簡稱R的某半導體廠導入SPC (Statistical Process Control system) 系統，輔助廠務監控系統FMCS (Facility Monitoring and Control System) 提升預警功能。結果成效不彰，主因是無效警報(False alarms)過多，工程師處理無效警報負荷過大。 如何使無效警報數量降低到參與人員可以接受的水位，讓系統在真實警報不遺漏之下，發揮應有的預警功能，是本文研究探討的主題。本文將說明R廠如何運用訊號偵測原理 (Signal Detection Theory) 的概念，提出新模型改善原有SPC系統，並結合異常事故管理審查機制，進一步強化新模型，並以實廠數據模擬、驗證新模型可有效控制無效警報數量，並且不會削弱原本預警功能，達到運用即時製程統計管理強化廠務監控系統以提升半導體廠廠務系統供應可靠度及安全性的目標。 ;A semiconductor factory named R introduced a SPC (Statistical Process Control system) system into the monitoring system of the facility for its utility quality control. Too many False alarms make engineers overloaded. Eventually, the SPC system cannot function as expected. How to reduce the number of false alarms to an acceptable level under the condition of none real alarms compromised is the main topic of this study. This thesis proposes a new model explaining how to make use of Signal Detection Theory method to improve the old SPC systems. Moreover, I also embed an abnormal event management reviewing mechanism to perfect the new model. The performance of new model is verified by real data simulation. The simulation results show the model capability of controlling the number of false alarms without downgrading the original warning functions of the system. The new model is proven to be a better feasible system that can strengthen the FMCS (Facility Monitoring and Control System). The proposed real-time statistical process control system is on the way to improve the reliability and the safety of the semiconductor plant R. This thesis proposed some useful information for the company which introduces the SPC into their facility monitoring program.