Modeling Research of Electrochemical Migration Failure on Printed Circuit Board

With the integration of electronic products and progress in electronic industrial technology, the assembly density of printed circuit board (PCB) is greatly improved, so that the spacing between wires (solder joints) is becoming smaller and smaller. At the same time, the working environment of electric products is becoming more and more diversified. High temperature, high relative humidity, and environmental contamination brings huge risk for the electrochemical migration (ECM) of PCB with high density. It is very necessary to build up the life model, which describes the relation between the time to failure (TTF) of ECM and the environmental factors of the temperature, the relative humidity, and the bias voltage in the circuits. In this paper, two methods to establish the life model of ECM on PCB were proposed and compared. One is the modeling based on the failure physics of ECM; the other is the modeling by machine learning. To fit the unknown coefficients in the model of failure physics and to train the model by machine learning, the data of TTF of ECM on PCB were obtained by accelerated experiments under temperature humidity bias (THB) conditions. Finally, two models were evaluated by the test data set. The model established by random forest regression algorithm in machine learning has better life prediction performance. By this research on modeling methods, it was verified that the modeling method by machine learning can be applied in the field of ECM failure under complex environmental conditions.