Exploring the Data-Driven Modeling Methods for Electrochemical Migration Failure of Printed Circuit Board

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

“…The NMSE is 0.328, 0.247, and 0.271 respectively, which is significantly more accurate than the model based on failure physics (The NMSE is 0.908.). It has been preliminarily verified that it is valid for machine learning in the modeling of life prediction of circuit board based on ECM failure [10].…”

“…It can be seen that when the influence parameters are increased to three, the model based on the failure physics of ECM has been very complex. Moreover, through the verification of experimental data, the standard mean square error (NMSE) of TTF calculated by (4) is 0.908, indicating a rather large prediction error [10]. However, if the soluble salt in dust contamination is further introduced, it is difficult to directly derive a theoretical model of ECM failure of PCB based on the chemical reaction on the electrode.…”

“…Then, the effects of the temperature T, the relative humidity H, the bias voltage U and the soluble salt concentration C on the ECM failure of the circuit board are considered as the effect of the four independent factors to multiply corresponding to (1) ~ (3) and (9), as (10).…”

Using machine learning algorithms to predict failure on the PCB surface under corrosive conditions

“…Furthermore, there are some studies on PCB using ML algorithms have performed like fault recognition regarding the PCB glue [38], detection of various types of defects in PCB inspection [39], reliability prediction of solder joints failures [40], prediction of solder joint health [41], and evaluation of the life prediction effect of ECM on PCB using three regression methods [42], which these only used some of the ML algorithms, or they used one of the classifications or regression analysis…”