在維修產業中,判斷故障原因是複雜且技術性的步驟,往往需要由有經驗的 維修工程師來執行,造成薪資成本高且培育不易等種種人事問題,因此這個問題 一直困擾著維修產業。本研究運用了動態統計、模糊決策樹、權重式案例推理、 及依案例績效與理想解相似度排序之技術等方法,建立了一個智慧型維修系統, 並提出了五個績效指標,以某手機維修單位一個月份的真實資料為分析資料,其 中分別以三種維修量較大的廠牌手機進行分析。以四種故障判斷方法所得的績效 結果與現行作法加以比較,結果顯示本研究所提出之四種方法均優於現行作法, 而其中以模糊決策樹和權重式案例推理的績效最優,修復手機的時間改善量分別 高達現行作法的21%與16%,若換算成薪資,其數目將更為可觀。 本研究的貢獻包括: 1. 四種故障判斷方法,在績效上皆優於現行故障判斷方法。 2. 以四種方法論工具為基礎,設計一個自動化的智慧型故障判斷系統。並 以手機維修產業真實案例為分析資料,對三大廠牌手機進行分析,結果 對實務界有一定程度的幫助。
In the repair industry, diagnosing the cause of failure is complicated and time consuming. It usually takes experienced repair engineers and trial-and-error processes are often involved causing man power availability and high-cost problems. The study used dynamic statistics, Fuzzy Decision Tree (FDT), Case Based Reasoning (CBR), and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to establish an intelligent repair system. One month width of real data on the repairs of cellular phones from three leading brands were collected and analyzed by the intelligent repair system. The results shown that all form tested diagnostic methods are better than the existing. In particular, the FDT and CBR methods show improvements of 21% & 16% in time to locate and fix problems. Contributions of this research include: 1. Proposing four methods of repair diagnosis approach which are all better than the current approach. The FDT method is especially recommended. 2. Establishing an intelligent cellular phone diagnosis system which can be used in the industry.