基于混沌多目标粒子群算法的测点优选方法

Test point selection based on chaotic multi-objective DPSO algorithm

模拟电路的测点优选问题旨在寻找总测试代价最低、测试性能最好的测点集合。之前的研究通常把最高测试性能作为约束条件,以单目标优化思想搜索代价最小的测点集合。而实际应用中,对测点集合的需求是多样化的,需要根据实际需求调整性能和代价间的关系。提出了混沌多目标粒子群算法,该算法采用多目标优化思想,能同时找到多样化的方案,并根据多目标优化问题的特点,加入了混沌机制提高算法的搜索能力。实验结果表明,该算法能找到对应不同测试性能的最优测点集合,与其他算法相比,算法成功率最高、找到的方案最多且运行速度较快。对模拟电路的可测性设计和故障诊断很有帮助。

Analog test point selection aims at finding the minimal test point set and obtains the highest fault isolation rate （FIR）. Almost all previous researches considered the highest FIR as a constraint and proposed many methods to find the minimal test point set. Of course a higher FIR is better, but practically, especially in large analog circuits, to obtain the highest FIR usually takes a great deal of time and money. Quite often, if just to obtain an approximate highest FIR, the expense of time and money can be greatly reduced. Therefore, in this paper, a novel chaotic muhi-objective discrete particle swarm algorithm is proposed for analog test point selection. This method finds more solutions than previous methods because it uses Pareto optimality theory to solve the test selection and it compares the solutions according to Pareto dominance. Each particle maintains a Pareto set as personal best, and global best is also a Pareto set. To improve its global-searching ability, chaos technique is applied in this method. The experimental results showed the following： 1 ） The proposed method found not only the solutions with the highest FIR, but also found the solutions with approximate highest FIR; 2） It found the minimal test point set for each FIR; 3） Compared with other multi-objective algorithms, the proposed method achieved the highest accuracy, found the most solutions and consumed a shorter time. It can provide more solutions with different FIR for circuit designers to choose, which is beneficial for design-for-testability and fault diagnosis.