Data-Driven Life Modeling of Electrochemical Migration on Printed Circuit Boards Under Soluble Salt Contamination

Under the air pollution environment, the dust enters the electronic products and deposits on the printed circuit boards (PCB) during the service life. Soluble salts in the airborne dust can reduce the critical humidity and increase the ion concentrations on condensed water film on PCB, which has been proven to aggravate the insulation failure of the high-density PCB caused by electrochemical migration (ECM). It is practical application significance for how to establish a life model to evaluate the ECM of PCB under the interaction of soluble salt with temperature, relative humidity, and electric field strength. In this article, through the temperature humidity bias acceleration experiment, the ECM failure of PCB under the contamination of the different concentrations of NaCl solution is simulated, and the ECM characteristics and failure mechanism are studied through the change of surface insulation resistance (SIR) and the morphology and element compositions of migration products. The time to the insulation failure of PCB under different conditions are obtained by the analysis of SIR curves. Based on the data-driven method, the life modeling of ECM failure of PCB under soluble salt contamination is studied by multivariate non-linear regression and machine learning methods, such as support vector regression, gradient boost regression tree, and random forest regression. It is proved that it is valid to use machine learning to establish the ECM failure life model of PCB in complex environments with limited life data.