基础的一种计算机数字控制,它已成为目前实现智能控制的一种重要而又有效的形式尤其是模糊控制和神经网络、遗传算法及混沌理论等新学科的融合,正在显示出其巨大的应用潜力。实质上模糊控制是一种非线性控制,从属于智能控制的范畴。模糊控制的一大特点是既具有系统化的理论,又有着大量实际应用背景。本文简单介绍了模糊控制的概念及应用,详细介绍了模糊控制器的设计,其中包含模糊控制系统的原理、模糊控制器的分类及其设计元素。关键词:模糊控制;模糊控制器;现状及展望AbstractFuzzycontroltheoryisbasedonfuzzymathematics,usinglanguagerulerepresentationandadvancedcomputertechnology,itisahigh-levelcontrolstrategywhichcanmakedecisionbythefuzzyreasoning.Fuzzycontrolisacomputernumericalcontrowhichbasedfuzzysettheory,fuzzylinguisticvariablesandfuzzylogic,ithasbecometheeffectiveformofintelligentcontrolespeciallyintheformoffuzzycontrolandneuralnetworks,geneticalgorithmsandchaostheoryandothernewintegrationofdisciplines,whichisshowingitsgreatpotential.Fuzzycontrolisessentiallyanonlinearcontrol,andsubordinatesintelligentcontrolareas.Amajorfeatureoffuzzycontrolisbothasystematictheoryandalargenumberoftheapplicationbackground.Thisarticleintroducessimplytheconceptandapplicationoffuzzycontrolandintroducesdetaillythedesignofthefuzzycontroller.Itcontainstheprinciplesoffuzzycontrolsystem,theclassificationoffuzzycontrolleranditsdesignelements.Keywords:FuzzyControl;FuzzyController;StatusandProspects.引言传统的常规PID控制方式是根据被控制对象的数学模型建立29 模糊控制考核论文,虽然它的控制精度可以很高,但对于多变量且具有强耦合性的时变系统表现出很大的误差。比例调节是根据被调量和设定值之间的差值来变化的,也就是说比例控制中余差不可避免。积分调节最终实现无余差调节,但是超调比较大。模糊控制是建立在人工经验基础之上的,它能将熟练操作员的实践经验加以总结和描述,并用语言表达出来,得到定性的、精确的控制规则,不需要被控对象的数学模型。并且模糊控制易于被人们接受,构造容易,适应性好。TheintroductionTraditionalwayofconventionalPIDcontrolwasestablishedaccordingtothemathematicalmodelofcontrolledobject,althoughitcanbeveryhighcontrolprecision,butforthemulti-variable29 andtime-varyingsystemswithstrongcouplingshowedgreaterror.Proportionalcontrolisbasedonthedifferenceinvaluebetweensetvalueandquantityofthemodulatedtochange,thatistosay,proportionalcontrolofresidualisinevitable.Integralregulationachieveeverythinginaglanceatpoorregulation,buttheovershootisbigger.Fuzzycontrolisbasedontheartificialexperience,itcanskilledoperator'spracticalexperiencesummarizedanddescribed,andthelanguageexpression,getaqualitative,precisecontrolrules,notneedmathematicalmodelofcontrolledobject.Andfuzzycontroliseasytobeacceptedbypeople,easystructure,goodadaptability.第一章模糊控制概述1.1模糊控制的概念及应用“模糊”是人类感知万物,获取知识,思维推理,决策实施的重要特征。“模糊”比“清晰”所拥有的信息容量更大,内涵更丰富,更符合客观世界。模糊逻辑控制(FuzzyLogicControl)简称模糊控制(FuzzyControl),是以模糊集合论、模糊语言变量和模糊逻辑推理为基础的一种计算机数字控制技术。模糊控制理论是由美国著名的学者加利福尼亚大学教授Zadeh·L·A于1965年首先提出,它是以模糊数学为基础,用语言规则表示方法和先进的计算机技术,由模糊推理进行决策的一种高级控制策。在1968~1973年期间Zadeh·L·A先后提出语言变量、模糊条件语句和模糊算法等概念和方法,使得某些以往只能用自然语言的条件语句形式描述的手动控制规则可采用模糊条件语句形式来描述,从而使这些规则成为在计算机上可以实现的算法。1974年,英国伦敦大学教授Mamdani·E·H研制成功第一个模糊控制器,并把它应用于锅炉和蒸汽机的控制,在实验室获得成功。这一开拓性的工作标志着模糊控制论的诞生并充分展示了模糊技术的应用前景。模糊控制实质上是一种非线性控制,从属于智能控制的范畴。模糊控制的一大特点是既具有系统化的理论,又有着大量实际应用背景。模糊控制的发展最初在西方遇到了较大的阻力;然而在东方尤其是在日本,却得到了迅速而广泛的推广应用。其典型应用的例子涉及生产和生活的许多方面,以下为模糊控制在工业和生活方面的一些应用实例:29 1净水场药品注人控制、上下水道处理系统2各种溶沪:电气炉水泥生成炉的控翻、原子能发电供水控制、金属板成形控制3城市垃圾焚烧炉的控制4随道盾构机械、油压掘进机械、集装箱吊装的控制5高速公路隧道的排气、换气控制6汽车的定速行走控制、发动机的控制、模糊AT、自动观光船7飞机离着陆控制、直升飞机控制、海上救难船控制8机器人的控制:扫除机械人、花道机械人、激光切割机器人、钓鱼机器人、9升降机群管理、自动枪票机、自动门开关装置、自动贩卖机10空调控制制冷、制热机、多路空调系统、铁道车辆、11造纸机、清酒酿造控制12自动声音调整器、传感器位置选择13电视会议系统、簇像机、电子喷水器、录像机、照像机、复印机、绘图机29 14家电制品:洗衣机、吸尘器、干澡机、冷藏箱、电子微波炉,电饭锅、电动剃须刀、1.2模糊控制的优点1简化系统设计的复杂性,特别适用于非线性、时变、模型不完全的系统上。2利用控制法则来描述系统变量间的关系。3不用数值而用语言式的模糊变量来描述系统,模糊控制器不必对被控制对象建立完整的数学模式。4模糊控制器是一语言控制器,使得操作人员易于使用自然语言进行人机对话。5模糊控制器是一种容易控制、掌握的较理想的非线性控制器,并且抗干扰能力强,响应速度快,并对系统参数的变化有较强的鲁棒性和较佳的容错性。6从属于智能控制的范畴。该系统尤其适于非线性,时变,滞后系统的控制。1.3模糊控制的缺点1模糊控制的设计尚缺乏系统性,这对复杂系统的控制是难以奏效的。所以如何建立一套系统的模糊控制理论,以解决模糊控制的机理、稳定性分析、系统化设计方法等一系列问题;2如何获得模糊规则及隶属函数即系统的设计办法,这在目前完全凭经验进行;3信息简单的模糊处理将导致系统的控制精度降低和动态品质变差。若要提高精度则必然增加量化级数,从而导致规则搜索范围扩大,降低决策速度,甚至不能实时控制;4.如何保证模糊控制系统的稳定性即如何解决模糊控制中关于稳定性和鲁棒性问题还有待完善。Thefirstchapterissummaryoffuzzycontrol29 1.1theconceptandapplicationoffuzzycontrol"Fuzzy"humanperceptioniseverything,toacquireknowledge,thinking,reasoning,decision-makingofimportantfeatures."Fuzzy"than"clear"havetheinformationcapacityofalarger,moreabundantconnotation,moreinlinewiththeobjectiveworld.FuzzyLogicControl(FuzzyLogicControl)referredtoas"FuzzyControl(FuzzyControl),basedontheFuzzysettheory,FuzzylanguagevariableandFuzzyLogicreasoningisthebasisofacomputernumericalControltechnology.FuzzycontroltheoryisbythefamousscholarattheuniversityofCaliforniaprofessorZadeh,l.a.firstproposedin1965,itisbasedonfuzzymathematics,expressedinthelanguagerulesmethodandadvancedcomputertechnology,bythefuzzyreasoningtomakedecisionsofanadvancedcontrolstrategy.Duringtheperiodof1968~1973,Zadeh,l.a.successivelyproposedlanguagevariable,fuzzyalgorithmandthefuzzyconditionalstatementconceptsandmethods,makesomeofthepastcanonlyusenaturallanguageformofconditionalstatementsdescribethemanualcontrolrulescanbeusedtodescribefuzzyconditionalstatementform,soastomaketheserulescanbeimplementedoncomputeralgorithm.In1974,aprofessorattheuniversityofLondonMamdani,E,H,successfullydevelopedthefirstfuzzycontrollerandapplyittotheboilerandthecontrolofthesteamengine,tosucceedinthe29 laboratory.Thispioneeringworkmarksthebirthofthefuzzycontroltheoryandfullyshowstheapplicationprospectoffuzzytechnology.Fuzzycontrolisessentiallyakindofnonlinearcontrol,frombelongstothecategoryofintelligentcontrol.Fuzzycontrolisoneofthebiggestcharacteristicisbothasystematictheory,andwithalargenumberofpracticalapplicationbackground.Thedevelopmentoffuzzycontrolisfirstencounteredinthewestthelargerresistance;Intheeast,especiallyinJapan,however,hasobtainedtherapidandextensivepopularizationandapplication.Itstypicalapplicationexampleinvolvesmanyaspectsofproductionandlife,thefollowingissomeapplicationsoffuzzycontrolintheaspectofindustrialandlivingexample:1Druginjectionwaterpurificationfieldpopulationcontrol,sewagetreatmentsystem2AllkindsofsolubleShanghai:cementgeneratedinelectricfurnacecontrolturn,nuclearpowergeneration,watersupply,sheetmetalformingcontrol3Controlofurbangarbageincinerator4Thetunnelshieldmachine,hydraulicexcavatingmachinery,containerliftingcontrol5Exhaustandventilationcontrolofhighwaytunnel29 6Inconstantspeedcontrol,automotiveenginecontrol,automaticfuzzyats,sightseeingboats7Aircraft,helicoptersfromlandingcontrolcontrol,maritimerescueboat8Therobot'scontrol:cleaningrobots,ikebanarobots,fishing,lasercuttingmachine,9Liftthefleetmanagement,gunticketmachine,automaticswitchingdevice,automaticvendingmachines10Refrigerationandairconditioningcontrolsystemofengine,multi-channelairconditioningsystem,railwayvehicle,11Papermachine,winebrewingcontrol12Automaticsoundregulator,sensorlocationchoice13Videoconferencesystem,clusterlikemachine,electronicsprinkler,VCR,camera,copymachine,drawingmachine14Homeappliancesproducts:washingmachines,vacuumcleaners,drybathmachine,cooler,electronicoven,electricricecooker,electricrazor1.1theadvantagesoffuzzycontrol1tosimplifythecomplexityofsystemdesign,especiallyfornonlinear,time-varyingandmodelincompletesystem.29 2controllawisusedtodescribetherelationshipbetweenthesystemvariables.3languageinsteadofnumericaltypeoffuzzyvariablestodescribethesystem,thefuzzycontrollerdon'tneedtoestablishacomprehensivemathematicalmodelofcontrolledobject.4controller,fuzzycontrollerisalanguagethatoperatorseasytousenaturallanguagefortheman-machinedialogue.5fuzzycontrollerisakindofeasytocontrolandmasteryoftheidealnonlinearcontroller,andstronganti-jammingcapability,fastresponsespeed,andthechangeofsystemparametershasstrongrobustnessandbetterfaulttolerance.6frombelongstothecategoryofintelligentcontrol.Thissystemisespeciallysuitablefornonlinear,timevaryingandlagcontrolsystem.1.3thedisadvantageoffuzzycontrol1thedesignofthefuzzycontrolisstilllackofsystematic,thecontrolofcomplexsystemsisdifficulttowork.Sohowtoestablishasystemoffuzzycontroltheory,inordertosolvefuzzycontrolmechanism,stabilityanalysis,systematicdesignmethodforaseriesofproblems;2howtoobtainfuzzyrulesandmembershipfunctions,systemdesign,completewithexperienceforthisinthepresent;3simplefuzzyinformationprocessingwillreducecontrolprecisionofthesystemandthedynamicqualitybecomespoor.Ifyouwanttoimprovetheaccuracyofinevitabletoincreasethequantitativeseries,leadingtorulesearchscope,reducethedecision-makingspeed,can'tevenreal-timecontrol;4.Howtoensurethestabilityofthefuzzycontrolsystemishowtosolvethefuzzycontrolonthestabilityandrobustnessproblemsremainstobeperfect.29 第二章模糊控制器的设计2.1模糊控制系统的原理模糊控制作为以模糊集合论、模糊语言变量及模糊逻辑推理为基础的一种计算机数字控制,它已成为目前实现智能控制的一种重要而又有效的形式尤其是模糊控制和神经网络、遗传算法及混沌理论等新学科的融合,正在显示出其巨大的应用潜力。控制规律执行机构被控对象传感器给定量图1常见负反馈控制系统方框图由测量装置、控制器、被控对象及执行机构组成的自动控制系统,就是人们所悉知的常规负反馈控制系统。其结构如图1所示。然而经过人们长期研究和实践形成的经典控制理论,虽然对于解决线性定常系统的控制问题非常有效。随着计算机尤其是微机的发展和应用,基于由于式中μ模糊量,所以为了对被控对象施加精确的控制,还需要将其清晰化转换为精确量u,然后经D/A得模拟量送给执行机构,对被对象进行第一步控制。然后中断等待第二次采样,进行第二步控制...这样循环下去就实现了对被控对象的模糊控制。2.2模糊控制器的基本结构模糊控制器的基本结构包括知识库、模糊推理、输入量模糊化、输出量精确化四部分。29 1.知识库知识库包括模糊控制器参数库和模糊控制规则库。模糊控制规则建立在语言变量的基础上。语言变量取值为“大”、“中”、“小”等这样的模糊子集,各模糊子集以隶属函数表明基本论域上的精确值属于该模糊子集的程度。因此,为建立模糊控制规则,需要将基本论域上的精确值依据隶属函数归并到各模糊子集中,从而用语言变量值(大、中、小等)代替精确值。这个过程代表了人在控制过程中对观察到的变量和控制量的模糊划分。由于各变量取值范围各异,故首先将各基本论域分别以不同的对应关系,映射到一个标准化论域上。通常,对应关系取为量化因子。为便于处理,将标准论域等分离散化,然后对论域进行模糊划分,定义模糊子集,如NB、PZ、PS等。同一个模糊控制规则库,对基本论域的模糊划分不同,控制效果也不同。具体来说,对应关系、标准论域、模糊子集数以及各模糊子集的隶属函数都对控制效果有很大影响。这3类参数与模糊控制规则具有同样的重要性,因此把它们归并为模糊控制器的参数库,与模糊控制规则库共同组成知识库。2.模糊化将精确的输入量转化为模糊量F有两种方法:(1)将精确量转换为标准论域上的模糊单点集。精确量x经对应关系G转换为标准论域x上的基本元素,则该元素的模糊单点集F为 uF(u)=1ifu=G(x)(2)将精确量转换为标准论域上的模糊子集。精确量经对应关系转换为标准论域上的基本元素,在该元素上具有最大隶属度的模糊子集,即为该精确量对应的模糊子集。3.模糊推理 最基本的模糊推理形式为: 前提1 IFATHENB 前提2 IFA′ 结论 THENB′ 其中,A、A′为论域U上的模糊子集,B、B′为论域V上的模糊子集。前提1称为模糊蕴涵关系,记为A→B。在实际应用中,一般先针对各条规则进行推理,然后将各个推理结果总合而得到最终推理结果。4.精确化29 推理得到的模糊子集要转换为精确值,以得到最终控制量输出y。目前常用两种精确化方法:(1)最大隶属度法。在推理得到的模糊子集中,选取隶属度最大的标准论域元素的平均值作为精确化结果。(2)重心法。将推理得到的模糊子集的隶属函数与横坐标所围面积的重心所对应的标准论域元素作为精确化结果。在得到推理结果精确值之后,还应按对应关系,得到最终控制量输出y。2.3模糊控制器的分类模糊控制的类型有:(1)基本模糊控制器:一旦模糊控制表确定之后,控制规则就固定不变了;(2)自适应模糊控制器:在运行中自动修改、完善和调整规则,使被控过程的控制效果不断提高,达到预期的效果;(3)智能模糊控制器:它把人、人工智能和神经网络三者联系起来,实现综合信息处理,使系统既具有灵活的推理机制、启发性知识与产生式规则表示,又具有多种层次、多种类型的控制规律选择。2.4模糊控制器的设计模糊控制器在模糊自动控制系统中具有举足轻重的作用,因此在模糊控制系统中,设计和调整模糊控制器的工作是很重要的。模糊控制器的设计包括以下几项内容:1、确定模糊控制器的输入变量和输出变量;2、设计模糊控制规则,并计算模糊控制规则所决定的模糊关系,建立模糊控制表;3、确立模糊化和非模糊化方法;4、合理选择模糊控制算法的采样时间。2.4.1模糊控制器的输入输出变量29 由于模糊控制器的控制规则是通过模拟人脑的思维决策方式提出的,所以在选择模糊控制器的输入输出变量时,必须深入研究人在手动控制过程中是如何获取和输出信息的。由于人在手动控制过程中,主要是根据误差、误差的变化及误差的变化的变化来实现控制的,所以模糊控制器的输入变量也可有三个,即误差、误差的变化及误差的变化的变化,输出变量一般选择控制量的变化。通常将模糊控制器输入变量的个数称为模糊控制的维数。由于一般情况下,一维模糊控制器的动态控制性能并不好,三维模糊控制器的控制规则过于复杂,控制算法的实现比较困难,所以,目前被广泛采用的均为二维模糊控制器,这种控制器以误差和误差的变化为输入变量,以控制量的变化为输出变量。整个论域即在定义这些模糊子集时应注意使论域中任何一点对这些模糊子集的隶属度的最大值不能太小,否则会在这样的点附近出现不灵敏区,以至于造成失控,使模糊控制系统控制性能变坏。2.4.2建立模糊控制器的控制规则建立模糊控制规则的基本思想:当误差大或较大时,选择控制量以尽快消除误差为主,而当误差较小时,选择控制量要注意防止超调,以系统的稳定性为主要出发点。模糊控制规则的来源有3条途径:基于专家经验和实际操作,基于模糊模型,基于模糊控制的自学习。模糊控制器的控制规则作为人工手动控制策略的语言描述,它通常用条件语句表示。其主要形式可概括如下:IfAthenBIfAthenBelseCIfAandBthenCIfAthenifBthenCIfAorBandCorDthenEIfAthenBandifAthenCIfAthenB,CIfAthenB1elseifA2thenB2知道上述条件语句之后。以二维模糊控制器为例,假设条件语句形式为ifE=AthenifC=BjthenU=Cij(i=1,2,...,n;j=1,2...,m),式中AiBjCij分别定义在误差、误差变化和控制量论域X,Y,Z上的模糊集;E,C,U分别代表误差、误差变化和控制模糊变量。2.4.3确立模糊化和精确化化方法一模糊化方法由于计算机采样输入的变量均为精确量,所以为便于实现模糊控制算法,须经过模糊量化处理变为模糊量。29 模糊化一般采用如下两种方法:1、将在某区间的精确量x模糊化成这样的一个模糊子集,它在点x处隶属度为1,除x点外其余各点的隶属均取0。如所选模糊集合论域为X={-n,-n+1,...,0,...,n-l,n},而输入的基本论域为[-e,e],输入精确量为e。2、首先同上算法得到L,其次查找语言变量赋值表,找出1位置上与最大隶属度所对应的语言值所决定的模糊量,该模糊量便为e的模糊化量。二精确化方法在模糊控制系统中,由于对建立的模糊控制规则通过模糊推理决策出的控制变量是一个模糊子集,它不能直接控制被控对象,所以还需要采取合理的方法将其转换为精确量,以便最好的发挥出模糊推理结果的决策效果。精确化过程的方法很多,主要有MIN-MAX重心法、代数积-加法-重心法、模糊加权型推理法、函数型推理法、加权函数型推理法、选择最大隶属度法、取中位数法。2.4.4采样时间的选择选择采样时间是计算机控制中的构性问题,所以模糊控制作为计算机控制的一种类型,也存在合理的选择采样时间的问题。香农采样定理给出了选择采样周期的下限.即式中为采样信号的上限角频率。在此范围内,采样周期越小,就接近连续控制。但也不能太小,它需要综合考虑执行机构响应时间、计算机控制算法所需时间、计算机字长、抗干扰性能等多方面因素Thesecondchapter,thedesignoffuzzycontroller2.1theprincipleoffuzzycontrolsystem29 Fuzzycontrolasfuzzysettheory,fuzzylanguagevariableandfuzzylogicreasoningonthebasisofacomputernumericalcontrol,ithasbecometherealizationofintelligentcontrolisanimportantandeffectiveespeciallyintheformoffuzzycontrolandneuralnetwork,geneticalgorithmandthefusionofnewdisciplinessuchaschaostheory,isshowingitsgreatpotentialapplications.Acommonnegativefeedbackcontrolsystemblockdiagraminfigure1Bymeasuringdevice,controllerandcontrolledobjectandactuatoroftheautomaticcontrolsystem,isthatpeopleknowknowtheregularfeedbackcontrolsystem.Itsstructureisshowninfigure1.Yetafteralong-termresearchandpracticeofclassicalcontroltheory,althoughforsolvingthecontrolproblemoflineartime-invariantsystemisveryeffective.Alongwiththecomputer,especiallythedevelopmentandapplicationofmicrocomputerbasedonthetypeofmufuzzyquantity,soinordertoexertprecisecontrolonthecontrolled,stillneedtoconverttheirmotivationtoaccuratequantitytou,andthentheD/Aanalogtoactuators,forthefirststepinthecontrolobject.Thenstop29 waitingforthesecondsample,carriesonthesecondstepcontrol...Thisloopisrealizedwithfuzzycontrolofthecontrolledobject.2.2thebasicstructureoffuzzycontrollerThebasicstructureoffuzzycontrollerincludesknowledgebase,fuzzyreasoning,fuzzinessofinput,output,high-precisionfourparts.1.TheknowledgebaseLibraryknowledgebaseincludingfuzzycontrollerparametersandfuzzycontrolrulebase.Onthebasisoffuzzycontrolrulesbasedonlinguisticvariable.Languagestatevariableisthe"big","","small29 ",suchasthefuzzysubset,thefuzzysubsettosubordinatefunctionshowsthatthebasictheoryofprecisionvaluebelongstothefuzzysubsetofthedomain.Therefore,inordertoestablishfuzzycontrolrules,requirestheaccuratevaluesonthebasictheoryofdomainbasedonmembershipfunctionareincorporatedintothefuzzysubset,tousethelanguagevariablevalues(large,mediumandsmall,etc.)insteadoftheaccuratevalues.Thisprocessrepresentsthepeopleofobservedvariablesinthecontrolprocessandcontroltheamountoffuzzypartition.Duetothedifferentvariablescope,sothefirstwillbethebasictheoryofdomainrespectivelyindifferentcorrespondingrelations,mappedtoastandardizedtheoryfield.Usually,thecorrespondingrelationshipbetweenoffforquantitativefactors.Foreaseofhandling,BiaoZhunLundomainsuchasbulkchemicalseparation,andthentofuzzypartitionofdiscourse,definethefuzzysubset,suchasNB,PZ,PS,etc.Thesamefuzzycontrolrulebase,fuzzypartitiontothefundamentaltheoryofdomainisdifferent,thecontroleffectisalsodifferent.Specifically,correspondence,BiaoZhunLundomain,thenumberoffuzzysubset,andthemembershipfunctionoffuzzysubsethasagreatinfluenceonthecontroleffect.Thesethreekindsofparametersandthefuzzycontrolruleshavethesameimportance,thereforetomergethemintofuzzycontrollerparameterdatabase,togetherwiththefuzzycontrolrulebaseoftheknowledgebase.2.TheblurConvertaccurateinputintofuzzyquantityFtherearetwoways:(1)convertsgaugedBiaoZhunLunfuzzysinglepointsetsonthedomain.GaugedbythecorrespondingrelationbetweenxxGintoBiaoZhunLundomainonthebasicelements,thentheelementsofthefuzzysinglepointsetFUF(u)=1ifu=G(x)(2)convertsgaugedBiaoZhunLundomainoffuzzysubsets.GaugedbythecorrespondingrelationshipintoBiaoZhunLundomainonthebasicelements,ontheelementhasthemaximummembershipdegree29 offuzzysubsets,namelyforthepreciseamountcorrespondingfuzzysubset.3.ThefuzzyinferenceThemostbasicformoffuzzyreasoningis:1IFATHENBPremise2IFA'ConclusionTHENB'Amongthem,A,A'forfuzzysubsetonthetheoryofdomainU,B,B'forthetheoryoffuzzysubsetVonthedomain.Premise1iscalledthefuzzyimplicationrelations,toAandB.Inpractice,thegeneralrulesoffirstinviewoftheindividualreasoning,thenthereasoningresultsumandeventuallyreasoningresultsareobtained.4.AccurateReasoningoffuzzysubsettoconvertaccuratevalue,togetthefinalcontroloutputy.Twoaccuratemethodscommonlyusedatpresent:(1)themaximummembershipdegreemethod.Inthereasoningoffuzzysubset,theselectionmembershipdegreeofthelargestBiaoZhunLundomainelementastheaverageoftheaccurateresults.(2)thegravitymethod.WillgetthemembershipfunctionoffuzzysubsetreasoningabscissaandtheareaaroundthecenterofgravityofthecorrespondingBiaoZhunLundomainelementsasaccurateresults.Afterthereasoningresultsprecisionvalue,stillshouldaccordingtocorrespondingrelation,getthefinalcontroloutputy.2.3theclassificationofthefuzzycontrollerThetypeoffuzzycontrolare:(1)thebasicfuzzycontroller:afterfuzzycontroltablewasdetermined,controlrulesarefixed;29 (2)theadaptivefuzzycontroller:intheoperationoftheautomaticchange,improveandadjusttherulesandmakethecontroleffectofincreasingthechargedprocessandachievethedesiredeffect;(3)theintelligentfuzzycontroller,itbringspeople,artificialintelligenceandneuralnetwork,andrealizetheintegratedinformationprocessing,makethesystembothwithflexibleinferencemechanism,theheuristicknowledgeandproductionrulesays,butalsohasmanylayers,choiceofmultipletypesofcontrollaw.窗体顶端2.4thedesignoffuzzycontrollerFuzzycontrollerplaysanimportantroleinfuzzyautomaticcontrolsystem,sothefuzzycontrolsystem,thedesignandadjustmentoffuzzycontrollerisveryimportant.Thedesignoffuzzycontrollerincludesthefollowingcontent:1,todeterminethefuzzycontrollerinputvariablesandoutputvariables;2,thedesignoffuzzycontrolrules,andcalculatethefuzzycontrolrulesdeterminedbyfuzzyrelation,fuzzycontroltable;3,establishfuzzyandthefuzzymethod;4,therationalselectionofsamplingtimefuzzycontrolalgorithm.2.4.1fuzzycontrollerinputandoutputvariablesBecausethecontrolrulesoffuzzycontrollerisbysimulatingthehumanbrainthinkingdecision-makingmodeisputforward,sowhenchoosinginputandoutputvariablesofthefuzzycontroller,mustfurtherstudyintheprocessofmanualcontrolishowtoobtainandoutputinformation.Becausepeopleintheprocessofmanualcontrol,mainlyaccordingtothechangeoftheerror,errorchangeanderrorchangestorealizethecontrol,sotheinputvariableoffuzzycontrolleralsocanhavethree,i.e.,thechangeoferror,errorchangeanderrorchanges,thequantityofoutputvariablesusuallyselectedcontrol.29 Willthenumberofinputvariableoffuzzycontrollerusuallyreferredtoasthedimensionsofthefuzzycontrol.Becauseingeneral,aone-dimensionaldynamiccontrolperformanceoffuzzycontrollerisnotgood,thethreedimensionalfuzzycontrollercontrolrulesaretoocomplex,therealizationofthecontrolalgorithmismoredifficult,sothecurrentwidelyusedarebasedontwo-dimensionalfuzzycontroller,thiscontrollerwitherroranderrorchangeasinputvariables,tocontrolthechangeofthequantityofoutputvariables.Theentireuniverseisindefiningthefuzzysubsetshouldpayattentiontomakeanyofthesefuzzysubsettheoryfieldofthemaximummembershipdegreenottoosmall,otherwisewillappearneartheinsensitivearea,sothatcauseoutofcontrol,thefuzzycontrolsystemcontrolperformancedeteriorate.2.4.2establishcontrolrulesoffuzzycontrollerSetupthebasicideaoffuzzycontrolrules:whentheerrorisbigorlarge,choosecontrolquantityinordertoeliminatetheerrorisassoonaspossibleLord,whentheerrorissmall,selectcontrolshouldpayattentiontopreventovershoot,systemstabilityasthemainstartingpoint.Thesourceofthefuzzycontrolruleshasthreeways:basedontheexpertexperienceandtheactualoperation,basedonthefuzzymodel,basedonthefuzzycontrolofselflearning.Controlrulesoffuzzycontrollerasamanualcontrolstrategyoflanguagedescription,itisusuallyexpressedinconditionalstatements.Itsmainformscanbesummarizedasfollows:IfAthenBIfAthenBelseCIfAandBthenCIfAthenifBthenCIfAorBandCorDthenEIfAthenBandifAthenCIfAthenB,C29 IfAthenB1elseifA2thenB2Afterknowtheconditionalstatement.Intwo-dimensionalfuzzycontroller,forexample,supposethatconditionalstatementsformasifE=AthenifC=BjthenU=Cij(I=1,2,...,n;j=1,2,...,m),typeofAiBjCijrespectivelydefinedinthetheoryoferror,errorchangeandcontroltheamountofX,Y,Zinfuzzysets;E,C,Urepresenttheerror,errorchangeandcontroloffuzzyvariables.2.4.3establishfuzzyandaccuratemethodAfuzzymethodThecomputersamplingtheinputvariablesarebasedonaccuratequantity,sothefuzzycontrolalgorithmforeaseofimplementation,mustbemadeviafuzzyquantizationprocessintofuzzyquantity.BlurgenerallyUSESthefollowingtwomethods:1,willbeinacertainrangeofgaugedxfuzzyintosuchafuzzysubset,itatpointxmembershipfor1,inadditiontothexeachpointintherestofthemembershipare0.DomainssuchastheselectedfuzzysettheorytoX={-n-n+1,...,0,...,n-l,n},andinputthebasictheoryofdomainfor[-e,e],accuratequantityinputise.2,firstdittoalgorithmtogettheL,secondlanguagevariableassignmenttablelookup,findout1positionwiththemaximummembershipdegreeisdeterminedbythelinguisticvaluesofthecorrespondingfuzzyquantity,thefuzzymeasureforetheblur.AccuratemethodInafuzzycontrolsystem,asaresultofthefuzzycontrolrulesbyfuzzyinferencedecisionoutofcontrolvariableisafuzzysubset,itcannotdirectlycontrolthecontrolledobject,soyoualsoneedtotakereasonablemethodtoconvertthemintoaccuratequantity,sothatthebestplaythedecisioneffectoftheresultoffuzzyreasoning.Accurateprocessisalotofmethods,includingMIN-MAXgravitymethod,algebraicproduct-addition-gravitymethod,theweighted29 fuzzyreasoningmethod,functionalreasoningmethodandweightedfunctiontypeinferencemethod,choosethemaximummembershipdegreemethod,themedianmethod.2.4.4thechoiceofsamplingtimeChoosethesamplingtimeisthestructureproblemincomputercontrol,sofuzzycontrolasatypeofcomputercontrol,thereisalsoareasonablechoiceofsamplingtime.Shannonsamplingtheoremisgiventhechoiceofsamplingperiodlimit.NamelyTypefortheupperlimitofsamplingsignalintheangularfrequency.Withinthisrange,thesmallerthesamplingperiod,isclosetoacontinuouscontrol.Butnottoosmall,itneedscomprehensiveconsiderationactuatorresponsetime,longtimeneededforcomputercontrolalgorithm,computerword,suchasanti-jammingperformancefactors第三章模糊控制应用研究现状及展望3.1模糊控制应用研究现状:模糊控制具有良好控制效果的关键是要有一个完善的控制规则。但由于模糊规则是人们对过程或对象模糊信息的归纳,对高阶、非线性、大时滞、时变参数以及随机干扰严重的复杂控制过程,人们的认识往往比较贫乏或难以总结完整的经验,这就使得单纯的模糊控制在某些情况下很粗糙,难以适应不同的运行状态,影响了控制效果。29 常规模糊控制的两个主要问题在于:改进稳态控制精度和提高智能水平与适应能力。在实际应用中,往往是将模糊控制或模糊推理的思想,与其它相对成熟的控制理论或方法结合起来,发挥各自的长处,从而获得理想的控制效果。由于模糊规则和语言很容易被人们广泛接受,加上模糊化技术在微处理器和计算机中能很方便的实现,所以这种结合展现出强大的生命力和良好的效果。对模糊控制的改进方法可大致的分为模糊复合控制,自适应和自学习模糊控制,以及模糊控制与智能化方法的结合等三个方面。1、模糊复合控制: Fuzzy-PID复合控制:即模糊PID控制,通常是当误差较大时采用模糊控制,而误差较小时采用PID控制,从而既保证动态响应效果,又能改善稳态控制精度;一种简便有效的做法是模糊控制器和I调节器共同合成控制作用。 模糊-线性复合控制:如模糊-前馈补偿控制等,实际利用了模糊控制是变增益PI控制器的特点,在实际系统的控制中取得了较好的效果。 史密斯-模糊控制器:针对系统的纯滞后特性设计,用模糊控制器替代PID可以解决常规史密斯-PID控制器对参数变化适应能力较弱的缺陷;此外模糊推理和模糊规则的运用有利于在一定程度上适应时延的变化,在更复杂的情况下对对象的纯滞后进行有效的补偿。 三维模糊控制器:一种是利用误差E,误差变化Ec和误差变化速率Ecc作为三维变量,可以解决传统二维模糊控制器的快速响应与稳定性要求之间的矛盾;另一种方法是利用E,Ec和误差的累积和ΣE,这相当于变增益的PID控制器,提高了模糊控制的稳态精度。 多变量模糊控制:一般采用结构分解和分层分级结构,利用多个简单的模糊控制器进行组合,并兼顾多规则集之间的相互关系。2、自适应和自学习模糊控制: 自校正模糊控制器:修改控制规则的自校正模糊控制器,从响应性能指标的评价出发,利用模糊集合平移或隶属函数参数的改变,来实现控制规则的部分或全面修正,也可通过修正规则表或隶属函数本身来进行调整;基于模糊模型的自校正模糊控制器,包括利用模糊集理论辨识系统模型的语言化方法,基于参考模糊集的系统模糊关系模型辨识方法,以及由I/O数据建立模糊规则模型,并以此作为自校正控制器设计的基础等。 参数自调整模糊控制:自调整比例因子的模糊控制,引入性能测量和比例因子调整的功能,在线改变模糊控制器的参数,较大的增强了对环境变化的适应能力;基于模糊推理的PID自整定控制,如参数自整定模糊PD控制,以及类似的PI及PID控制等。29 模型参考自适应模糊控制器:利用参考模型输出与控制作用下系统输出间的偏差来修正模糊控制器的输出,包括比例因子、解模糊策略、模糊控制规则等。 具有自学习功能的模糊控制:包括多种对外扰影响或重复任务的性能具有自学习功能的模糊控制方法,以及自寻优模糊控制器等,其关键在于学习和寻优算法的设计,尤其是提高其速度和效率。 自组织模糊控制器:将参考模型和自组织机制相结合的模糊模型参考学习控制,及自适应递阶模糊控制等更高级的自组织形式具有很大的发展潜力。3、模糊控制与其它智能控制方法的结合:尽管模糊控制在概念和理论上仍然存在着不少争议,但进入90年代以来,由于国际上许多著名学者的参与,以及大量工程应用上取得的成功,尤其是对无法用经典与现代控制理论建立精确数学模型的复杂系统特别显得成绩非凡,因而导致了更为广泛深入的研究,事实上模糊控制已作为智能控制的一个重要分支确定了下来。4、专家模糊控制:专家系统能够表达和利用控制复杂过程和对象所需的启发式知识,重视知识的多层次和分类的需要,弥补了模糊控制器结构过于简单、规则比较单一的缺陷,赋予了模糊控制更高的智能;二者的结合还能够拥有过程控制复杂的知识,并能够在更为复杂的情况下对这些知识加以有效利用。5、基于神经网络的模糊控制:神经网络实现局部或全部的模糊逻辑控制功能,前者如利用神经网络实现模糊控制规则或模糊推理,后者通常要求网络层数多于三层;自适应神经网络模糊控制,利用神经网络的学习功能作为模型辨识或直接用作控制器;基于模糊神经网络的隶属函数及推理规则的获取方法,具有模糊连接强度的模糊神经网等,均在控制中有所应用;模糊系统与遗传算法相结合的控制器设计方法则提供了更为新颖的思路。此外,模糊预测控制,模糊变结构方法,模糊系统建模及参数辨识,模糊模式识别等的研究,也都属于较为前沿的研究方向。3.2模糊控制研究方向展望29 模糊控制仍然是一个充满争议的领域。由于它的发展历史还不长,理论上的系统性和完善性,技术上的成熟性和规范性都还是不够的,有待人们的进一步提高。模糊系统理论还有一些重要的理论课题没有解决。其中两个重要的问题是:如何获得模糊规则及隶属函数,这在目前完全凭经验来进行;以及如何保证模糊系统的稳定性。大体说来,在模糊控制理论和应用方面应加强研究的主要课题为: 适合于解决工程上普遍问题的稳定性分析方法,稳定性评价理论体系;控制器的鲁棒性分析,系统的可控性和可观测性判定方法等。 模糊控制规则设计方法的研究,包括模糊集合隶属函数的设定方法,量化水平,采样周期的最优选择,规则的系数,最小实现以及规则和隶属函数参数自动生成等问题,进一步则要求我们给出模糊控制器的系统化设计方法。 模糊控制器参数最优调整理论的确定,以及修正推理规则的学习方式和算法等。 模糊动态模型的辨识方法。 模糊预测系统的设计方法和提高计算速度的方法。 神经网络与模糊控制相结合,有望发展一套新的智能控制理论。 模糊控制算法改进的研究:由于模糊逻辑的范畴很广,包含大量的概念和原则;然而这些概念和原则能真正的在模糊逻辑系统中得到应用的却为数不多。这方面的尝试有待深入。 最优模糊控制器设计的研究:依据恰当提出的性能指标,规范控制规则的设计依据,并在某种意义上达到最优。Thethirdchapterfuzzycontrolresearchsituationandprospectsofapplication3.1thefuzzycontrolapplicationresearchstatus:Fuzzycontrolhasbettercontroleffectofthekeyistohaveaperfectcontrolrules.Butduetothefuzzyrulesaresummarizedoforobjecttoprocessfuzzyinformation,thehighorder,nonlinear,largedelay,time-varyingparametersoftherandomdisturbanceandseverecomplexcontrolprocess,theunderstandingofthepeopletendtobepoorordifficulttosummarizetheexperiencesofcomplete,thismakesthesimplefuzzycontrol,insomecases,veryrough,difficulttoadaptto29 differentrunningstatus,hasaffectedthecontroleffect.Conventionalfuzzycontrolofthetwomainproblemis:toimprovethesteadystatecontrolprecisionandimprovethelevelofintelligenceandabilitytoadapt.Inpracticalapplication,oftenisthethoughtoffuzzycontrolandfuzzyreasoning,andotherrelativelymaturecontroltheoryormethod,playtotheirstrengths,togetidealcontroleffect.Duetothefuzzyrules,andlanguageisveryeasytobewidelyacceptedbypeople,combinedwiththefuzzytechnologyinthemicroprocessorandcomputerthatcanbeeasilyrealized,sothecombinationshowastrongvitalityandgoodeffect.Toimprovethemethodoffuzzycontrolcanberoughlydividedintofuzzycompoundcontrol,adaptiveandself-learningfuzzycontrol,andthecombinationoffuzzycontrolwithintelligentmethodthreeaspects.1,fuzzycompoundcontrol:Fuzzy-PIDcompoundcontrol,theFuzzyPIDcontrol,usuallywhenthelargererrorofFuzzycontrol,andthePIDcontrolerrorissmall,astoguaranteetheeffectofdynamicresponse,andcanimprovethesteady-stateaccuracy;AsimpleandeffectiveapproachisacommonsyntheticfuzzycontrollerandtheIregulatorcontrolfunction.-linearfuzzycompoundcontrol,suchasfuzzy-feedforwardcompensationcontrol,theactualuseofthefuzzycontrolisavariablegaincharacteristicsofPIcontrollerintheactualsystemcontrolachievedgoodeffect.Smith-fuzzycontroller:basedonthepurelagcharacteristicsofsystemdesign,replacePIDwithfuzzycontrollercansolvetheconventionalSmith-PIDcontrollerthedefectsofweakadaptabilitytoparameterchange;Moreoverfuzzyreasoningandtheuseoffuzzyrulesisbeneficialtoadapttothechangeoftimedelayinthecertaindegree,inmorecomplexsituationsofobjectofpurelagcompensationeffectively.Fuzzycontrollerinthreedimensions:oneisusingtheerrorEanderrorchangeofEcandtheerrorchangerateofEccasthree-dimensionalvariable,cansolvethetraditionaltwo-dimensionalfuzzycontroller29 ofthecontradictionbetweenthefastresponseandstabilityrequirements.AnothermethodistouseE,EcanderroraccumulationandΣE,thisisequivalenttoavariablegainPIDcontroller,improvethesteady-stateaccuracyoffuzzycontrol.Multivariablefuzzycontrol:generallyUSESthestructuredecompositionandhierarchicalclassificationstructure,usemoresimplefuzzycontrollerarecombined,andbalancetherelationshipbetweenmultiplerulesets.2,adaptiveandself-learningfuzzycontrol:Self-tuningfuzzycontroller:modifythecontrolrules,aself-tuningfuzzycontrollerfromtheresponseontheevaluationofperformance,theuseoffuzzysettranslationorthechangeofthemembershipfunctionparameter,torealizethecontrolrulesofpartialorfullcorrection,canalsobeusedbyfixedrulestablesortoadjustmembershipfunctionitself;Self-tuningfuzzycontrollerbasedonfuzzymodel,includingtheuseoffuzzysettheorytoidentifysystemmodellanguagemethod,basedonthereferencesystemfuzzyrelationshipmodelidentificationmethodoffuzzysetsandfuzzyrulesareestablished,andbytheI/Odatamodel,andasthebasisofself-tuningcontrollerdesign,etc.Parameterself-tuningfuzzycontrol,self-tuningscalingfactorsoffuzzycontrol,theintroductionofthefunctionofperformancemeasurementandthescalingfactoradjustment,changetheparametersofthefuzzycontrolleronline,largerenhancestheabilitytoadapttotheenvironmentchanges;PIDself-tuningcontrolbasedonfuzzyreasoning,suchastheparameterself-tuningfuzzyPDcontrol,andthelikePIandPIDcontrol,etc.Modelreferenceadaptivefuzzycontroller:usingthereferencemodeloutputandthecontrolsystemundertheactionofthedeviationbetweentheoutputtocorrecttheoutputoffuzzycontroller,includingthescalefactor,fuzzy,fuzzycontrolrules,etc.Whichhasthefunctionofself-learningfuzzycontrol:includeavarietyofexternalinterferenceeffectortheperformanceofthe29 repetitivetaskwhichhasthefunctionofself-learningfuzzycontrolmethod,andtheoptimalfuzzycontroller,thekeyistolearnandthedesignoftheoptimumalgorithm,especiallytoimprovethespeedandefficiency.Self-organizingfuzzycontroller:thereferencemodelandtheorganizationmechanismofcombiningthefuzzymodelreferencelearningcontrol,andadaptivehierarchicalfuzzycontrolandothermoreadvancedformsofself-organizationhasagreatpotentialfordevelopment.3,thecombinationoffuzzycontrolwithotherintelligentcontrolmethods:Althoughthefuzzycontrolisstillexistmanydisputesintheconceptandtheory,butinthe90s,duetotheparticipationofmanyfamousscholars,internationalandsuccessonalargenumberofengineeringapplication,especiallyforcan'tuseclassicalandmoderncontroltheorytoestablishaccuratemathematicalmodelofcomplexsystemsisparticularlygradeisspecial,andthusledtoamorein-depthstudy,fuzzycontrolhasinfactasanimportantbranchofintelligentcontroltodeterminedown.4,theexpertfuzzycontrol:Expertsystemcanexpressandusingtheheuristicknowledgenecessarytocontrolthecomplexprocessandobject,attachimportancetotheneedsofmulti-levelandclassificationofknowledge,makeuptherulesoffuzzycontrollerstructureistoosimple,tocompareasingledefect,givesthehigherintelligentfuzzycontrol;Thecombinationofthetwoalsocanhaveacomplexprocesscontrolknowledge,andbeabletomorecomplexcasestomakeeffectiveuseoftheknowledge.5,fuzzycontrolbasedonneuralnetwork:Neuralnetworktorealizepartialorallofthefuzzylogiccontrolfunctions,suchasusingneuralnetworktorealizefuzzycontrolrulesandfuzzyreasoning,thelatterusuallyrequirenetworklayermorethanthreelayers;Adaptiveneuralnetwork,fuzzycontrol,neuralnetwork29 learningfunctionasmodelidentificationordirectlyusedasthecontroller;Membershipfunctionsandinferencerulesbasedonfuzzyneuralnetworkaccessmethod,afuzzyconnectionstrengthoffuzzyneuralnetwork,etc.,wereusedinthecontrol;Fuzzysystemwiththecombinationofgeneticalgorithmcontrollerdesignruleprovidesamorenovelideas.Inaddition,thefuzzypredictivecontrol,thefuzzyvariablestructuremethod,fuzzysystemmodelingandparameteridentification,fuzzypatternrecognition,etc,areallbelongtothefrontierresearchdirection.3.2fuzzycontrolresearchdirectioninfutureFuzzycontrolisstillacontroversialfield.Becauseofitsdevelopmenthistoryisnotlong,theoryofsystematicandintegrity,technicalmaturityandstandardizationarestillnotenough,tofurtherimprove.Thefuzzysystemtheoryandsomeimportanttheoryproblemnotsolve.Twooftheimportantquestionis:howtoobtainfuzzyrulesandmembershipfunctions,thisinthecurrentruleofthumbforcompletely;Andhowtoensurethestabilityofthefuzzysystem.Generallyspeaking,thefuzzycontroltheoryandapplicationaspectsshouldbestrengthenedforthemaintopicsoftheresearch:Suitableforsolvingthecommonproblemsintheengineeringstabilityanalysismethod,theoryofstabilityevaluationsystem;Controllerrobustnessanalysis,thesystemcontrollabilityandobservabilitydeterminationmethods,etc.Thestudyofthedesignmethodoffuzzycontrolrules,includingthesettingmethodofthefuzzysetmembershipfunctiontoquantifylevels,samplingperiod,theoptimalchoice,coefficientoftherulesofthesmallestimplementationaswellastherulesandmembershipfunctionparametersautomaticallygenerated,furtherdemandsourmethodinsystematicdesignoffuzzycontrollerisgiven.29 Fuzzycontrollerparametersdeterminationoftheoptimaladjustmenttheory,aswellasthecorrectlearningmethodandalgorithmofinferencerules,etc.Theidentificationmethodoffuzzydynamicmodel.Thedesignmethodofthefuzzypredictionsystemandthemethodofimprovethecomputingspeed.Combinedwithfuzzycontrol,neuralnetworkisexpectedtodevelopanewintelligentcontroltheory.Fuzzycontrolalgorithmtoimproveresearch:duetothecategoryoffuzzylogicisverywide,containalargeamountofconceptsandprinciples;However,theseconceptsandprinciplescanbetrulyappliedinthefuzzylogicsystemisoneofthefew.Trytobethoroughinthisregard.Theoptimalfuzzycontrollerdesignresearch:onthebasisofproperperformanceoftheproposed,specificationandthedesignprincipleofthecontrolrules,andachievetheoptimalinsomesense.参考文献1章乃荛,神经网络与模糊控制[M],清华大学出版社,19982于海生,计算机控制技术[M],机械工业出版社,2008:173-2063王耀南,神经网络自适应模糊控制在温度控制系统中制,19964胡寿松,自动控制原理[M].北京:科学出版社,2008:219-2585李士勇等,模糊控制和智能控制理论与应用[M],哈尔滨工业大学出版社,1990:250-2806龙升照,汪培庄,Fuzzy控制规则的自动调整问题[M],(3):105-1127诸静,模糊控制原理与应用及现状,电子技术杂志referenceChapter1isRao,neuralnetworkandfuzzycontrol[M],tsinghuauniversitypress,19982hai-shengyu,computercontroltechnology[M],mechanicalindustrypublishinghouse,2008:173-2063yao-nanwang,neuralnetworkadaptivefuzzycontrolsysteminthetemperaturecontrolsystem,19964shou-songhu,automaticcontrolprinciple[M].Beijing:sciencepress,2008:219-2585LiShiyong,fuzzycontrolandintelligentcontroltheoryandapplication[M],Harbinindustrialuniversitypress,1990:250-280Accordingtoloosen,we,theFuzzycontrolrulesofautomatic29 adjustmentproblems[M],(3):105-1127ZhuJing,fuzzycontrolprincipleandapplicationandthestatusquo,electronicsmagazine29
此文档下载收益归作者所有