針對故障診斷過程中的特徵選擇問題，提出一種基於多種群量子粒子群約簡演算法；該方法引入多種群分群的改進策略，利用兩層結構分別實現雙向搜索，保證群體進化快速收斂於最小約簡，選出真正代表故障的特徵變數；通過對田納西－伊斯曼（Tennessee-Eastman, TE）公司的故障診斷應用，並與主元分析法等實驗對比，結果表明，基於MQPSO的粗糙集屬性約簡演算法應用於故障診斷具有非常實際的意義，能夠最大程度上提高故障診斷的正確率。

Considering the feature selection problem in a fault diagnosis process, the author proposes a method based on quantum-behaved particle swarm attribute reduction by a multi-swarm algorithm. This method introduced a multi-swarm clustering strategy, using a two-tier structure to realize a bidirectional search, to ensure fast convergence to the minimum population reduction, and select feature variables that truly represent the characteristics of the fault. Through the application of Tennessee-Eastman's fault diagnostic and other experiment using principal component analysis, the results show that a rough set attribute reduction algorithm based on MQPSO is practical and can improve the accuracy of fault diagnosis to the greatest extent.