中文摘要：

针对常规三角分解方法计算模拟电路规范模糊组和可测试值存在误差的问题，提出一种基于测试矩阵基向量的变异粒子群算法．该算法分析了电路的各阶规范模糊组参数成分，根据粒子速度和位置初始化找到所有二阶模糊组，由低阶模糊组确定高阶模糊组的构成参数，通过粒子速度变异找到所有规范模糊组．实例验证表明：该算法不通过三角分解就能够找出所有的规范模糊组，提高了计算精度，同时也降低了计算复杂度．

英文摘要：

A variation particle swarm optimization algorithm based on the matrix basis to search the ambiguity in analog circuit is proposed to improve the round-off error caused by routine triangle decomposition method. It firstly analyzes the components of canonical ambiguity groups, finds all second order ambiguity groups via the initialization of particle swarm, and then chooses the components of higher order ambiguity groups based on lower order ambiguity groups to get all canonical ambiguity groups through variation of particle velocities. It is demonstrated by an example that the proposed algorithm can find all canonical groups without triangle decomposition and heightens the precision and decreases the complexities of computation.