National Kaohsiung First University of Science and Technology
FirstTech Institutional RepositoryNot a member yet
18584 research outputs found
Sort by
A Study of the relationship among Campus Innovative Climate, Entrepreneurial Characteristic and Entrepreneurial intention：An Analysis of Students at University of Science and Technology in Southern Regions of Taiwan
No full text[[abstract]]人塑造環境，環境塑造人，Lewin(1951)場地論的基本觀念，指人的一切行為與所處環境具有交互作用的關係，人所表現出來的所有行為，都是個人與環境兩方面因素交互作用後所得的結果，進而產生情緒、態度、信念、理想、目標、壓力與緊張等，因此在政府致力推廣三創教育(創新、創意、創業)下，希望藉由創業教育來培養學生的創業家特質及意圖，以及觀察學生個人參與創業行為的活躍度是否會影響其創業意圖之程度。因此，根據文獻整理分析，本研究旨在探討學校創新氣氛、創業家特質與創業意圖之間關係，並以Lewin(1951)的場地理論為基礎，研究採用問卷調查法，針對南部科技大學的在學學生為研究對象，實際有效問卷221份。在統計方法上，以t檢定及單因子變異數分析檢測學生在學校創新氣氛、創業家特質與創業意圖之差異分析；以皮爾森積差相關及多元迴歸分析檢測創學校創新氣氛、業家特質、創業意圖之關聯性分析的關聯性及預測力，研究結果發現：學校創新氣氛與創業意圖呈顯著正相關、創業家特質與創業意圖呈顯著正相關，且創業家特質，顯著正向影響創新氣氛與創業意圖之間關係。本研究根據研究目的與問題、文獻探討內容，經分析結果提出以下之研究建議：(1)三創教育的推廣與評鑑制度(2)將學術研究成果轉換成具有商業價值的創造(3)培育創業創客(4)邀請創業人士來校分享創業經驗並提供適時的實務學習(5)幫助大學生培育創業家特質[[abstract]]Peopleshapingtheenvironment,theenvironmentshapingpeople,Levin(1951)fieldtheorysay”Allthebehaviorofpeopleandtheirenvironmenthasaninteractiverelationship”,thebehaviorofpeopleistheresultoftheinteractionbetweentheindividualandtheenvironment,andthentheemotions,attitudes,beliefs,ideals,goals,stressandtension,sothegovernmentiscommittedtopromotingthreecreativeeducation(InnovationCreative,Entrepreneurial),hopingtocultivatestudents'entrepreneurialcharacteristicsandintentionsthroughentrepreneurialeducation,andwhetherthedegreeofactivityofstudents'participationinentrepreneurialbehaviorwillaffecttheirentrepreneurialintentions.Accordingtotheliterature,thegoalofthisstudyistodiscusstherelationshipbetweenschoolinnovationatmosphereandentrepreneurialintentions,andalsotodiscussmoderatingeffectsonthisrelationshipafteraccedingtoentrepreneurialcharacteristicsvariable.BasedonthetheoryoffieldtheoryofLewin(1951),thisquestionnaireissetupthroughthescholarproposedmeter.ThetargetsubjectsofthisstudyarestudentsatuniversityofscienceandtechnologyinsouthernregionsofTaiwan;thenumberofusefulquestionnairesare221copies.Statistically,thisstudyusedt-testandANOVAtoexaminethediversityofcampusinnovativeclimate,entrepreneur'scharacteristicsandentrepreneurialintentionfromdifferentstudents;Pearsoncorrelationandmultipleregressionanalysisareexaminedtheconnectionandanticipationofthecampusinnovativeclimate,entrepreneurialcharacteristicandentrepreneurialintention.Researchresultsare:Campusinnovativeclimateispositivelycorrelatedtoentrepreneurialintentionandalsotoentrepreneurialcharacteristic,andentrepreneurialcharacteristicshassignificantlymoderatingeffectontherelationbetweenCampusinnovativeclimateandentrepreneurialintention.Accordingtotheresultsofstatisticalanalysis,thespecificsuggestionsareprovidedforreference.Basedonthepurposeandproblemofthestudyandthecontentoftheliterature,thefollowingresearchsuggestionsareputforward:(1)Elevatethethreecreativeeducations(innovationCreative,entrepreneurial)anditsevaluationsystem(2)transformationofacademicresearchresultsintoacommercialvalueofthecreation(3)cultivateMaker(4)invitingentrepreneurstoshareentrepreneurialexperienceandprovidetimelypracticallearning(5)Tohelpstudentscultivatethecharacteristicsofentrepreneurs
Performance Analysis of Quantitative Investment Portfolios-Case Study for Taiwanese Stock Market
No full text[[abstract]]近年來「量化投資」已成為國際投資界興起的一種新方法：乃藉由蒐集大量的數據為依據，做為進行投資決策的基石。此法主要是利用電腦運算幫助人腦處理大量數據資訊，進一步協助歸納並總結出市場的規律，從中建立優化並可以重複使用的投資策略，導引出能夠提高成功概率的投資決策過程。本研究目的採用專業價值投資達人提倡的九大指標為選股策略，以台灣股票市場2000年至2016年的所有上市上櫃公司為研究樣本，資料來源為台灣新經濟日報。研究結果顯示年化報酬率為14.53%，長期績效的累積報酬率高達509%。[[abstract]]Quantitativeinvestmenthasbecomeapopularmethodamongglobalinvestmentenvironmentbysearchingalargeamountofinformationtoconstructinvestmentstrategieswithpotentialgoodperformance.Thismethodusesrule-basedalgorithmtohelphumandealwithagreatquantityofdata,todiscoverthemarketpatterns,tocreateandoptimizetheinvestmentperformance,andtoresultinaprocessofinvestmentdecisionwithhighwinprobabilities.TheaimsofthisstudyistoadoptstockselectionstrategiesextendedfromaprofessionalinvestorinTaiwanesemarket.Usingthedatarangingfrom2000to2016withallstockslistedinTaiwanstockmarketwiththedatasourcefromTEJ.Theperformanceofthequantitativeinvestmentstrategyisupto14.53%annuallywithcumulativereturnsupto509%
Investor's Market Timing Decisions on ETF：An Application of Anchoring Theory
No full text[[abstract]]本研究以2003年6月至2016年12月之間的台灣50ETF作為研究對象，並立基於定錨理論建構「52週低點買進的基金擇時策略」，以探討基金投資人本身的擇時能力。實證結果顯示：(1)52週低點買進的基金擇時策略能夠擊敗基金買進持有策，平均而言，52週低點買進的基金擇時策略高於基金買進持有策略約17%的年化報酬率；(2)當設定52週低點再下探3%進場的投資策略報酬率更為顯著；(3)在考慮交易成本的條件下，52週低點買進的基金擇時策略仍具有效性。本文最重要研究貢獻在於為提供一個創新的基金擇時策略；由於過去較少文獻探討基金投資人本身的擇時能力，因此本篇研究結果填補此文獻缺口。[[abstract]]UsingthedataonTWSE50ETFduringJune2003toDecember2016,thispaperconstructsa“anchoringtheory-based52-weeklow”mutualfundmarkettimingstrategyandtestsitseffectiveness.Ourresultsshowthat1)52-weeklowmarkettimingstrategybeatsthebuy-and-holdstrategy,withahighermutualfundperformanceat17%annually;2)Whenweset52-weeklowmarkettimingstrategywithadeclinedreturnby3%,the52-weeklowstrategyismorepowerful;and3)Our52-weeklowstrategystillworkswhenconsideringtransactionscosts.Ourstudycontributestotheliteraturebyprovidinganewmarkettimingstrategyformutualfundtrading
Announcement Effects of Private Equity Placements
No full text[[abstract]]本研究以2002年至2015年台灣上市(櫃)私募股權之公司為樣本，並分別探討私募股權之宣告效果以及在不同景氣、不同產業下之私募股權宣告效果。並以事件研究法進行實證。結論歸納如下：(1)私募宣告日前後五日之累積平均異常報酬為正；長期亦有正向累積平均異常報酬。(2)不論在多頭或是空頭市場下進行私募，皆有異常報酬率之產生(3)非電子產業之正向私募宣告效果大於電子產業。[[abstract]]Usingtheeventstudymethodology,thisstudyexaminesinvestors’reactionstofirmannouncementsofprivateplacement.Asampleof654privateplacementannouncementsduringJanuary2002andDecember2015inTaiwanwasobtained.Theempiricalfindingsareasfollows:1.Theshort-runeffectsofprivateequityplacement:Thecumulativeaverageabnormalreturnofprivateplacementannouncementandtheannouncementonthefirstdayafterannouncementarebothsignificantlypositive.Itshowsplacementcanbringpositivegainsforshareholders.Thelong-runeffectsofprivateequityplacement:TheCARissignificantpositive.2.Regardlesswhetherthemarketisbullishorbearish,abnormalreturnsispositive.3.Stockmarketreactiontotheannouncementsofprivateplacementdiffersbyindustry.Theaveragecumulativeabnormalreturnsforfirmsinelectronicindustryaresmallerthanthoseforfirmsinotherindustries
The Optimal Control Design of Intelligent Guided Vehicle for Cargo Application
No full text[[abstract]]本論文旨在應用直流馬達於自走車控制上。本研究以單晶片(ArduinoUNO)為控制自走車之開發，其核心架構包含車體結構、驅動控制以及伺服馬達設備，藉由最佳化方法找出理想的控制參數，將行程時間縮短為設計目的。實驗設計考量主要將自走車的行程時間最小化，設計主要著重於車子的輕量化，因此使用直流馬達為自走車的動力供應來源，並藉由ICL293D來控制馬達之正反轉；控制方式，運用對車體的設計，並思考控制馬達的參數由單晶片(ArduinoUNO)中的控制邏輯來對自走車的行為控制。本研究運用田口法對影響母群平均數與母群變異數相關因素進行最佳化分析，分析後的結果開發出最佳化設計。並以模糊理論對母群平均數與母群變異數等二品質特性求得可控因子的最佳水準組合來佐證田口法的設計，再由雙反應曲面法驗證多目標參數之正確性，使得實驗解果達到最佳化。關鍵字：直流馬達、田口法、Arduino、自走車、雙反應曲面法[[abstract]]ThepurposeofthispaperistoapplytheDCmotortotheself-propelledvehicle.Inthisstudy,thedevelopmentofself-propelledvehicleiscarriedoutwithArduinoUNO.Thecorearchitectureincludesbodystructure,drivecontrolandservomotorequipment.Theoptimalcontrolparametersareobtainedbyoptimizingthemethod,andthetraveltimeisshortenedtoaimofdesign.Thedesignofthedesignmainlytominimizethetraveltimeofthecar,thedesignismainlyfocusedontheweightofthecar,sotheuseofDCmotorpowersupplyfortheself-propelledsource,andbyICL293Dtocontrolthemotorforwardandreverse;,Theuseofthedesignofthebody,andconsidertheparametersofthecontrolmotorfromthesingle-chip(ArduinoUNO)inthecontrollogictocontrolthebehaviorofthecar.Inthisstudy,theTaguchimethodwasusedtooptimizethefactorsaffectingtheaveragenumberofmothersandthenumberofvariationoftheparentgroup,andtheresultswereanalyzed.AndthefuzzystandardisusedtosupportthedesignofTaguchimethod,andthenthecorrectnessofthemulti-objectiveparametersisverifiedbythedoublereactionsurfacemethod,andthecorrectnessofthemulti-objectiveparametersisverifiedbythefuzzytheory.Theexperimentalsolutionisoptimizeddesign.Keywords：DCMotors、AutomaticGuidedVehicle、Arduino、Taguchimethod、ResponseSurfaceMethodolog
Effect of Stamping Process on Iron Loss of Electrical Steel
No full text[[abstract]]本研究以冲切參數對馬達矽鋼片鐵損劣化的影響，釐清馬達矽鋼片鐵損劣化的原因，以建立高效率低鐵損劣化的矽鋼片成型技術。針對中鋼料號50CS470、50CS1300及35CS210矽鋼片，分析冲切間隙量(10~25μm)、冲切速度(250~510Spm)以及壓料力(635~3175Kgf)等冲切參數對矽鋼片鐵損值的影響，以及退火處理對於矽鋼片鐵損劣化改善。實驗結果顯示：(1)50CS470最低鐵損劣化百分比為10.1%，冲切參數為冲切間隙量3.5%，壓料力10%，冲切速度並無明顯影響。(2)50CS1300最低鐵損劣化百分比為5.2%，冲切參數為冲切間隙量2%，壓料力10%，冲切速度並無明顯影響。(3)35CS210最低鐵損劣化百分比為46%，冲切參數為冲切間隙量5%，壓料力12%，冲切速度並無明顯影響。(4)退火溫度在860℃時會因為殘留應力釋放使鐵損劣化獲得改善，鐵損劣化百分比會與較沒有殘留應力的電化學加工法相當。(5)相同板厚之50CS470與50CS1300在經冲切成型後，50CS470的晶粒尺寸為65~70μm，相較於50CS1300的晶粒尺寸30~40μm鐵損劣化百分比平均高出約11%。[[abstract]]Theaimofthisstudyistobuildhighefficiencyandlowironlossdeteriorationonsiliconsteel(ChinaSteelCorporation:50CS470、50CS1300&35CS210)withtechnology.Incomparisontoelectrochemicalmachining,stampingprocessinducesresidualstressinsidethematerialwhichdeterioratestheironlossproperty.Importantparameterstodecreasethedeterioratedironlosspropertyareanalyzed,includingcommonlyadjustedtrimmingtolerance(10~25μm),stampingspeed(250~510SPM)andblankholdforce(635~3175Kgf).Inadditionannealingtheprocessafterstampingprocess.Theresultsshow:(1)50CS470:thelowestpercentageofironlossdeteriorationis10.1%,while3.5%trimmingtoleranceand10%blankholdforceareapplied.(2)50CS1300:thelowestpercentageofironlossdeteriorationis5.2%,while2%trimmingtoleranceand10%blankholdforce.(3)35CS210:thelowestpercentageofironlossdeteriorationis46%,while5%trimmingtoleranceand12%blankholdforce.(4)Ironlosspropertyofsiliconsteelisaffectedbyresidualstressesandgrainsize,whenannealingtemperatureislowerthan860℃,theeffectofresidualstressdominatesandtheironlosspropertyimprovewiththeannealingtemperature.Withtheoptimumannealingconditionof860℃holdingtemperature,3hoursholdingtimeat5.5╳10-4Torrvacuumlevel,theironlossisveryclosetoelectrochemicalmachining.(5)Thegrainsizeincreasesfrom30~40μmto65~70μmafterstampingandtheironlossdeteriorationrateisfrom10%to21%
Improvement Method for Enterprise Software Utilization based on Usage Analysis and Dynamic Authorization
No full text[[abstract]]大型企業在電腦軟體利用上，除版權的合法性之外，常止於軟體的版權和數量上安裝的管控，難以分析所購買軟體適用性與足夠性，進而造成購置過多版權的浪費或版權購置不足的困擾。因此，如何查核企業內安裝軟體的合法性、利用率與實際使用情況，實為企業管理軟體的挑戰之一。本研究基於軟體的安裝數使用分析，提出一動態授權的機制，開發出一企業軟體利用分析系統。在架構上，以瀏覽器、web伺服器、與MongoDB建立MVC架構。在資料蒐集上，首先透過前端使用蒐集器，可依組織、軟體、時間、與使用行為蒐集使用紀錄，並將之彙整於DB中。在系統功能上，本系統可依軟體版權清單，分析軟體安裝率、使用率、與前景執行率等，提供軟體實際使用情況。以一大型企業為例，透過本系統可確實稽核該企業內部的軟體使用情形。以所分析的5套軟體而言，區分成三種授權，分別為安裝授權數、全廠安裝授權、與全廠執行授權等。如MS-Visio、MS-Project、與Acrobat為安裝授權數，而MS-office為全廠安裝授權，兩者重點都在非授權版本的減少與新授權版本的集中。而AutoCAD則為全廠執行授權，重點在執行階段的有效性。配合動態授權的機制，可觀察到非授權版本安裝率已逐月下降；而在全廠執行授權軟體的前景執行率和軟體使用率，亦分別提升約25%和50%。[[abstract]]Thesoftwaremanagementofanenterpriseshouldnotonlyauthorizethesoftwarecopyrightsbutalsoevaluateapplicabilityandsufficiencyoftheinstalledsoftware.Currently,itischallengingenterpriseshowtoefficientlyauditlegalityandeffectivelyevaluateutilizationoftheinstalledsoftware.Thispaperbasedsoftwareusagelogsproposesadynamicauthorizationsystemforevaluatingsoftwareutilization.Specifically,thesystemisstructuredwithaMVC(ModelViewController)architecturebywebbrowsers,awebserverandaMongoDB.Indatacollection,throughfrontendusagecollectors,thesystemisabletostorethesoftwareusagesbyorganization,software,time,andbehaviorintheMongoDB.Infunction,thesystemprovidesanalysisofsoftwareinstallationratio,usageratio,andforegroundexecutionratioaccordingtoasoftwarelist.Inalargeenterprisercase,thesystemenablesuserstoeffectivelyauditsoftwareutilization.Asforthefivesoftwarepackagesoftheenterprise,therearethreetypesoflicense:byindividuals,site,andexecution.Forexample,MS-Visio,MS-projectandAcrobatarecategorizedforlicensesbyindividualswhereasMS-officebeingsitelicense,inwhichthekeyofbothtypesistoreducetheun-licensedversionandtoincreasethenewlicensedversion.Belongingtotheexecutionlicense,AutoCADisneededtoevaluateitsexecutionefficiency.Inaddition,theresultstoapplyingtheproposedsystemwithdynamicauthorizationindicatethatinstallationratiooftheunlicensedversionwasreducedmonthbymonthandtheforegroundexecutionratioandutilizationratiooftheexecutionlicensesoftwarehaveincreased25%and50%,respectively
Study of Aligning Control for a Multi-Layer Strip Winding System
No full text[[abstract]]在高速多帶盤捲製程中捲型控制是一門重要的課題，為了避免各帶盤捲位置不一而造成過多的捲型溢出而裁切使得成本提升，在盤捲的過程中需準確的控制帶材的橫向位置，以確保獲得較佳的捲型，而本研究在於利用糾偏系統改善高速盤捲系統在盤捲速度變化下之捲型。研究中發現線速度會與糾偏系統反應息息相關，在盤捲過程中盤捲初始速度會慢慢提升至指定之最高速，而盤捲速度可能也會因為製程關係而隨時改變，但會因為改變時的線速度與原本的線速度不一而使用不匹配的控制器，其使得帶材進行糾偏時捲型不佳，故我們欲設計一種能夠隨帶材線速度變化而控制器參數隨著變化的適應性PI及PID調變控制器，首先我們先利用Matlab建立糾偏系統之模型，接著給予模型2m/s、4m/s、6m/s、8m/s、12m/s及16m/s的線速度命令並設計PI補償器及PID補償器，並在其中找尋線速度與PI/PID補償器數值之關係式，便可以藉由此關係式在不同帶材線速度下得到相對應控制器參數，此即為可隨線速度變化而調變之適應性PI及PID調變控制器。經由Matlab模擬結果顯示，使用PI控制器下，以第一個糾偏系統模型為例，在線速度1m/s至16m/s下若考慮中間輪之干擾，頻寬範圍為1.37rad/s至17.4rad/s，而不考慮中間輪之干擾，頻寬範圍為2.56rad/s至25.4rad/s，而在此線速度範圍下不考慮中間輪干擾之系統頻寬會比有考慮中間輪的系統平均高1.64倍。經由Matlab模擬結果顯示，使用適應性PI調變控制器下，以第一個糾偏系統模型為例，帶材從線速度0m/s等加速至2m/s的過程中，而糾偏命令給予1mm之步階位移命令，其響應之超越百分比為5%，而未使用適應性調變控制器的系統為65%，同一系統帶材從線速度0m/s等加速至16m/s的過程中，而糾偏命令給予1mm之步階位移命令其響應之超越百分比為5.5%，而未使用適應性調變控制器的系統則為62.5%；此外，在使用適應性PID調變控制器下，以第一個糾偏系統模型為例，帶材從線速度0m/s等加速至2m/s的過程中，而糾偏命令給予1mm之步階位移命令，其響應之超越百分比為11%，則未使用的系統為45%，同一系統帶材從線速度0m/s等加速至16m/s的過程中，而糾偏命令給予1mm之步階位移命令其響應之超越百分比為3.5%，而未使用適應性調變控制器的系統則為60%，由以上之數據可知適應性PI/PID調變控制器之系統對於線速度的高低在變化時都具有良好的超越百分比，其中適應性PI調變控制器對系統的調變效果較佳，但使用適應性PID調變控制器之系統反應頻寬較高。本研究係以PC-Based控制器之三帶盤捲實驗台為基礎，經實驗後可知第一個糾偏系統及第二個糾偏系統之等速盤捲標準差為0.24~0.89mm，適應調變PI控制器以2m/s、4m/s及8m/s之線速度下未使用適應調變PI控制器之系統標準差範圍為0.05~4.04mm，標準差平均值為1.42mm，而使用適應調變PI控制器之系統標準差範圍為0.04~2.24mm，標準差平均值為1.21mm，以實驗結果來說使用適應性PI控制器之糾偏系統響應有明顯的改善，其中對於線速度變化時對於較大偏移量較明顯的改善。[[abstract]]Aligningcontrolisimportantinhighspeedmulti-layerwindingprocess.Toachievehighwindingquilty,thelatetalpositioncontorlofthewindingstripintheprocessisoneofimportantissues.Thisresearchfocusesondevelopinganadaptivealigningcontrollertoimprovewindingqualityinhighspeedmulti-layerwindingsystem,inwhichthreestripsareseparatelytransportedforward,pressedtogether,andfinallywindedtogetheronawindingrollers.Theresponseofthewindingcontrolsystemishighlyrelatedtothewindingspeed(i.e.,linespeed),whichmayvaryduringthewindingprocess,e.g.,accelerateintheinitialtransientstate,ordeaccelerateintheendingtransientstate.Hereitisaimedatdesigninganadaptivecontrollerthatcouldcopewithwindingspeedvariation.Firstly,anmathematicalmodelisderivedforthewindingsystem.ThenaPIcontrollerandaPIDcontrolleraredesignedforlinespeeds2m/s,4m/s,6m/s,8m/s,12m/s,and16m/s.Finally,amathematicalrelationshipisestabishedforthecontrollerparametersbasedonthelinespeed.MATLABsimulationisconductedtoverifythefeasibilityoftheaboveadaptivecontroller.Simulationresultsshowthat,fortheno.1aligningsystemwithaPIcontrolandanintermediaterollerbeforethepressroller,thesystembandwidthisbetween1.37rad/sand17.4rad/sunderdifferentlinespeedsfrom1m/sto16m/s.Asforonewithoutanintermediateroller,thebandwidthrangewouldbebetween2.56rad/sand25.4rad/s.Thatis,thebandwithforthesystemwithoutanintermediaterollerisabout1.64timesofthatforonewithanintermediateroller.Simulationresultsalsoshowthattheno.1aligningcontrolsystemwithanadaptivePIcontrollerpossessesa5%overshootunderan1mmstepdisplacementcommandduringlinespeedacceleratingform0m/sto2m/s.Onthecontrary,thealigningcontrolsystemwithoutanadaptivePIcontrollerpossessesa65%overshootunderthesameconditions.Simulationalsoshowthattheno.1aligningcontrolsystemwithanadaptivePIcontrollerpossessesa5.5%overshootunderan1mmstepcommandduringlinespeedacceleratingform0m/sto16m/s.Underthesameconditions,theonewithoutanadaptivePIcontrollerhasa62.5%overshoot.Inaddition,theno.1aligningcontrolsystemwithanadaptivePIDcontrollerpossessesan11%overshootunderan1mmstepdisplacementcommandduringlinespeedacceleratingform0m/sto2m/s,whiletheonewithoutanadaptivePIDcontrollerhasa45%overshoot.Andtheno.1aligningcontrolsystemwithanadaptivePIDcontrollerhasa3.5%overshootunderan1mmstepdisplacementcommandduringlinespeedacceleratingform0m/sto16m/s,whiletheonewithoutanadaptivePIDcontrollerhasa60%overshoot.Fromtheaboveresults,theovershootinthesystemwithanadaptivePI/PIDcontrollerwouldbemuchlowerthanonewithoutanadaptivePI/PIDcontroller.Inaddition,theadaptivePIcontrollerismoreeffectivethantheadaptivePIDcontrollerfromthepointofviewofanovershoot,buttheadaptivePIDcontrollerismoreeffectivethantheadaptivePIcontrollerforconsideringthevalueofitsbandwidth.Experimentalverificationisconductedbasedonathree-layerwindingsystemwithaPC-basedcontroller.Theresultsshowthatthestandarddeviationsoftheno.1windingsystemandno.2windingsystemarebetween0.24and0.89mmunderlinespeeds1m/s,4m/sand8m/s.Inaddition,thestandarddeviationofthesystemwithanadaptivePIcontrollerisbetween0.05mmand4.04mmandtheaveragestandarddeviationis1.21mm.ThestandarddeviationofthesystemwithoutanadaptivePIcontroller(i.e.,withapurePIcontroller)isbetween0.04mmand2.24mmandtheaveragestandarddeviationis1.41mm.Therefore,itcanbeconcludedthattheadaptivePIcontrollerhasbetterperformancethanthepurePIcontroller
Study on the Design Automation System of Progressive Dies for the Automobile Radiator Cover
No full text[[abstract]]連續沖壓模具有快速大量生產產品之優勢，且具高經濟價值，在工業製造上被廣泛的使用，而連續沖模設計的過程繁瑣耗時，且須應對不同設計需求，因此客製化的模具設計自動化將成為未來趨勢。本研究之主要目的是探討連續沖壓模具自動化設計系統，並以汽車散熱器水箱蓋為例。本研究將系統架構於三維電腦輔助設計軟體CATIA，並應用VisualBasic6.0(VB)程式語言進行CATIA的二次開發來建構此自動化設計系統，利用知識及電腦化的優勢來縮短設計時間，讓模具設計工程師能更快速且簡單的進行連續模具設計。本研究是以將程式模組化的方式建構此自動化設計系統，系統內包含了使用者介面、模具主結構設計模組、標準零件設計與組裝模組、2D圖出圖模組、材料表匯出模組，以及為了提升系統設計彈性的零件即時顯示點放功能，系統在結構設計主要是以將基礎模塊之外型特徵與料條上工程特徵進行替換，替換後基礎模塊外型即改變成實際所需之形狀，以此方式將逐步將模具結構設計堆疊而出。本系統依照使用者介面輸入之參數值，與匯入的料條資料使系統自動執行連續沖壓模具結構設計、標準零件設計與裝配、零件2D圖出圖、材料表匯出。以汽車散熱器水箱蓋實例驗證，結果顯示本系統可大幅縮短模具設計時間提昇模具開發效率。[[abstract]]Theprogressivediehassuperiorityofthelargenumberproduction,iswidelyusedinthemanufacturing,theprocessofdesignprogressivedieistime-consumingandtediousalsoneedstodealwithdifferentdesignrequirements,thereforecustomizeddesignautomatedsystemofdieswillbecomethefuturetrend.Thepurposeofthisresearchistostudythedesignautomatedsystemofprogressivediesfortheautomobileradiatorcover.Thisresearchstructureon3DcomputeraideddesignsoftwareCATIA,andtheapplicationofVBprogramminglanguagetodothere-developmentofCATIAforconstructionthedesignautomationsystem,reducethetimeofdesignbyusingtheknowledgeandthecomputerizationadvantage.Letengineerscanmorequicklyandeasilytodesignprogressivedies.Theconceptofthesystemisdividedintoanumberofprogrammodules.Themodulescontainsuserinterface,structuraldesignofprogressivedie,assemblyforstandardlibrariesofprogressivedie,2DDraw,BillofMaterial,andinstantdisplaydesignofpartsthatinordertoimprovethesystemdesignflexibility,thesystemusedthemethodtodesignstructuralofdiesthatmaketheappearancecharacteristicofbasemoduleschangewithappearancecharacteristicsofmaterial.Thesystemaccordingtotheinputparameteronuserinterfaceandmaterialdatatodothestructuraldesignofprogressivedies,assemblyforstandardlibrariesofprogressivedies,2DDraw,BillofMaterial.Usingtheautomobileradiatorcoverasanexampletoverifythissystem.Theresultsshowedthatthesystemcansignificantlyreducethetimeofdesignandenhancetheefficiencyofdiedesign
Study of a Surface Roughness Modulation System for CNC End Milling
No full text[[abstract]]端銑削加工之表面品質會受到加工參數的影響，而過去傳統加工參數的設定都是選擇較保守的加工參數來進行加工，難以掌握表面品質，且花費許多時間。因此，在需求的表面品質下，研究能預測表面品質，以及未達到需求表面品質的加工參數調變的機制。本文主要研究表面粗糙度調變系統，系統主要分兩個部分表面粗糙度預測模型和表面粗糙度調變機制。表面粗糙度預測模型是使用類神經網路建立，其輸入為主軸轉速、進給速度、切削深度、切削力特徵輸入和-1，其中輸入-1為閥值提供了額外的自由度，使得訓練神經網絡的過程得到加速，依據直交表做端銑削實驗建立出75組數據集來做類神經網路訓練，而切削力特徵選取分別使用RMS、PEAK和絕對值平均，切削力軸向分別選用以下這五種不同輸入，X軸向和Y軸向、XY軸合併、Z軸向、X軸向Y軸向Z軸向和XY軸合併Z軸向，網路結構之隱藏層分別取用單層與雙層兩種，神經元分別為1~10個，輸出為表面粗糙度，綜合以上可組成300種不同組合，之後單層和雙層分開比較誤差與平均精度選擇出單層與雙層最佳組合，最後比較單層與雙層最佳組合，選擇切削力特徵MEAN軸向為XY軸合併和Z軸向，網路結構6-10-1，訓練平均精度為95.21%，測試平均精度為92.72%，作為表面粗糙度預測模型。表面粗糙度調變機制，建立過程分成兩個部分建立數據庫和表面粗糙度調變控制系統。首先建立數據庫，第一步驟設定加工參數配置(固定主軸轉速和切削深度只調變進給速度)，第二步驟使用加工參數預測表面粗糙度之類神經網路模型，模型輸入為主軸轉速、進給速度、切削深度和-1，輸出為表面粗糙度，網路結構4-7-7-1，訓練平均精度為93.52%，測試平均精度為92.36%，使用此模型進行預測表面粗糙度，第三步驟將參數配置和預測表面粗糙度的結果以相同主軸轉速和切削深度為一組，經由排列組合建立出具有進給速度和表面粗糙度變化量之數據集。表面粗糙度調變控制系統分成自動調變和手動調變，自動調變使用非線性迴歸分析和類神經網路進行建模，輸入為加工參數和表面粗糙度差值，輸出為進給速度調變量，回歸分析模型預測平均精度為90.02%，類神經網路模型預測平均精度為88.97%，因為目前類神經網路數據集分成訓練和驗證數據集，而回歸分析還沒有，所以使用類神經網路，網路結構5-3-1，過程先用加工參數預測表面粗糙度模型，得到表面粗糙度預測值，計算預測表面粗糙度與設定表面粗糙度兩者差值，把表面粗糙度差值和加工參數代入進給速度調變量預測模型，得到進給速度調變量，控制表面粗糙度低於設定值，模擬結果顯示，表面粗糙度自動調變系統可將模擬範例的初始表面粗糙度0.425μm調降到0.385μm，低於設定的表面粗糙度0.4μm。手動調變使用加工參數預測表面粗糙度之類神經網路模型，以此模型輸入加工參數預測表面粗糙度，判別有無在設定表面粗糙度範圍內，如果在範圍外有兩種情況，高於設定表面粗糙度範圍，表示粗糙度精度比設定值差則調低進給速度，低於設定表面粗糙度範圍，表示粗糙度比設定值佳則調高進給速度，每次調變進給速度為10mm/min，持續調整到設定範圍內，模擬結果顯示，表面粗糙度手動調變系統可將初始0.425μm調降至0.408μm，介於需求的表面粗糙度0.4±0.01μm範圍內，表面粗糙度自動調變系統與表面粗糙度手動調變系統的實驗驗證將列為未來工作之一。[[abstract]]Thesurfacequalitybytheendmillingprocessisaffectedbyprocessparameters.Inthepasttheprocessparametersareusuallysetaccordingtosuggestionsoftoolingcompanies,andthusitisdifficulttocontrolthesurfacequalityorsometimesittakesalotoftimetoachievegoodsurfacequalitybytrialanderror.Therefore,thisresearchaimsatdevelopingthemechanismofpredictingthesurfacequalityforendmillingprocessesandalsothemodificationmethodoftheprocessparameterstoachievethedesiredsurfacequality.Thesurfaceroughnessmodulationsystemmainlyconsistsoftwosubsystems,i.e.,thesurfaceroughnesspredictionmodelandthesurfaceroughnessmodulationmechanism.Thesurfaceroughnesspredictionmodelwasestablishedusingabackpropagationneuralnetwork(NN),inwhichtheinputsarespindlespeed,feedrate,depthofcut,cuttingforcefeaturesandabias-1.Theinput-1providesanadditionaldegreeoffreedomforthethreshold,sothatthetrainingprocessoftheneuralnetworkisaccelerated,whilethecuttingforcefeaturesconsidertheRMSvalues,vibrationamplitudesandmeanvaluesoffivedifferentcuttingforces,whicharethecuttingforcesintheXandYdirections,theresultantonesintheXYplanes,theonesintheZdirection,theonesintheX,Y,andZdirections,andtheresultantonesintheXYplanescombinedwiththeonesintheZdirection.Asfortheneuralnetworkstructure,oneandtwohiddenlayersareconsidered,andthenumberoftheneuronsineachhiddenlayerarefrom1to10.TheoutputoftheNNmodelisthesurfaceroughness.TheabovevariationintheNNstructureresultsin300differentcombinations.ThentheMATLABNNtoolboxisutilizedtoconductNNtrainingandthebestNNstructureistheonewiththeMEANcuttingforcefeaturefortheresultantforcesintheXYplanescombinedwiththeonesintheZdirection,theneuronnumber6-10-1ineachlayer.ThisbestNNmodel,withthetrainingaccuracy95.21%andthetestingaccuracyis92.72%ischosenasthesurfaceroughnesspredictionmodel.Asfordevelopingthesurfaceroughnessmodulationmechanism,adatabaseisfirstestablishedandthenasurfaceroughnessmodulationsystemisbuiltupbasedonthedatabase.Toestablishthedatabase,thefirststepistosettheconfigurationofmachiningparameters,i.e.,fixspindlespeedandcuttingdepthbutonlyadjustthefeedrate.Thesecondstepistoestablishaneuralnetworkmodelwhichutilizesspindlespeed,feedrate,anddepthofcut(thefeaturesofcuttingforcesisnotincludedhere)topredictsurfaceroughness.TheresultantNNmodelwithastructureof4-7-7-1possessesatrainingaccuracyof93.52%andatestingaccuracyof92.36%.Thethirdstepistoprocessthemachining-parameterconfigurationandthenusetheNNmodeltopredictthecorrespondingsurfaceroughness.Intheconfiguration,asetofadatagroupisarrangedtopossessthesamespindlespeedanddepthofcut,butthemodulatedfeedrateandthecorrespondingsurfaceroughnesschange.Nextthesurfaceroughnessmodulationsystemadoptstwostrategies,i.e.,automaticandmanualmodulation.Theautomaticmodulationusesnonlinearregressionanalysisandneuralnetworks,respectively,toestablishtherelationshipbetweentheinputs(i.e.,spindlespeed,feedrate,depthofcut,anddesiredsurfaceroughnesschange)andtheoutput(i.e.,feedrateadjustment).Theaverageaccuracyofthenonlinearregressionmodelis90.02%,whilethatoftheneuralnetworkmodelmodelwiththestructure5-3-1is88.97%.Herethedatafortrainingtheneuralnetworksaredividedintothetrainingandvalidationsets,butthosefortheregressionanalysisarenotdivided.Therefore,theneuralnetworkmodelisadoptedfortheautomaticsurfaceroughnessmodulationsystem.Asforthemanualmodulation,itusesmachiningparameterstopredictthesurfaceroughnessbyaneuralnetworkmodel.Ifthesurfaceroughnessisworsethanthepresetvalue,thefeedrateisadjustedtobe10mm/minloweruntilachivingthepresetroughness.Simulationshowsthatbothoftheautomaticandmanualmodulationmethodscouldachievethedesiredsurfaceroughness.Theexperimentalverificationisleftasoneofthefutureworks