| Popis: |
When a system is in idle/starting-up state, Field-Testing is a promising way to guarantee the reliability of an advanced system. However, the extremely limited test application time obstructs the implementation of field test. In this paper, we introduce a test pattern partitioning approach by using two well-known machine learning algorithms: Simulated Annealing (SA) and Support Vector Machines (SVM), to derive an optimal solution for pattern partitioning that minimizes the test latency for high reliability. From the experimental results on benchmark circuit we show that both SA and SVM based method can significantly improve the test latency of partition test, and SVM is much more efficient than SA. Those results confirm the feasibility of machine learning algorithm for the pattern partition problem. |
