摘  要：針對目前復雜電路測試、診斷過程中故障樣本獲取困難的問題，提出一種隨機變異生成策略。首先定義適用于隨機方法的8種限定變異算子，可覆蓋實際電路中的大多數真實故障；然后應用層次分析法（AHP），分別考慮電路組成元素間的相對故障概率和故障類型，計算出不同變異位置以及對應變異算子的被選權重；考慮連續型故障參數的分布，通過隨機采樣確定具體的參數值，使故障模型更為真實；最后，以一個串聯調整型穩壓電路為實例，為其生成500個隨機變異體。實驗結果表明：該方法能夠快速有效地為多層次模擬電路生成大規模、高質量的故障樣本。
關鍵詞：自動故障注入；隨機變異生成；變異算子；層次分析法
文獻標志碼：A       文章編號：1674-5124（2016）01-0012-04
A stochastic mutant generation method for multi-level analog circuits
TANG Xiaofeng1，2， XU Aiqiang1， LI Wenhai1
（1. Department of Scientific Research，Naval Aeronautical and Astronautical University，Yantai 264001，China；
2. Unit 92514 of the PLA，Yantai 264007，China）
Abstract: A stochastic mutant generation method has been proposed to acquire automatically multiple fault samples from complex circuits for testing and diagnostic purposes. First， eight restricted mutation operators are defined to cover the majority of actual circuit faults. Analytical hierarchy process （AHP） is used to analyze the relative fault probabilities and fault types of circuit elements and evaluate the weights of different mutation locations and operators selected. Then， the probability distributions of continuous fault parameters are considered and the fault model is made more realistic by random sampling of fault parameter values. In the end， 500 stochastic mutants are generated for a regulator circuit. The experimental results show that the proposed method can rapidly generate massive high-quality fault samples for multi-level analog circuits fast and well.
Keywords: automatic fault injection； stochastic mutant generation； mutation operator； AHP