夾具類外文翻譯-基于事例推理的夾具設(shè)計(jì)研究與應(yīng)用【中文4050字】【PDF+中文WORD】,中文4050字,PDF+中文WORD,夾具,外文,翻譯,基于,事例,推理,設(shè)計(jì),研究,應(yīng)用,中文,4050,PDF,WORD 【中文4050字】 基于事例推理的夾具設(shè)計(jì)研究與應(yīng)用 摘要:根據(jù)基于事例的設(shè)計(jì)方法,提出采用工序件的特征信息和夾具的結(jié)構(gòu)特征 信息來描述夾具的相似性,并建立了包括這2方面主要特征信息為基礎(chǔ)的事例索 引碼,設(shè)計(jì)了事例庫的結(jié)構(gòu)形式,創(chuàng)建了層次化的事例組織方式;同時,提出了基 于知識引導(dǎo)的夾具事例檢索算法,以及事例的修改和采用同族事例碼進(jìn)行相似 事例的存貯,形成了基于事例推理的夾具設(shè)計(jì).所開發(fā)的原型系統(tǒng)在型號工程夾 具設(shè)計(jì)等項(xiàng)目的設(shè)計(jì)過程中得到了應(yīng)用,并取得了令人滿意的使用效果. 關(guān)鍵詞: 基于事例的推理 夾具設(shè)計(jì) CAD 夾具是以確定工件安全定位準(zhǔn)確為目的的裝置，并在加工過程中保持工件 與刀具或機(jī)床的位置一致不變。因?yàn)閵A具的結(jié)構(gòu)依賴于產(chǎn)品的特點(diǎn)和在企業(yè)規(guī) 劃中加工工序的地位,所以它的設(shè)計(jì)是制造過程中的瓶頸,制約著效率的提高. 夾具設(shè)計(jì)是一個復(fù)雜的過程,需要有從大量的設(shè)計(jì)論文中了解質(zhì)量知識的經(jīng)驗(yàn)， 這些設(shè)計(jì)論文包括工件的結(jié)構(gòu)設(shè)計(jì)、涉及加工工藝，和加工環(huán)境。當(dāng)用這些擅 長繪制詳細(xì)設(shè)計(jì)圖的傳統(tǒng)的 CAD 工具（如 Unigraphics、CATIA、Pro/E)時,這 仍然是一項(xiàng)非常耗時的工作，但是利用以往的設(shè)計(jì)經(jīng)驗(yàn)和資源也不能提供一些 益處,而這正是提高效率的關(guān)鍵因素. 基于事例推理 (CBR) 的方法適應(yīng)以往個 案解決的辦法,建立一個新問題的方法,主要有以下四步驟:檢索、利用、修改,并 保留.這是一個比用專業(yè)系統(tǒng)模仿人類思維有用的使用方法，因?yàn)樘岢鲆粋€類似 的情況,和采用一些修改,似乎不言自明,而且比人類更直觀.所以支持不同事例的 設(shè)計(jì)工具已經(jīng)在諸多領(lǐng)域中發(fā)展起來，如在注射成型及設(shè)計(jì)、建筑設(shè)計(jì)、模具 設(shè)計(jì)投死, 規(guī)劃過程中,還有夾具設(shè)計(jì). 孫用六個數(shù)字組成代碼參數(shù),包括工件的 形狀、機(jī)械部分、軸襯,第一定位裝置,第二定位裝置和夾緊裝置. 但這個系統(tǒng)不 能用于除鉆床夾具外的其他夾具類型,不能解決儲存需要保留的同一參數(shù)代碼 的問題,這在CBR 中是非常重要的. 1 事例參數(shù)和事例圖書館的建立 1.1 事例參數(shù) 事例參數(shù)應(yīng)該由工件的所有的特征組成，來區(qū)別不同的夾具. 使用他們能 夠使操作方便. 因?yàn)榱慵男螤钍嵌喾N多樣的, 在生產(chǎn)企業(yè)中制造的技術(shù)要求 也不斷發(fā)展,許多特征作被用做事例參數(shù)將會使搜索速度降低,其主要特征是不重要的,因?yàn)榉峙浣o每個特征的比重必須減少. 另一方面,事例參數(shù)包含所有的 特征是困難的。 因此,考慮到實(shí)際和快速設(shè)計(jì)的需求,事例參數(shù)要包含工件的主要特征和夾 具的結(jié)構(gòu)。事例參數(shù)代碼由16 位數(shù)組成：13 位數(shù)是事例特征 3 位數(shù)是事例識 別數(shù)字。 前 13 位數(shù)代表 13 個特征。 每個數(shù)字與特征的一個屬性相一致,這可能是 "*"、"？"、"1"、"2"，…，"A"、"B",…，"Z"，…，等其中的一個。其中，"*" 是指任何一個,"?"代表不確定,"0"代表沒有。 系統(tǒng)規(guī)定:夾具的類型，工件的形狀，位置模式不能是"*"和"?"。在設(shè)計(jì)系 統(tǒng)時,三個項(xiàng)目的屬性信息沒有這些選擇,這就意味著必須選擇確定的屬性。 最后三位數(shù)是事例識別號碼,如果事例特征的13 位數(shù)是一樣的,這三個數(shù)字 就用來區(qū)別他們。 該系統(tǒng)還規(guī)定:"000"是用于修正的一個典型事例,其他事例"001"、"002"、…, 這些是用于設(shè)計(jì)師查找參考事例的. 如果其中一個偶爾需要改變成典型事例,首 先它必須要求改成"000",前面的自動變成參考事例. 事例索引碼的結(jié)構(gòu)如圖1 所示。 1—夾具類型； 6—工件重量； 11—夾緊模型； 2—工件形狀； 7—工件剛度； 12—夾具體； 3—工件材料； 8—加工內(nèi)容； 13—其他； 4—批 量； 9—過程所有物； 14 到16—事例識別碼； 5—工件比例； 10—定位模型； 圖1 事例索引碼的結(jié)構(gòu) 1.2 事例庫 事例庫由許多預(yù)定義的事例組成。事例的描述是基于事例推理的最重要的 問題之一。所以由索引碼復(fù)合。 1.3 事例的層次化 夾具的結(jié)構(gòu)相似被認(rèn)為是整個夾具，成分和內(nèi)容相似。所以，整個夾具事 例庫，成分事例庫，夾具的成分事例庫形成相同。整個夾具的設(shè)計(jì)資料通常是 由工件資料和工件加工資料組成,這就意味著夾具的設(shè)計(jì)應(yīng)滿足特別功能的需 求.全部夾具事例是由功能成分組成，它是用功能成分的名字和數(shù)字來進(jìn)行描述 3 的。成分事例代表成員（成分功能和其他結(jié)構(gòu)成分，主要驅(qū)動參數(shù)，數(shù)字，和 它們的約束關(guān)系）。成分事例（夾具的最低層）是功能成分和和其他成分的結(jié)構(gòu)。 在現(xiàn)代夾具設(shè)計(jì)中有很多參數(shù)化準(zhǔn)件和普通非標(biāo)準(zhǔn)件。所以成分事例圖書館應(yīng) 記錄特殊參數(shù)和保持它們的方法。 2 事例修改的策略 在基于事例的夾具設(shè)計(jì)中，最重要的是相似點(diǎn)的修改，這樣能有助于獲得 最相似的事例，以及縮短適應(yīng)時間。根據(jù)夾具設(shè)計(jì)的需求，事例修改的策略使 最接近的事例方法和知識指導(dǎo)結(jié)合起來。首先在深度上查找，然后在寬度上； 知識指導(dǎo)策略意味著在來自客觀事物根源的知識規(guī)則上查找，這就要首先查找 固定類型，然后查找工件的形狀，第三查找定位方法。例如，如果事例索引碼 包括夾具類型的磨削夾具，就只查找所有的磨削夾具，然后查找工件形狀的盒 子，第三查找一個平面兩個銷的定位方法。如果沒有合適的，就查找深度標(biāo)點(diǎn)， 然后回到最上層，然后再找所有與寬度相關(guān)的事例。 修改方法： 1） 根據(jù)夾具事例庫的事例索引信息，查找有關(guān)事例庫。 2） 將事例索引碼與事例庫的每個事例碼匹配，然后計(jì)算相似尺寸的價(jià) 值。 3） 整理相似尺寸的次序，最大的架子是最類似的事例。 兩個事例之間的相似點(diǎn)是基于兩個事例特征之間的相似點(diǎn)。相似點(diǎn)尺寸的 計(jì)算依靠特征的類型。相似點(diǎn)的價(jià)值可以通過數(shù)字化的價(jià)值來計(jì)算，例如比較 重量分別是 50kg 和 20kg 的工件。非數(shù)字化的價(jià)值也能計(jì)算，例如，現(xiàn)在前 13 位索引碼都是非數(shù)字化的價(jià)值。一個夾具的相似尺寸的計(jì)算公式如下： 其中S 表示通用夾具的相似尺寸，n 表示索引特性數(shù)，w i表示每個特性的重量，si ( f a i f b ji )表示事例庫中特性f a i和相關(guān)夾具的特性f b ji 的相似尺寸。同時，0≤s i ( f ai , f b ji )≤1 ，數(shù)值計(jì)算如下： 其中f a i表示第i 個特征的索引特性值，f b ji表示事例庫中第j 個事例的第i 個特 征的特性值。 所以有兩種方法選擇相似夾具。一個方法是建立數(shù)值。如果通用事例的相 似尺寸值比給定的數(shù)值小，這些事例就不能選來作相似事例。事例庫最初建立 的時候，只有一些事例，數(shù)值可以建小一點(diǎn)。如果有大量的相似事例，數(shù)值就 應(yīng)該建的大一些。另外一個方法是只建立相似事例的數(shù)字（例如 10），這是類 型單里相似尺寸的最大值。 3 事例的修改和存儲 3.1 事例的修改 夾具設(shè)計(jì)中相似事例的修改包括以下三個階段： 1） 成分的替代 2） 保持形式不變，調(diào)整成分的特性 3） 模型重新設(shè)計(jì) 如果夾具的成分是普通的物品，它們能通過使用工具被修改，代替以及刪 除，這些已經(jīng)被設(shè)計(jì)好了。 3.2 事例的存儲 在將一個新的事例保存到事例庫之前，設(shè)計(jì)者必須考慮保存是否有價(jià)值。如 果這個事例不能增加系統(tǒng)的知識，就沒有必要把它保存到事例庫里。如果它有 價(jià)值的話，設(shè)計(jì)者在保存之前必須分析一下，看看這個事例是否作為標(biāo)準(zhǔn)事例 或參考事例被存儲了。一個標(biāo)準(zhǔn)事例是一個描述同族事例主要特征的標(biāo)準(zhǔn)。一 個同族事例是有事例庫中索引碼前 13 位相同而最后三位不同的那些事例組成 的。一個標(biāo)準(zhǔn)事例的最后三位通常是“000”。一個參考事例屬于同族標(biāo)準(zhǔn)事例， 5 最后三位用不同數(shù)字區(qū)分。 從被解釋的概念中，可采用以下方法： 1）如果一個新的事例和任何一個存在的事例族一致，和一個存在的標(biāo)準(zhǔn) 事例的前13 位數(shù)相同，那么這個事例就不能存儲因?yàn)橐呀?jīng)這種標(biāo)準(zhǔn)事例了?；?者只能作為一個參考事例保存（最后三位不是“000”，而且和其它的不一樣） 在事例庫中。 2）如果一個新的事例和任何一個存在的事例族一致，并且被認(rèn)為代替這 個事例族要比以前的標(biāo)準(zhǔn)事例好，那么這個標(biāo)準(zhǔn)事例就被這個新的事例代替， 以前的標(biāo)準(zhǔn)事例作為一個參考事例保存。 3）如果一個新的事例和任何一個存在的事例族不一致，一個新的事例族 將會自動產(chǎn)生，并作為標(biāo)準(zhǔn)事例保存到事例庫中。 4 夾具設(shè)計(jì)中基于事例推理的過程 根據(jù)夾具設(shè)計(jì)的特性，夾具設(shè)計(jì)的基本信息，例如夾具的名字，零件，生 產(chǎn)和設(shè)計(jì)者等等，必須先輸入。然后，輸入或設(shè)計(jì)工件的模型。輸入有關(guān)工件 的細(xì)節(jié)信息，建立事例索引碼，然后CBR 開始依靠相似尺寸查找相似事例，選 出最相似的事例。如果需要的話，事例要滿足通用性設(shè)計(jì)，再存儲到事例庫中。 程序流程圖如圖2 所示 圖2 基于事例推理的夾具設(shè)計(jì)流程圖 6 5 基于事例推理的夾具設(shè)計(jì)說明 這是一個工件如圖3 所示。材料是45 鋼，底座，形狀為塊狀，生產(chǎn)批量 為中批等。需要設(shè)計(jì)成一個用來旋轉(zhuǎn)孔的旋轉(zhuǎn)夾具。 圖3 需要設(shè)計(jì)夾具的一個工件 （最大尺寸80mmx49mmx22mm） 工件的特征值，屬性值，事例索引碼和重量在表1 中列出。 表1 工件的事例索引碼和重量 特征名稱 特性值 索引碼 重量 夾具類型 車床夾具 1 100 工件形狀 塊狀 9 90 工件材料 中碳鋼 3 70 批量 中批 2 60 工件比例 小 5 60 工件重量 輕 5 60 工件剛度 硬度強(qiáng) 1 60 加工內(nèi)容 孔 3 80 程序要求 完成加工 3 70 定位方法 三個平面 1 100 夾緊方法 不確定 ？ 90 夾具體 復(fù)合 4 80 其他 沒有 0 60 通 過 查 找 和 計(jì) 算 相 似 點(diǎn) ， 最 相 似 的 事 例 的 事 例 索 引 碼 是 19325513321402000，細(xì)節(jié)信息在表2 中列出。 表2 最相似事例的事例索引碼 特征名稱 特性值 索引碼 夾具類型 車床夾具 1 工件形狀 塊狀 9 工件材料 中碳鋼 3 批量 中批 2 工件比例 小 5 工件重量 輕 5 工件剛度 硬度強(qiáng) 1 加工內(nèi)容 孔 3 程序要求 完成加工 3 定位方法 三個平面 1 夾緊方法 不確定 ？ 夾具體 復(fù)合 4 其他 沒有 0 相似點(diǎn)的計(jì)算如下： 所以夾具的相似尺寸值是0.806，這是在事例庫中用于設(shè)計(jì)的最相似的事 例，最相似的事例的結(jié)構(gòu)如圖4 所示 圖4 最相似的夾具 當(dāng)成分替代，修改定位模型和夾緊模型，以及調(diào)節(jié)相關(guān)尺寸之后，新的夾 具被設(shè)計(jì)出來，圖形如圖5 所示 圖5 需要設(shè)計(jì)的新夾具 因?yàn)樵谑吕龓熘袥]有相似夾具，新夾具被儲存到事例庫中。事例索引碼是 19325523311402000。 6 結(jié)論 基于事例推理，作為一個問題解決的方法，是一個比模仿人類思想的專業(yè)系 統(tǒng)更有效的方法，已經(jīng)在很多難獲取知識的領(lǐng)域里得到發(fā)展?；谑吕评淼?優(yōu)點(diǎn)如下：它和人類的思想很相似；一個事例庫通過保存新事例獲得自學(xué)能力， 它比有慣例庫更快更容易，它可以更好的傳遞和解釋新的知識，這和慣例庫有 9 很大的不同。基于事例推理中提出的一個夾具設(shè)計(jì)的框架已經(jīng)被實(shí)行了，使用 的是支持基礎(chǔ)數(shù)據(jù)的VC++，UG 電腦繪圖軟件。這個框架也已經(jīng)和普通成分庫 和典型夾具庫結(jié)合起來。這個發(fā)展的標(biāo)準(zhǔn)系統(tǒng)，用于航空項(xiàng)目，幫助夾具設(shè)計(jì) 者提高設(shè)計(jì)效率和重新使用先前的設(shè)計(jì)資源。? Journal of Southeast University ( English Edition) Vol. 20 No. 2 June 2004 ISSN 1003 ) 7985 Application and development of case-based reasoning in fixture design Wang Qi1, 2 Liao Wenhe1 ( 1College of M echanical and Electrical Engineering , N anjing U niversity of A er onautics and Astronautics, Nanjing 210016, China) ( 2 Department of M echanical Eng ineering, Y ancheng Institute of T echnology , Yancheng 224002, China) Abstract: Based on the case-based designing ( CBD) methodology, the fixture similarity is in two respects: the funct ion and the structure information. T hen, the computer-aided fixture design system is created on case- based reasoning ( CBR) , in which the att ribut es of the main features of workpiece and struct ure of fixture as case index code are designed for the ret rieve of the similar cases, and the structure and hierarchical relation of case library are set up for st ore. M eanw hile, the algorithm based on the knowledge- guided in the retrieve of the similar cases, the strategy of case adaptation and case storage in which the case identificat ion number is used to dist inguish from similar cases are presented. The applicat ion of the system in some projects improves the design efficiency and gets a good result. Key words: case-based reasoning; f ixture design; computer aided design ( CAD) Fix tures are dev ices that serve as the purpose of holding the w orkpiece securely and accurately, and maintaining a consistent relationship w ith respect to the tools w hile machining. Because the fixture structure depends on the feature of the product and the status of the process planning in the enterprise, it s design is the bottleneck during m anufacturing, w hich restrains to im prove the ef ficiency and lead- time. And fixture design is a com plicated process, based on ex perience that needs comprehensive qualitative know ledg e about a number of desig n issues including w orkpiece conf iguration, manufacturing processes involved, and machining environment. This is also a very t ime- consum ing w ork w hen using traditional CAD tools ( such as Unigraphics, CATIA or Pro/ E ) , which are good at perform ing detailed design tasks, but prov ide f ew benefits for taking advantage of the prev ious design ex perience and resources, w hich are precisely the key factors in improving the efficiency. T he m ethodology of case- based reasoning ( CBR) adapts the solution of a previously solved case to build a solution for a new problem w ith the f ollow ing four steps: retrieve, reuse, revise, and retain[ 1] . This is a more useful method than the use of an expert system to simulate hum an thought because proposing a similar case and applying a few modifications seems to be sel-f Received 2003-10-22. Foundation items: The Nat ional H igh Technology Research an d Development Program of C hina ( 863 Program ) ( No. 2002AA 411030) , the Natural S cientific Research Program College of University of Jiangsu Province ( No. 03KJD520246) . Biographies: Wang Qi ( 1962 ) ) , male, doctor, associate prof essor, w qi003 @ sohu. com; Liao W enhe ( corresponding author) , male, doctor, professor, cnw ho@ nuaa. edu. cn. ex planatory and more intuitive to humans. So various case- based design support tools have been developed for numerous areas[ 2- 4] , such as in injection molding and design, architectural design, die- casting die design, process planning, and also in f ix ture design. Sun used six digitals to compose the index code that included w orkpiece shape, machine portion, bushing, the 1st locating dev ice, the 2nd locating device and clamping device[ 5] . But the system cannot be used for other f ixture types except for drill fixtures, and cannot solve the problem of storage of the same index code that needs to be retained, w hich is very important in CBR[ 6] . 1 Construction of a Case Index and Case Library 1. 1 Case index The case index should be composed of all f eatures of the w orkpiece, which are distinguished f rom dif- ferent f ix tures. Using all of them would make the op- eration in convenient. Because the f orms of the parts are diverse, and the technology requirements of m an- ufacture in the enterprise also develop continuously, lots of features used as the case index w ill m ake the search rate slow , and the main feature unimportant, for the reason that the relative weig ht w hich is allot- ted to every feature must diminish. And on the other hand, it is hard to include all the features in the case index. Therefore, considering the practicality and the demand of rapid design, the case index includes both the major feature of the w orkpiece and the structure of fixture. The case index code is m ade up of 16 dig- Application and development of case- based reasoning in fixture design 175 its: 13 digits for case features and 3 dig it s for case -i dentification number. T he first 13 digits represent 13 features. Each dig it is corresponding to an at t ribute of the feature, w hich m ay be one of / * 0, / ?0, / 10, / 20, ,, / A0, / B0, ,, / Z0, ,, etc. In w hich, / * 0 means anyone, / ?0 uncertain, / 00 nothing. T he system rules: fixture type, w orkpiece shape, locating model cannot be / * 0 or / ?0. When the system is designed, the at tribute information of the three it ems does not have these options, w hich means the certain at t ribute must be selected. And the default of the other at tributes of the feature is w orkpiece material as / 30 ( m iddle carbon steel) , batch size as / 20 ( middle batch ) , w orkpiece scale / 30 ( medium ) , workpiece w eight / 30 ( m iddle ) , w orkpiece toughness / 10 ( much stiff er) , machining content / 9 0 ( relative ) , process property / 3 0 ( finished machining ) , clamping method / ?0 ( uncertain) , and others ( They are degree, profile modeling, etc. ) / 00 ( nothing ) . And the default v alues can be changed by the system administrator. For / * 0 and / ?0, w hen in retrial, / * 0 means matching anyone, / ?0 means matching none of them, and w hen saved, the certain means of / * 0 and / ?0 must be g iven. The last three digits are the case identif ication number, w hich means the 13 digits of the case feature are the same, and the number of these three digits is used for distinguishing them, and saving them in the same case family if needed. The system also rules: / 0000 is a prototype case, which is used for retrieval, and other cases are / 0010, / 0020, ,, which are used for reference cases to be searched by designers. If occasionally one of them needs to be changed as the prototype case, f irst it must be required to apply to change the one to / 0000, and the former is changed to referential case automatically. The construction of the case index code is show n in Fig. 1. 1. 2 Case library The case library consists of lots of predefined cases. Case representation is one of the most important issues in case- based reasoning . So compounding w ith the index code, the structure of case library is show n in Tab. 1. Fig. 1 Construction of case index code Tab. 1 Structure of case library Field name Data type Description Case ID Auto number Specify the case. s place in case l ibrary Case index Verchar Case index code: * * * * * * * * * * * * * - * * * Fixture type Verchar 1) Lathe f ixt ure; 2) Drilling jig; 3) M illing fixture; 4) Boring fixture; , Workpiece shape Verchar 1) Box; 2) Shaft; 3) Round; 4) Sheat h; 5) Seat ; 6) Piston rod; 7) Frame; 8) Plate-shel; 9) Block; A) Compound; , 5) M iddle hardness aluminum; , Batch size Verchar * ) Random; 1) Large batch; 2) M iddle batch; 3) Low batch; 4) Few Workpiece scale Verchar 1) M uch bigger; 2) Bigger; 3) M edium; 4) Less bigger; 5) Small Workpiece w eight Verchar 1) M uch heavier; 2) Heavier; 3) M iddle; 4) Less heavy; 5) Light Workpiece toughness Verchar ?) Uncertain ; 1) M uch stif fer; 2) Stiff er; 3) M iddle; 4) Les st iff er; 5) W eak M achining content Process property Verchar Verchar 1) Cylinder; 2) End face; 3) H ole; 4) Plane; 5) Step plane; 6) Slot; 7) Pocket; 8) Hole of dril-l ream; 9) Relative; A) Compounding 1) Blank; 2) Unfinishing machining; 3) Finish machining Workpiece material Verchar * ) Anyone; 1) Cast iron; 2) Low carbon steel; 3) M iddle carbon steel; 4) Alloy steel; Locating method Verchar 1) 3plane; 2) 1plane+ 2pin; 3) Big plane+ short pin; 4) Long V+ plane; 5) Double V+ plane; 6) Long pin+ little plane; 7) Small taper shaft; 8) centred ?) Uncertain ; 0) Nothing; 1) T op clamping; 2) T op+ side clamping; 3) S ide clamping; 4) Tilted Clamping method Verchar clamping; 5 ) M ult-i point top clamping; 6) Compounded clam ping; 7 ) Centred clamping; 8 ) Vacuum clamping; , Fixture body Verchar ?) Uncertain; 1) Plan e; 2) Cast ing; 3) W elding; 4) Compounding Others Verchar ?) Uncertain ; 0) Nothing; 1) Horizon-move-divided degree; 2) Horizon-round- divided degree; 3) Radia-l round-divided degree; 4) Profile modeling; , Case identification number Verchar Case identificat ion number: 000; 001; , Case mapping Verchar T he location of case modeling stored in hard disk of com puter 1. 3 H ierarchical f orm of Case T he structure similarity of the f ix ture is represented as the w hole fixture similarity, components sim ilarity and component similarity . So the w hole f ix ture case library, com ponents case library, component case library of f ix ture are f ormed corresponding ly. Usually design information of the w hole fix ture is composed of workpiece information and w orkpiece procedure information, w hich represent the fixture satisfying the specifically designing function demand. The w hole f ix ture case is made up of function components, w hich are described by the function components. names, numbers, and relationship to object- workpiece which are the body of the knowledge. The com ponents case represents the members. ( function component and other structure components) main- driven parameter, the number, and their constrain relations. The component case ( the low est layer of the f ixture) is the structure of function component and other components. In the modern f ix ture design there are lots of parametric standard parts and common non- standard parts. So the component case library should record the specif ication parameter and the w ay in w hich it keeps them. The hierarchical form of the f ix ture case library is show n in Fig. 2. cut dow n the t im e of adaptation. According to the requirement of fixture design, the strategy of case retrieval combines the w ay of the nearest neighbor and know ledge- guided. That is, first search on depth, then on breadth; the knowledge- guided strategy means to search on the know ledge rule f rom root to the object, w hich is f irst ly searched by the f ix ture type, then by the shape of the workpiece, thirdly by the locating method. For ex ample, if the case index code includes the m illing fixture of f ix ture type, the search is just for all milling fixtures, then for box of w orkpiece shape, the third f or 1plane+ 2pine of locating method. If there is no match of it, then the search stops on depth, and returns to the upper layer, and retrieves all the relative cases on breadth. Retrieval algorithms: 1 According to the case index information of f ix ture case library, search the relevant case library; o M atch the case index code w ith the code of each case of the case library, and calculate the value of the sim ilarity measure; ? Sort the order of similarity measure, the biggest value, w hich is the most analogical case. Sim ilarity betw een two cases is based on the similarity between the two cases . features. T he calculation of similarity measure depends on the type of the feature. The value of similarity can be calculated for num erical values, for example, compare w orkpiece with the w eight of 50 kg and 20 kg . T he value can also be calculated betw een non- numerical values, for example, now the first 13 digits index code is all non- numerical values. It is important that a CBR system supports a multitude of sim ilarity measure, and, ideally, allows the definition of new types. The sim ilarity measure of a f ix ture is calculated as follow s: n E ( w i si ( f ai , f bji ) ) i= 1 S = n E w i i= 1 i = 1, 2, ,, n Fig. 2 Hierarchical form of fix ture case library 2 Strategy of Case Retrieval In the case- based desig n of fixtures, the most important thing is the retrieval of the similarity, w hich can help to obtain the most sim ilar case, and to w here S is the similarity measure of current fixture, n is the number of the index feature, w i is the w eight of each f eature, si ( f ai , f bji ) is the sim ilarity measure of the at tribute f ai of the i- th feature w ith the attribute f bj i of relative feature of the j- th case in the case library . At the same t ime, 0 [ si ( f ai , f bji ) [ 1, the value counts as follows: If ( f ai and f bji are non- numerical values) If ( f ai = f bji ) si ( f ai , f bji ) = 1; Else si ( f ai , f bj i ) = 0; If ( f ai and f bj i are numerical values) f ai - f bj i so the case is not saved because it is represented well by the prototype case. Or is just saved as a reference case ( the last 3 digits are not / 0000, and not the same with si ( f ai , f bj i ) = 1- max( f ai , f bj i ) . others) in the case library. w here f ai is the value of the index at t ribute of the i- th feature, and f bj i is the value of at tribute of the relative i- th f eature of the j- th case in case library. So there are two methods to select the analogical fixture. One is to set the value. If the values of similarity measure of current cases were less than a given value, those cases would not be selected as analogical cases. When the case library is init ially set up, and there are only a few cases, the value can be set smaller. If there are lots of analogical cases, the value should get larger. The other is just to set the number of the analogical cases ( such as 10) , which is the largest value of similarity measure from the sorted order. 3 Case Adaptat ion and Case Storage 3. 1 Case adaptation The modification of the analog ical case in the fixture design includes the f ollowing three cases: 1) T he substitution of components and the component; 2) Adjusting the dimension of components and the component w hile the form rem ains; 3) T he redesig n of the model. If the components and component of the fixture are com mon objects, they can be edited, substituted and deleted with tools, w hich have been designed. 3. 2 Case storage Before sav ing a new fixture case in the case library, the designer must consider w hether the saving is valuable. If the case does not increase the know ledge of the system, it is not necessary to store it in the case library . If it is valuable, then the designer must analyze it before sav ing it to see w hether the case is stored as a prototy pe case or as reference case. A prototy pe case is a representation that can describe the main features of a case f amily. A case family consists of those cases w hose index codes have the same f irst 13 digits and different last three dig it s in the case library. The last three digits of a prototy pe case are alw ays / 0000. A reference case belongs to the same family as the prototype case and is distinguished by the different last three digits. From the concept that has been ex plained, the follow ing strategies are adopted: 1) If a new case mat ches any existing case family, it has the same first 13 digits as an existing prototype case, 2) If a new case matches any existing case family and is thought to be better at representing this case family than the previous prototype case, then the prototype case is substit uted by this new case, and the previous prototype case is saved as a reference case. 3) If a new case does not match any existing case f amily, a new case family w ill be generated automatically and the case is stored as the prototype case in the case library. 4 Process of CBR in Fixture Design According to the characteristics of f ix ture design, the basic information of the fixture design such as the name of f ix ture, part, product and the designer, etc. must be input first . T hen the f ix ture f ile is set up autom at ically, in which all components of the fixture are put together. Then the model of the w orkpiece is input or desig ned. The detailed information about the workpiece is input, the case index code is set up, and then the CBR begins to search the analogical cases, relying on the sim ilarity measure, and the most analog ical case is selected out. If needed, the case is adapted to satisfy the current design, and restored into the case library. T he f low chart of the process is show n in Fig. 3. Fig. 3 Flowchart of case- based reasoning for fixture deign 5 Illustrating for Fixture Design by CBR This is a workpiece ( see Fig. 4) . Its material is 45# steel. Its name is seat. Its shape is block, and the product batch size is middle, etc. A f ix ture is turning fixture that serves to turn the hole, w hich needs to be designed. The value of f eature, at tribute, case index code and w eight of the w orkpiece is show n in Tab. 2. The similarity is calculated as follow s: S= 1@ 100+ 1 @ 90+ 2 @ 1 @ 70+ 5 @ 1@ 60+ 2 @ 1@ 80+ 0 @ 100+ 0 @ 90 100+ 90+ 70+ 60+ 60+ 60+ 60+ 80+ 70+ 100+ 90+ 80+ 60 = 790= 0. 806 980 Fig. 4 A workpiece requested to design a fix ture ( M aximum size is 80 mm @ 49 mm @ 22 mm) Tab. 2 T he case index code and w eight of workpiece Feature name At tribut e value Index code We ig ht Fix ture ty pe La the fixt ure 1 100 Workpiece shape Block 9 90 Workpiece ma terial M iddle carbon steel 3 70 Bat ch size M iddle 2 60 Workpiece scale Small 5 60 Workpiece weight L ig ht 5 60 Workpiece toughness M uch stiffer 1 60 Machining content Hole 3 80 Pro cess propert y F inish machining 3 70 L ocating method 3plane 1 100 Clam ping method Uncert ain ? 90 Fixt ure body Compounding 4 80 Others No thing 0 60 T hrough searching, and calculating the similarity, the case index code of the most similar case is 1932551332140- 000, and the detailed information is shown in Tab. 3. Feature name Attribute value Index code Fixture type Lathe f ixture 1 W orkpiece shape Block 9 W orkpiece material M iddle carbon st eel 3 Batch size M iddle 2 Workpiece scale Small 5 W orkpiece w eight Light 5 W orkpiece toughn ess M uch st if fer 1 M achining content Hole 3 Process property Finish machining 3 Locating method 3plane 2 Clamping m ethod Uncertain 1 Fixture body Compounding 4 Others Nothing 0 Tab. 3 Case index code of t he most similar case So the value of similarity measure of the f ix ture w hich needs to be designed w ith the most analog ical case in case library is 0. 806, and the structure of the most analogical case is show n in Fig. 5. Fig. 5 M ost analogical fixture After having been substituted the com ponent, modified the locating model and clamp model, and adjusted the relative dimension, the new fix ture is designed, and the fig