Time to Failure Prediction on a Printed Circuit Board Surface Under Humidity Using Probabilistic Analysis

This paper presents the probabilistic study of time to failure (TTF), which is caused by combinations of various important controllable factors on a printed circuit board (PCB) surface under humidity. The study investigated the impact of four changeable factors including pitch distance, temperature, contamination, and voltage, each at three levels upon the surface insulation resistance test boards. Constant 98% relative humidity with adipic acid as contamination related to flux residue was used for a 20-h parametric experiment. Two main states were considered on the whole output current measurements: the stable part before the short transition phase and the unstable part after due to electrochemical migration (ECM) on the PCB surface. Leakage current (LC) in the first state and TTF at the beginning of the second stage was measured with five replications for each condition as the predictive indicator in all models. The trend of LC and TTF was also investigated on three levels of each factor. In addition, probabilistic distribution analysis using fitted Weibull distribution, multivariate regression analysis, and the classification and regression tree (CART) analysis were used to predict the probability of TTF and failure risk prediction on the PCB surface. All the prediction models had an acceptable prediction of TTF at diverse accuracy levels, according to changing factors/levels. Nevertheless, the multivariate regression analysis had the best prediction, highest R2, and lowest error compared to the other models.