A feasibility study on potential of stress sensors to detect solder joint defects toward prognostics of power modules

Abstract

The superiority of stress sensors over temperature sensors for detecting defects in solder joints was investigated. The effects of joint defects on device stress and temperature in a single-sided cooling structure were obtained via thermal stress simulation. The peripheral, edge, and central defects were assumed as the defect distributions. The central defects did not result in a remarkable occurrence rate of the stress waveform change. In contrast, the periphery and edge defects caused a large occurrence rate of the stress waveform change compared to that of the temperature. In general, the thermal strain of solder joints in a power module causes its initial degradation from the joint edge or periphery. Thus, the stress waveform provides useful information for detecting initial degradation. The change pattern of the temperature waveform because of the solder joint defects only caused an amplitude increase. In contrast, stress waveform changes exhibited four different patterns: amplitude increase, amplitude decrease, waveform inversion, and waveform disappearance. These results confirm that stress sensors are better than temperature sensors for detecting waveform changes caused by slight joint defects at peripheral or edge, particularly with a small number of sensors. Application of the data acquired via stress sensors to machine learning will allow estimation of the joint defect distribution. Furthermore, time-series stress waveform data can provide prognostics of solder joint degradation.