本研究選定台灣某一半導體DRAM製造公司的三個廠區共27個部門為對象樣本，變數包含部門特質外生變數（人數、機台數、化學品數、作業項數等6項）、安全衛生管理系統（OHSAS-18001）績效評估分數與四類異常事故件數共11項；研究架構依據ISO-14031績效評估將以上三類變數（variables）視為投入（input）、中介變數（process）與產出（output），依據此一架構運用相關矩陣、多元迴歸與多變量變異數分析等統計與數量方法驗證投入產出變數間的關連性，結果顯示：組織特質與安全績效、管理成績相關，而組織安全管理對於和部門安全績效的影響沒有組織特質來的明顯。 其次運用資料包絡分析方法（Data Envelopment Analysis,DEA）針對各項投入與產出變數進行部門別安全管理效率評估。在兩階段安全效率的分解結果顯示：第一階段效率值約在0.4-1之間，顯示個案公司成立十餘年，各項管理措施與規章制度完備；而第二階段管理措施有效性的效率值差異落差介於0.000001-1之間，顯示對於部份部門而言，管理措施對於減少與預防各類異常事故的發生，沒有發揮應有的功能與作用。 最後以廠別、作業特質、管理成績等級等三項質性變數，針對DEA所得效率值以無母數方法進行差異檢定，結果發現除廠區別間安全效率沒有差異外，其餘二者不同群組間的效率值均有顯著差異，同時管理成績越差的群組，其組內安全效率的分配有偏低的趨勢。 本論文為國內外首次將DEA方法應用於工業安全績效評估領域的研究，証實資料包絡分析方法可適用於製造業工廠部門別安全績效的評估，並提供分析步驟/流程相關參考建議，就DEA模式的選用，建議採規模變動報酬VRS與單方面的投入或產出導向Envelopment Model即可，同時在負向產出的處理上，採行分離（separating）與INP模式即可。藉由DEA方法的運用，可以突破一般安全衛生績效評量，只使用一個負面的指標(受傷及疾病的數據)來進行量測的限制，同時考量多個投入與產出變數，同時變數間不須訂定具有爭議性的權重，可充分滿足實務運作針對績效評估的需求。 ;Selected a Taiwan DRAM semiconductor manufacturing company with three plants and total 27 departments as case and samples. Study variables included six departmental characteristics variables(the number of machines、 chemicals、employees、operation items), one management variable（performance evaluation scores of OHSAS-18000), and four safety performance variables（number of abnormal events, such as gas leak/injury..）. Research framework based on ISO-14031 assessing the six departmental characteristics variables as input (input), OHSAS evaluation scores as intermediate variables (process), and four safety performance variables as outputs . On this framework use correlation matrix, multiple regression and multivariate analysis of variance statistical technical to verify the relationships between variables. The results showed that: safety performance variables and OHSAS evaluation scores were related to departmental characteristics variables. OHSAS evaluation scores did not affect the safety performance obviously when combine departmental characteristics variables . Next, use Data Envelopment Analysis（DEA）and the input/output variables to evaluate safety management efficiency of the 27 departments. The resaults of two-stage DEA efficiency showed：The efficiency of first stage was about 0.8-1, indicated this company have well managerial regulations. And the efficiency of secondary stage was about 0.1-1, indicated these managerial regulations could not well prevent the incidents. Finally, used category variables such as category of plants/operating characteristics, to test the significance of DEA efficiency. The reseaults showed that difference between operating characteristics and evaluation scores were significant（p<0.05）. This is the very first thesis using DEA in the region of industrial safety management and comfirm DEA being a good method to evaluate safety performance. Traditional Safety performance evasluation using sole injury frequency/servity as indicator, or using controversial weightings to combine several indicators into one final number. DEA make breakthroughs of these limits, and fulfill the needs of factory safety performance evaluation.
