英语词根_百度知道
知道的教一下！
在前cept=take 拿, leg, lig=choose, gather 选，收 elect 选举lev=raise 举、升 elevate 抬起，前缀改变单词词义，后缀决定单词词性，拿 different 不同的,圆 circle 圆，圈，环状物cid，长命fact, fac=do, make 做，作 factory 工厂fer=bring, carry 带，持 exhibit 展出，展览hospit，摄影的
photographer(n：ache=ache 痛 earache 耳痛ag=do, cours=run 跑 cruise巡航舰di=day 日 diary 日记dit=give 给 edit 编辑dict，钦佩mort=death 死 mortal 终有一死的mot=move 移动 motion 运动，动nomin=name 名 nominal 名义上的，有名无实的nov=new 新 novel 新的，新奇的numer=number 数 numeral 数字，[语]数词oper=work 工作 operation 手术，工作，操作ori=rise 升起 orient 东方，东方的paci=peace 和平 pacify 使和平，抚慰past=feed 喂，食 pasture 放牧、牧场，吃草pel=push, drive 推，逐，驱 propel 推动pend, pens=hang 悬挂 pendent 悬空的，悬而未决的pet=seek 追求 compete 竞争，比赛phon=sound 声音 phone 电话plen=full 满，全 plenty 大量，丰富pone=put 放置 postpone 推后，推迟popul=people 人民 population 人口，全体居民port=carry 拿，带，运 import 输入，进口pos=put 放置 expose 揭露，揭发preci=price 价值 precious 宝贵的，珍贵的pur=pure 清，纯 purify 使纯净rect=right, straight 正，直 correct 改正，纠正rupt=break 破 rupture 破裂，使裂开sal=salt 盐 salary 薪水sci=know 知 science 科学sec, sequ=follow 跟随 sequence 继续，连续sect=cut 切割 section 切开，一部分sent, sens=feel 感觉 sentiment 感情，思想感情son=sound 声音 sonic 声音的，音速的spect=look 看 spectate 出席，观看spir=breathe 呼吸 inspire 鼓舞，吸入tail=cut 切，割 tailor 裁缝，成衣商tain, ten=hold 握，持，守 contain 容纳，包含，内装tect=cover 掩盖 detect 侦查，发觉tempor=time 时 temporary 暂时的，临时的tend, tens=draw 拉 tension 拉紧，引力tent, tract=draw 拉，抽，引 tractor 拖拉机urb=city 城市 suburb 郊区，近郊ut=use 用 utility 效用, 有用vac, vacu=empty 空 vacancy 空白，空虚vad, vas=walk, go 行走 invasion 入侵，侵略vari=change 变化 variable 可变的，反复的ven=come 来 convene 召集（会议），集会vert, vers=turn 转 subvert 推翻，颠覆vi, via=way 路 via 取道，经由vis, vid=see 看 visible 可见的，看得见的vit=life 生命 vital 充满活力的viv=live 活 vivid 活泼的，有生气的前缀：以结合形式出现,与一词、词根或短语开头的一个音或连续几个音相接,或书写中一个字母或连续几个字母用以产生出派生或变化形式。汉语里指在词根前面的构词成分。如“阿哥”、“阿姨”中的“阿”英文中：（有详细的点击链接）一个英语单词可以分为三个部分：前缀（prefix），词根（stem）及后缀（suffix）。单词中位于词根前面的部分就是前缀。前缀，可以改变单词的意思。常见的前缀有如下几类：第一类：表示正负（或增减）的，如：un-in-im-il-ir-non-mis-mal-dis-anti-de-under-re-over-等；第二类表示尺寸的，如：semi-equi-mini-micro-macro-mega-等；第三类表示位置关系，如：inter-super-trans-ex-extra-sub-infra-peri-等；第四类表示时间和次序，如：ante-pre-prime-post-retro-等；第五类表示数字，如：semi-mono-bi-tri-quad-penta-hex-sept(em)-oct-dec-multi-等；其它类别，如：pro-auto-co-con-等。后缀是一种重要的构词法，通过后缀我们常常可以判断出一个词的词性。下面分四大类分别讲解一些常见后缀及其含义（和前缀不同的是，这是一些比较常见的与词性相关的:—tion、—tive等）名词后缀：1．-ster，-eer，-er（or）意为：从事某种职业或参与某种活动的人（personengagedinanoccupationoractivity）例词：gamester，gangster，songster，engineer，profiteer，mountaineer，auctioneer，driver，teacher，director，actor，professor2．-let意为：小或者不重要的东西（small，unimportantthings）例词：booklet小册子，leaflet小叶、嫩叶，starlet3．-ette意为：1）小的东西（small）例词：cigarette2）假的东西（imitation）例词：leatherette3）女性（female）例词：usherette4．-ess意为：女性（female）例词：actress，poetess，hostess，paintress5．-hood意为：时期（status；etc．）例词：boyhood，childhood，manhood6．-ship意为：才能，状态，资格，品质等（skill，state，condition，status，quality）例词：leadership，friendship，membership，lectureship，sportsmanship7．-ful意为：量（theamountwhichnouncontains）例词：cupful，handful，mouthful，spoonful8．-tion，-ion意为：1）状态，行动等（state；action；etc．）例词：action，oppression，possession，education，starva-tion2）机构等（institution；etc．）例词：organization，foundation9．-ment意为：状态，行动等（state；action；etc．）例词：movement，enslavement，pavement10．-al意为：动作（action）例词：arrival，refusal，revival，recital，removal11．-age意为：程度，数量等（extent；amount；etc．）例词：wastage，coverage，acreage，shrinkage，breakage，hostage12．-ness；-ity（ty）意为：状态，品质（state；quality；etc．）例词：happiness，usefulness，selfishness，kindness，rapidity，activity，sanity，changeability13．-ism意为：道义，主义，学说等（doctrineof，practiceof）例词：idealism，impressionism，absenteeism，racism动词后缀：1．-ify意为：转为，变为（toturninto，tomakeorbecome）例词：beautify，diversify，simplify2．-ize；-en意为：使……，变得……（tomakeorbecome；tomakeinto）例词：modernize，popularize，legalize，hospitalize，symbolize，ripen，widen，heighten，threaten3．-ate意为：增加，使……（giveoradd，makeorbecome）例词：originate，hydrogenate，validate，differentiate形容词后缀：1．-ful意为：充满，有（fullof；hav-ing；giving；etc．）例词：useful，pitiful，hopeful，helpful，forgetful，thankful，fearful2．-less意为：没有，无（without；notgiving）例词：speechless，childless，harmless，hopeless，meaningless3．-ly意为：有……品质的（havingthequalitiesof）例词：beastly，manly，brotherly，friendly4．-like意为：像……的（like）例词：childlike，statesmanlike，tiger-like5．-y；-ish意为：像……一般的（somewhatlike）例词：meaty，sandy，silky，hairy，leafy，watery，foolish，girlish，blackish，thinnish6．-some意为：像……一样的；引起……的；有……品质的（like；causing；havingthequalityof）例词：troublesome，burdensome，wholesome，tiresome，bothersome7．-able（ible）意为：能……的；可以……的（abletobe；capable）例词：changeable，readable，drinkable，comfortable，expansible，convincible8．-ed意为：有……的（having，etc．）例词：wooded，pointed，moneyed，odd-shaped9．-al意为：有……属性的，……类型的（natureof，typicalof）例词：cultural，personal，regional，musical10．-ary（ory）意为：属于……的，与……相连的（belongingto；connectedwith）例词：revolutionary，imaginary，contradictory11．-ous意为：富含……的；有……品质的；像……的（fullof；havingthequalityof；like）例词：glorious，erroneous，malicious，gracious12．-ic（ical）意为：……类的；属于……的（typicalof；belongingto）例词：historic，historical，methodic，methodical，dramatic，heroic13．-ive意为：有……属性的；有某种倾向的（havingthenatureorqualityof；givenortendingto）例词：attractive，talkative，restrictive，defensive，preventive，constructive，sensitive副词后缀：(常为ly)1．-ly意为：以……方式（ina...manner；etc．）例词：happily，boldly，attentive-ly，strangely2．-ward（s）意为：表示方式或动作的方向（manneranddirectionofmovement）例词：onward（s），backward（s），earthward（s），homeward（s），eastward（s）3．-wise意为：1）按照……方式（inthemannerof）例词：crabwise，clockwise2）就……而言（asfaras...isconcerned）例词：weatherwise，educationwise注释：（如果看不懂n./adj.这样的东西，对照一下名称看）n.名词v.动词pron.代词interj.感叹词pl.复数adj.形容词adv.副词num.数词art.冠词prep.介词conj.连词,dic=say 言..感兴趣，be intersted in+ doing sth, duct=lead 引导 conduct 引导，指导，形容事，农艺ann,mindful 记忆，记住的 memory 记忆，记忆力milit=soldier 兵 military 军事的，军队的mini=small,less 小 minimum 最小数mir=wonder 惊奇，惊异 admire 赞赏./sth.)interesting(令人感到有趣的事物：interest(n.兴趣,luck,accident 机会，偶发happen 发生，巧遇hibit=hold 拿,续集(We have a collecting)=collection更多，节略ced, curs.photo(照片){photography(n，注意 security 安全cur，拍照gress=go, walk 行走 progress 进步hap=chance，这是最基本的。首先举个例子。
photograph(n.)照片举例n，系列的v.collect(收集){collectioin(n.)可和collection通用。
collecting(n, ceed=go 行走 precedent 先行的，cruc=cross 十字 crucify把……钉在十字桇上；折磨cur=care 关心，挂念，排斥，拒绝接纳cogn=know 知道 cognition 认知cord=heart 心 cordial 衷心的，诚心的corpor, cor=body 体 corporation 团体，画 photograph 照相, man=hand 手 manuscript 手稿mar=sea 海 marine 海上的。词根可以是任何词性的词, shut 关闭 exclude 排除.)收集，走, cis=cut, kill 切，杀 suicide 自杀claim, clam=cry, shout 喊叫 exclaim 呼喊，取 exception 例外，除外circ=ring 环，, clus=close.)摄影师
collective(adj.)续集的，农田 agriculture 农业, aster=star 星 astronomy 天文学audi，流畅fus=pour 灌，流，倾泻 refuse 拒绝, act 做，动 agent 代理人agri, agro=agriculture 田地，拒受geo=earth 地 geography 地理学gon=angle角 triangle三角形grad=step, go, grade 步, hosp=guest 客人 hospitable 好客的idio=particular, own，惊叫clar=clear 清楚的，明白的 clarify 澄清，相异的flor, enn=year 年 annual 每年的，年度的astro，级 gradual 逐步的gram=write 写 telegram 电报graph=write, writing 写，航海的medi=middle 中间 mediate 居中调解，调停memor=memory，说 dialogue 对话loqu=speak 言说 eloquent 有口才的，雄辩的lun=moon 月亮 lunar 月亮的，似月的manu，社团cred=believe, trust 相信，信任 credibility 可信，可靠, olour=flower 花 florid 如花的，华丽的flu=flow 流 fluency 流利, audit=hear 听 audible 听得见的bell=war 战争 rebellion 反叛，反抗bio, bi=life 生命，生物 biology 生物学brev=short 短 abbreviate 缩短，说 dictator 独裁者，口授者duc；使清楚clud./adj.)摄影业，有趣，景点)是一个词根，有趣的)]就是和词根保持母系关系，词更又称母词，字母 literate 识字的，有文化的loc=place 地方 local 当地的log=speak 言, private, proper 特殊的，个人的，专有的 idiom 惯用语，方言insul=island 岛 insular 岛的，偏狭的it=go 行走 exit 出口，退出ject=throw 投掷 projection 投掷，发射lect，使升高liber=liberty 自由 liberation 解放lingu=language 语言 linguist 语言专家liter=letter 文字，词根能造出非常多的词，好似数学里的树形图。由interest衍生出的词[interested(对.，经营ed=eat 吃 edible 可以吃的，食用的ev=age 年龄, cour，时代 longevity 长寿单词一般由三部分组成：词根、前缀和后缀。词根决定单词意思
采纳率：27%
英语水平想提高要自己去努力。英语学习提高选培训，好的教学方法还是很重要的。在生活中不断的去积累，生活中点点滴滴的提高。
告诉我邮箱，我发给你
为您推荐：
其他类似问题
您可能关注的内容
英语词根的相关知识
换一换
回答问题，赢新手礼包
个人、企业类
违法有害信息,请在下方选择后提交
色情、暴力
我们会通过消息、邮箱等方式尽快将举报结果通知您。Please enable JavaScript in your browser to get the full Trove experience.
The work with identifier
has been deleted. There may be a .12-1812-1812-1812-1812-1812-1812-1812-1812-1812-18最新范文01-0101-0101-0101-0101-0101-0101-0101-0101-0101-0101-0101-0101-0101-0101-01温馨提示！由于新浪微博认证机制调整，您的新浪微博帐号绑定已过期，请重新绑定！&&|&&
你能读我吗？
临江仙 天涯孤旅只一人 四季日夜连晨昏 秋风如霜雨似针 长夜独行久 庭院几许深 忍看邻家天伦暖 遥闻欢颜笑语声 生非命薄却如今 梦醒人不见 泪眼映孤灯
http://dbank.vmall.com/
LOFTER精选
网易考拉推荐
用微信&&“扫一扫”
将文章分享到朋友圈。
用易信&&“扫一扫”
将文章分享到朋友圈。
阅读(1744)|
用微信&&“扫一扫”
将文章分享到朋友圈。
用易信&&“扫一扫”
将文章分享到朋友圈。
历史上的今天
在LOFTER的更多文章
loftPermalink:'',
id:'fks_087',
blogTitle:'外文期刊全称和缩写对照(2)',
blogAbstract:'
外文期刊全称和缩写对照
Journal Titles and Abbreviations
{if x.moveFrom=='wap'}
{elseif x.moveFrom=='iphone'}
{elseif x.moveFrom=='android'}
{elseif x.moveFrom=='mobile'}
${a.selfIntro|escape}{if great260}${suplement}{/if}
{list a as x}
推荐过这篇日志的人：
{list a as x}
{if !!b&&b.length>0}
他们还推荐了：
{list b as y}
转载记录：
{list d as x}
{list a as x}
{list a as x}
{list a as x}
{list a as x}
{if x_index>4}{break}{/if}
${fn2(x.publishTime,'yyyy-MM-dd HH:mm:ss')}
{list a as x}
{if !!(blogDetail.preBlogPermalink)}
{if !!(blogDetail.nextBlogPermalink)}
{list a as x}
{if defined('newslist')&&newslist.length>0}
{list newslist as x}
{if x_index>7}{break}{/if}
{list a as x}
{var first_option =}
{list x.voteDetailList as voteToOption}
{if voteToOption==1}
{if first_option==false},{/if}&&“${b[voteToOption_index]}”&&
{if (x.role!="-1") },“我是${c[x.role]}”&&{/if}
&&&&&&&&${fn1(x.voteTime)}
{if x.userName==''}{/if}
网易公司版权所有&&
{list x.l as y}
{if defined('wl')}
{list wl as x}{/list}A knowledge based system using fuzzy inference for supervisory control of bioprocesses_学霸学习网
A knowledge based system using fuzzy inference for supervisory control of bioprocesses
journal of blotechnologyELSEVIERJournal of Biotechnology 34 (8A knowledge based system using fuzzy inference for supervisory control of bioprocessesCamilla von Numers a,1, Mikio Nakajima a, Terhi Siimes b, Hajime Asama a, Pekka Linko b,., Isao Endo aaChemical Engineering Laboratory, The Institute of Physical and Chemical Research (RIKEN), Wako-shi, Saitama 351&01 Japan 9 Laboratory of Biotechnology and Food Engineering, Helsinki University of Technology, FIN-02150Espoo, Finland(Received 3 August 1993; accepted 12 January 1994)Abstract In this paper, a rule-based, real-time knowledge based system for bioprocess fault diagnosis and control is described. The system was designed to generate on-line advice for the operators and to supervise automatic control of bioprocesses, using biotechnical production of lactic acid as an example process. It consists of a real-time data acquisition and data processing system linked to a fuzzy expert system written in Smalltalk V / M a c . The expert knowledge was expressed in the form of a rule-based knowledge network, fuzzy membership functions and control strategies. The fuzzy expert system carries out on-line fault diagnosing on the basis of filtered specific rates calculated from process variable measurements, and provides suitable countermeasures to recover the process. Fault diagnosis was realized both by backward and forward chaining procedures. The system was constructed to allow three different control strategies (given here in Smalltalk syntax), change of Setpoint, FuzzyAnswer for each discovered fault, employing the fuzzy m e a n defuzzification method, and linguistic Advice to the operator. The system was successfully tested on-line with a laboratory scale process.Key word~: F Kn B Lactic acid cultivation1. Introduction* Corresponding author. Abbreviations: A, B = NB = NM = NS = P = product concentration (g 1-I); PB= PM= PS = S = substrate concentration (g l-l); U = u u i = an element of U; X = biomass dry weight (g 1 1); ZE = v = specific substrate consumption rate (g s-l); /x= specific growth rate (s-l); /ZA(Ui)= membership function of fuzzy set A; ~& = specific product formation rate (g s-l); ~ = u = n = intersection. Text in italics refers to Smalltalk code. 1 Visiting scientist from HUT at RIKEN. Present address: ValioData, Inc., PB 229 SF00101 Helsinki, Finland.In the field of bioprocess control, there are many kinds of objectives such as maintaining certain environmental conditions to optimize the process. T h e r e are methods which can serve these various objectives, but their applications to bioprocesses may cause some problems owing to considerable non-linearities, time varying parameters, and a n u m b e r of disturbances in the process. Thus, in bioprocess control one of the major problems is real-time estimation of the system/$07.00 ? 1994 Elsevier Science B.V. All rights reserved SSDI 0 1 6 8 - 1 6 5 6 ( 9 4 ) 0 0 0 0 7 - Y110C. yon Numers et al. /Journal o f Biotechnology 34 (8state. This is often difficult owing to the lack of reliable sensors, system complexity, model uncertainties and parameter variations. Ideally, the state of the process should be determined by means of measurements, and the knowledge of the process behaviour expressed in terms of a model. Since bioprocesses often cannot be completely described by a single mathematical model, the experience gained by working with actual processes provides valuable information. However, human operators require time-consuming training. Further, reserve personnel is often needed. The operators vary in reliability, consistency and emotions when dealing with the problems involved. The use of human operators is often unsuitable under hazardous circumstances. In order to overcome all such problems in a bioprocess control, knowledge based systems have been introduced (Linko, 1988). Knowledge based systems are constructed for emulating the reasoning process of a human operator. The knowledge obtained from experienced operators can be expressed as a set of rules or other form of heuristics. We have previously described a knowledge based fuzzy expert system developed on the basis of a shell called BIOTALK (Aarts et al., 1990). In this expert system shell, the knowledge base employs 'If, then' rules in the form of a knowledge network. According to Stephanopoulos (1987), rule-based expert systems are favoured by chemical and biochemical engineers to process verbally formulated knowledge. By collecting the knowledge from experts and cultivation processes into a knowledge network a number of example applications were developed for bioprocess fault diagnosis and control (Pokkinen et al., 1992; Siimes et al., 1992a,b). The goal of the present work was to incorporate online control ability to the expert system shell and to automate some of the supervisory tasks currently performed by expert operators according to the concept described by Endo et al. (1989).employed. The on-line sampling unit was connected to an HPLC analyzer, and a sample of cell-free medium was injected every 20 min to an ion exchanger column in order to obtain actual values of product and substrate concentrations (Endo et al., 1985). The sterilizable on-line laser turbidity electrode was used to measure the optical density to represent cell mass concentration (Nagamune et al., 1985). The bioreactor was monitored by a measurement and control system BIOACS (Bio Advanced Control System), installed in a Fujitsu A-240~ workstation used as a process computer (Endo et al., 1989). The workstation performed real-time data acquisition by collecting data from the process controller and from the HPLC unit, and carried out real-time data processing by filtering the actual values of substrate (S), product (P) and biomass (X) concentrations, and by calculating the respective specific rates (v), (rr) and (/x) (Pokkinen et al., 1992). A conventional fourthorder delay filter with a time constant of approx. 26 min was used. The calculated specific rates along with the measured values of the process variables (temperature, pH, agitation rate, and substrate product and cell mass concentrations) were transferred to the expert system implemented in a Macintosh IIci computer. An inference engine, a database for standard variable time-courses, fuzzy sets for the process and state variables, and a knowledge network representing 'If, then' rules were incorporated into the Smalltalk/V Mac-based object oriented fuzzy expert system shell. The inference was based both on backward and forward chaining described in detail below. The knowledge based system performed online, real-time diagnosing and control. This was made possible by multiplex communication between the system and the process computer through RS232C communication line as illustrated in Fig 1.2. System descriptionA 30-1 jar bioreactor equipped with a sampling unit and an on-line laser turbidity electrode was3. Knowledge expressionThe system shell was constructed in the object oriented Smalltalk V / M a c programming environ-C. yon Numers et al. /Journal of Biotechnology 34 (8111Jar f e r m e n t o r manual ~ . . . _ ~ _ operation~, ,,~adviceOperatorgc°ntr°st--ges 1 IIB.50pH35.0 I1~Tempnew set pointsiiiraw and pro.ssed--IIIIIIIIIIIIIIIIIIIIIII/./77 e o u ~O.D. AgitIIIIraw data setsngg OOgggExpert systemProcess controllerProcess computer Fig. 1. A schematic diagram of the system.ment. In the system, expert knowledge was expressed in a form of a rule-based frame network named knowledge network (Aarts et al., 1990). The network was constructed in a modular form in order to facilitate easy modifications and additions. An example of adding a new sub network to an existing network is shown in Fig. 2. Each frame consisted of a name representing a fact in the bioprocess, and of several slots indicating either stationary or dynamic values of knowledge. In the present work the knowledge network was constructed of nodes that have 'If, then' correlation. A number of different nodes such as aboth in the measured variables and in the causal relations between the nodes. For the StartNodes fuzzy sets were static, but for the EndNodes fuzzy sets could be dynamically changed with time. The fuzzy regions were defined on the basis of the standard deviations derived from experimentsi-NEW KNOWLEDGEo.oStartNode, EndNode, OrNode, AndNode, ForwardOrNode, ForwardAndNode, ActionNode and ForwardActionNode were defined in Smalltalksyntax as separate classes for various purposes. The knowledge network components are schematically shown in Fig. 3. The StartNodes represent the original causes for malfunctions, EndNodes the observable faults in the bioprocess, and OrNode, AndNode, ForwardOrNode and ForwardAndNode the fuzzy operations in question for backward and forward chaining procedures, respectively. ActionNodes and ForwardActionNodes represent the control actions in the case of backward chaining and forward chaining procedures to recover the process. Fuzzy membership functions were integrated into the network and used to handle uncertaintiesNEW ARCtIORIGINAL KNOWLEDGE 06 0.60.5.It !.... ~!~ i]s t.00.9,),5L_0.7_1Fig. 2. Adding new knowledge to the network. TruthValues written below the nodes are calculated according to the original network and TruthValues on top of the nodes are calculated after adding new knowledge. T h e TruthValues of the shadowed nodes remain u n c h a n g e d after the addition of the new subnetwork.112C. yon Numers et al. /Journal of Biotechnology 34 (8I i.o~,,l,,msiz~L.../ I :'l=ge'(~)~v(~ 0 OrNode StartNode ForwardOrNode ActionNode ForwardActionNodeI ire_culture i/i,- fI&m°='°°=Fig. 3. Two example rule chains of the current knowledge network for lactic acid cultivation, representing the following rules. (a) Forward chaining: 7 --, 4: If '/x = ' n o r m a l & then 'process is going well'; 4 --, 1: If 'process is going well' then 'temperature = 'ZE&. (b) Backward chaining: 8--, 6: If '~, = ' h i g h & then 'inoculum is 'strong&; 6 ~ 2: If 'inoculum is 'strong& then 'inoculum size is 'large&; 6 --, 3: or 'preculture time is 'long&; 2 ---,5: If 'inoculum size is 'large& then 'temperature = 'NS&..05 --'t000I:~~t ,~+4a - - +20.005~)&,..\.-015\p=\\',, _//,-.~-20&.`0.carried out under standard conditions. The standard deviations of the state variables presented in the EndNodes were much larger in the beginning of the cultivation due to signal noise, and decreased gradually towards the end of the process. An example for a state variable is shown in Fig. 4, which clearly demonstrates how four different fuzzy intervals for each state variable were used depending on the process time involved.4. Functional structure of the knowledge network In order to realize the inferencing in the system, several slots and procedures were implemented to the nodes in the knowledge network. Slots could store either static or dynamic values, of which the latter were calculated by the procedures given. Since an object oriented programming environment was used, procedures are re-p --IX= high0 -4a -2o 0 +20& +4CAgFig. 4. Dynamically changing fuzzy membership functions for the change of state variable ~. o- = standard deviation of A/~.C. yon Numers et al. /Journal of Biotechnology 34 (8113ferred to here as messages and slots as instance variables. The procedures and slots employed are described in a greater detail below. Slots named InNodes and OutNodes can be defined for each node in the knowledge network. The set of nodes connected to the right side of the certain node are defined as InNodes of that node and the nodes connected to the left side are called OutNodes. Together they indicate static knowledge of a causal relationship between two facts in a bioprocess. InNodes include the reasons for a fact in a node, and OutNodes show the results following from the reasons. A TruthValue is a slot, which indicates the degree of possibility of the fact as the basis of a single node. The TruthValue varied within a range from 0 to 1. In order to tie the knowledge of a certain bioprocess together a slot named Model was defined. This slot represents the model of the process under investigation. It includes a procedure to access to the process and state variables of the bioprocess in question, to the respective fuzzy membership functions, and to realize dynamic calculation of the TruthValue of the facts. Each node in the network is given a preliminary TruthValue by an initializing procedure performed by the node itself as follows: (a) OrNode, ForwardOrNode: The node initializes its TruthValue automatically to 0. b) AndNode, ForwardAndNode: The node initializes its TruthValue to 1. (c) StartNode, EndNode: The node calculates the TruthValue of its fact using a fuzzy membership function defined for the variable in question according to the slot Model. Subsequently, the TruthValue of a node is recalculated from the actual Truth Values of the InNodes through backward chaining or from its OutNodes in forward chaining. The calculation depends on the node type in the following way ( Truth Values /A,A, ~/'B denote different values according to the nodes in question in each case): (a) OrNode, ForwardOrNode: The TruthValue is set according to the maximum operation (FuzzyOr); a maximum of the Truth Values of the InNodes (OutNodes in case of ForwardOrNodes, respectively), (I.?A I.) /d,B = max[/.LA ' ~B]). (b) AndNode, ForwardAndNode: A minimumvalue (Fuzzy-And) of the Truth Values of its InNodes (OutNodes in case of ForwardAndNodes, respectively) is taken as a new Truth Value, (/z A n P~B = min[~A, /x~]). The TruthValue calculations described above are independent of possible later network modifications or additions. Consequently, possible newly installed parts of the knowledge network can be considered separately in this respect. Fig. 2 illustrates how a newly installed part only affects the nodes B and C of the original network, which have a causal relationship with the new part of the network. The Truth Values of the rest of the nodes, marked with shadow in the original network do not change.5. Backward and forward chaining procedures in fault diagnosisIn a typical deductive system, such as fault diagnosis, inference is done at query time using backward chaining. It is a method of reasoning by which goals are proven to be true by recursively proving that the sub-goals are true. In the present work, original causes of possible faults were selected as the goals in the fault diagnosis. Both backward and forward chaining procedures were needed in order to diagnose two diffe those which can be realized by a simple measurement and those which are more abstract and cannot be measured. In the backward chaining procedure, each node in the network, starting from the EndNodes, sends a message backward to each of its InNodes and as a result the TruthValue from each of them is returned. After receiving all of the Truth Values from the appropriate InNodes, the node in turn will recalculate its own TruthValue and send it further to its OutNodes or, in case there are no more OutNodes, to the inference engine. In the present work the forward chaining method was implemented to the fuzzy expert system shell in order to diagnose those characteristic phenomena that are not directly observable, such as microbial contamination during the process. In the present fault diagnosis system, the forward chaining is complementary to backward114C yon Numers et al. /Journal of Biotechnology 34 ( sis of the measurement data. As a result each EndNode returns a TruthValue for its fact. If this TruthValue is higher than the pre-defined threshold, the EndNode is regarded as the observed fault or fact in the bioprocess. In order to find possible reasons for this malfunction the inference engine sends a backward chaining message to the EndNode which further sends it to the InNodes in the network. During this procedure the nodes that are defined for forward chaining (ForwardOrNodes, ForwardAndNodes) will reply by similarly carrying out forward inference using the forward chaining procedure. When backward chaining reaches the StartNodes, which stand for the original causes for malfunctions, their Truth Values are calculated by evaluating the respective fuzzy operations defined in their facts. In this example (Figs. 5 and 6), the OrNodes (~-= 'low', conditions faulty) and AndNode (inoculum is weak) are initialized to 0.0 and 1.0, respectively, on the way to the StartNodes. The procedure is performed chain by chain, calculat-chaining. The inference begins with known facts and proceeds forward seeking to generate new facts by matching rules contained in the knowledge base (Russo and Peskin, 1987). Each node in the network sends the forward chaining message to each of its Outnodes, starting from the leftmost ones, and as a result the TruthValue from each of them is returned. After receiving all of the TruthValues from its OutNodes, the node in turn will recalculate its own TruthValue and send it further to its InNodes or, in case there are no more InNodes, to the inference engine.6. Fault diagnosisAn illustrative example of fault diagnosing employing backward chaining is shown in Figs. 5 and 6. First, the system receives the measured or calculated values of the process variables. Then, the inference engine sends a message in a consecutive order to each one of the EndNodes to calculate their respective Truth Values on the ba-?I temperature LI ='sho.'Fig. 5. A part of the knowledgenetwork for lactic acid cultivation.C. yon Numers et aL /Journal of Biotechnology 34 (8115Step 1Step 2Step 3?Step 4?Step 5Step 6(9??IStep 7 Step 8 Step 9? ??(9?Step 10-11Step 12Step 13@ ?Fig. 6. An example of fault diagnosis in lactic acid cultivation.116C. yon Numers et aL /Journal of Biotechnology 34 (8ing the TruthValue for each node on the way back to the EndNode (Tr ='low'). The Truth Values of the StartNodes 1 to 4 (temperature = ' v e r y high', agitation--'high', culture t i m e = ' l o n g ' , preculture time ='short') were 0.8, 0.7, 0.6 and 0.0, respectively. After termination of the backward chaining procedure, every node has its own Truth Value. The TruthValue of one node can be regarded as a TruthValue of a certain part of the network. Consequently, for example in the step 13 the OrNode 5 (conditions faulty) has a TruthValue of 0.8 obtained as the maximum of the Truth Values of the StartNodes 1 (temperature = ' v e r y high', TruthValue 0.8) and 2 (agitation ='high', TruthValue 0.7). After having set the TruthValue of every node by using the slots and procedures available, the inference engine starts to report all the chains in the network that have a TruthValue higher than a pre-defined threshold by displaying the faults andtheir reasons in a hierarchical order to the operator. As a result the appropriate EndNode (rr ='low') will be given a new TruthValue (0.8) calculated according to the Fuzzy-Or operation (the maximum of the Truth Values of the InNodes), representing the level of certainty at which the network could predict the original reason for the malfunction represented in the facts of the EndNodes at a given time. Fig. 7 gives an example display of the user interface at a given time during the diagnosing.7. C o n t r o l strategiesAfter the system has determined the cause of a process fault through inferencing as described above, it analyzes the appropriate control actions. All of the StartNodes which have a TruthValue greater than a pre-defined value are considered??File E d i t ~ D a t aInput Process LacticAcidReport Window Production :a=-%3i'IYYPHASEAT TIME: 05:40:00 15:'°°(6/exponential phase MYYPHASEATTIME: 06:00:00 IS: exponential phase i'IYYPHASEATTIrIE: 06:20:00 IS: transient phase4~ ptl820~5 ?C0I'IYYPHASEATTIHE: 06:40:00 IS: transient phaseJar, 26,
: 0 0 : 0 0 lll==n=mm~=~za~m==zam,~l ~0 ~i!i!iiiii~iiiiiiiiiii~iiiiiiiii~iii!i!i!~!iii!iii~iiiiiiiiiii!i~iiiiiii~i~i~i~iiiiiiiii!iii!iiiiiiiiiiiiiii~i~iii!iii~i~i~i~i~i~ 0&Faults detected Causes found from the AndNode:inoculum Js weak is 60~ from the StartNode:cultureti me = 'long' is 60~ from the StertNode:precultTime = 'short' is IOOR from the OrNode:conditions faultg is 2 ~ from the StartNode:pH = 'low' is 2~The fuzzg truth value of the observed fault: mggmeas='lo~v' is 100~I'~:;: The network found reasons for the observed fault: mggmeas='lo~,v' for 60% from the OrNode:conditions faultg is 2 5 from the 5tartNode:pH = 'low' is 2~ from the AndNodefinooulum is weak is 6 0 ~ from the StertNode:culturetime = 'long' is 60~ ii~i from the StartNode:precultTime = 'short' is 100~Fig. 7. User interface of the expert systemduring fault diagnosis of a lactic acid production process.c. yon Numers et al. / Journal of Biotechnology 34 (8 as possible reasons for the observed malfunction. With measurable faults each one of such nodes has its own control strategy represented in a form of a connected ActionNode, or similarly in the case of forward chaining with non-measurable faults a ForwardActionNode for sending a message on-line to the process computer for further actions. The control strategies were individually pre-determined for each one of such nodes on the basis of expert knowledge representing the key strategies to recover the process. The three basic types of control strategies used in the present work were, in Smalltalk syntax, SetPoint, Advice and FuzzyAnswer. In the SetPoint control strategy an exact value combined with the name of the process variable was sent to the process computer in order to adjust the process variable to its pre-determined set point. If the inference engine had reached the conclusion to recommend the control strategy called Advice, the process computer would draw the attention of the operator to this fact and give him advice for a manual operation. The FuzzyAnswer control strategy consisted of three parts, a process variable, the respective pre-defined fuzzy membership function, and the TruthValue of the discovered original cause. For each discovered fault a FuzzyAnswer is created. To obtain crisp values when using the control strategy. FuzzyAnswer, a defuzzifier based on the fuzzy mean (FM) method was employed according to Eq. 1 (Postlethwaite 1990):117original cause of the malfunction of /z being 'high' (see Fig. 3). Further, the fact '/z = 'normal& was connected to the ForwardOrNode 'process going well' with a TruthValue of 0.6. After finding the two facts two FuzzyAnswers were activated as countermeasures, (a) 'Change temperature to NS' ( Truth Value 0.4), and (b) 'Change temperature to ZE' ( Truth Value 0.6). The latter control strategy is applied to reduce radical changes in the set points when the process is actually going well according to the measured value of /z. The real control value (change of setpoint) for the temperature was calculated on the basis of the TruthValues of these two activated ForwardOrNodes. The expert system calculates the crisp values for the control variables separately using FM method for defuzzification. In the fuzzy expert system shell the membership functions for the defuzzification of any chosen variable can be easily set through the user interface. The crisp values obtained are sent back on-line to the process computer together with the appropriate control strategies for further actions. The system performed satisfactorily when tested on-line with a laboratory scale lactic acid production process.8. Conclusions It is difficult to avoid problems related to malfunctions in a production process. However, if faults are diagnosed and corrected before becoming too serious, various losses in the bioindustry can be significally reduced. For this reason an on-line, real-time fuzzy knowledge based diagnosing and control system was developed. A fuzzy expert system was installed to realize the diagnosing and control faculties, and a real-time data acquisition and processing system was used to carry out on-line actions. Fuzzy membership functions of each process and state variable, casual relationships to malfunctions, and defuzzification rules were determined on the basis of historical data and experience from the actual production processes. The system could satisfactorily perform on-line, real-time diagnosing and control.ffu.A(u) duAD= fsS_+p,A(U) du where A o is the crisp value of the fuzzy set A, /ZA(U) is the fuzzy membership function of A, and /ZA(U) = 0 for u & s_ or u & s+. To further illustrate the function of FuzzyAnswer an example case is described below. The EndNode messages of 'v = ' h i g h & and '/z = ' n o r m a l & were observed, and as a result of the fault diagnosis the StartNode 'inoculum size 'large& with a TruthValue of 0.4 was found as the (1)118C. yon Numers et al. /Journal of Biotechnology 34 (8AcknowledgementsThis study was s u p p o r t e d by special co-ordination f u n d s of the Science a n d T e c h n o l o g y A g e n c y of the J a p a n e s e G o v e r n m e n t . G r a n t s for two of us (CvN, TS) from the A c a d e m y of F i n l a n d are gratefully acknowledged.ReferencesAarts, R.J., Suviranta, A., Rauman-Aalto, P. and Linko, P. (1990) An expert system in enzyme production control. Food Biotechnol. 4, 301-315. Endo, I., Asama, H. and Nagamune, T. (1989) A database system and an expert system for realizing factory automation in bioindustries. In: Fiechter, A., Okada, H. and Tanner, R.D. (Eds.), Bioproducts and Bioprocesses. Springer-Verlag, Berlin, pp. 337-346. Endo, I., Nagamune, T. and Inoue, I. (1985) Autosampler. US Patent No. 4501161. Endo, I. and Nagamune, T. (1987) A database system for fermentation processes. Bioprocess Eng. 2, 111-114. Linko, P. (1988) Uncertainties, fuzzy reasoning and expertsystems in bioengineering. Ann. NY Acad. Sci. 542, 83101. Nagamune, T., Inoue, I. and Takematsu, N. (1985) Instrument for measuring concentration of substrate in suspension. US Patent No. 4561779. Pokkinen, M., Flores Bustamante, Z.R., Asama, H., Endo, I. and Linko, P. (1992) Diagnosing lactic acid fermentation based on specific rates of growth, substrate consumption and product formation. Bioprocess Eng. 7, 319-322. Postlethwaite, B. (1990) Basic theory and algorithms for fuzzy sets and logic. In: McGhee J., Grimble, M.J. and Mowforth, P. (Eds.), Knowledge-based Systems for Industrial Control. Peregrinus, London. pp, 34-46. Russo, M.F. and Peskin, R.L. (1987) Knowledge-based systems for the engineer. Chem. Eng. Progress 83, 38-43. Siimes, T., Nakajima, M., Yada, H., Asama, H., Nagamune, T., Linko, P. and Endo, I. (1992a) Knowledge based diagnosis of inoculum properties and sterilization time in lactic acid fermentation. Biotechnol. Tech. 6, 385-390. Siimes, T., Nakajima, M., Yada, H., Asama, H. Nagamune, T., Linko, P. and Endo, I. (1992b) Object-oriented fuzzy expert system for on-line diagnosing and control of bioprocesses. Appl. Microbiol. Biotechnol. 37, 756-761. Stephanopoulos, G. (1987) The future of expert systems in chemical engineering. Chem. Eng. Progr. 83, 44-51.
更多相关文档