联合mRMR算法和BP神经网络的集成电路测试方法

An Integrated Circuit Test Method Based on mRMR Algorithm and BP Neural Network

集成电路规模的急剧增大显著加了测试成本。针对集成电路测试成本过高的问题,提出了一种适应性测试方法。将最小冗余最大相关算法与BP神经网络相结合。首先通过最小冗余最大相关算法选择重要的测试项,仅测试重要的测试项并组成特征集合,然后使用BP神经网络模型预测测试结果。实验结果表明,相较于传统测试方法,该方法以牺牲0.1%的测试逃逸率为代价,降低了45%以上的测试成本。与其他适应性测试方法相比,该方法的测试逃逸降低91%以上,可以在测试成本和测试质量之间选择最优解。

The rapid increase in the scale of integrated circuits has led to test cost increasing issues.In order to solve the problem of much too high integrated circuit test cost,an adaptive test method was proposed.The method combined the minimum redundancy maximum relevance algorithm with the back propagation neural network.Firstly,the important test items were selected through the minimum redundancy maximum relevance algorithm,and the only important test items were tested to forme a feature set.Secondly,back propagation neural network was used to predict the test results with the feature set.The experimental results showed that,compared with the traditional test methods,the proposed method could reduce the test cost by more than 45%at the expense of 0.1%test escape rate.When compared with other adaptive test methods,the test escape of the proposed method could be reduced by more than 91%.An optimal solution between test cost and test quality could be chosen by this method.