Predicting X-Sensitivity of Circuit-Inputs Based on Random Forests

As modular circuit design becomes more and more complex,uninitialized timing cells,black boxes in the design,clock-domain interface and erroneous behavior of analog-to-digital converters may produce unknown logic value(X)at various circuit nodes.The existence of X-value decreases the test coverage of the test set.For the X-sources that X-value exists on the certain primary or secondary inputs of a logic circuit,this paper presented a method based on machine learning to predict the X-sensitivity which measures the effect of X-source on test coverage.Firstly,the basic structure parameters of the circuit are calculated by topology algorithm.Then the circuit is divided into three parts,and the specific circuit characteristic parameters are extracted as the original data set.Finally,the random forest model is used to train and predict the datasets which obtained in all circuits.Some of the circuits in ISCAS’89 and ITC’99 are selected as the data set source.Compared with the existing prediction methods,the test set accuracy of proposed method is 90.27%with 14.69%higher,and the large circuits accuracy of proposed method is 93.32%with 19.49%higher.Experimental results show the higher accuracy and better generalization ability of proposed method.