Feature Extraction and RUL Prediction of SAC Solder Alloy Packages by Different Statistical and Time-frequency Analysis Techniques under Simultaneous Temperature-Vibration Loads

This study focuses on finding feature vectors from the strain signals of electronic assemblies under vibration load of 5 G acceleration levels at different temperatures. The strain measurements measured from the strain gauges fixed at different locations of the PCB were used to acquire signals at the different time of vibration of the packages. Different analyses were done in finding the feature vector for predicting failure of the packages under combined temperature and vibration loads. These include statistical as well as time-frequency based signal processing techniques. Frequency based techniques were used to understand the frequency content of the signal during vibration. Time-Frequency based techniques like spectrogram and Periodogram were used to find the change in the frequency content of the signal with time during vibration. Statistical techniques were combined with parametric and non-parametric spectral estimation techniques for identification of the feature vector. Higher order time-frequency distribution techniques like Wigner-Vile and Pseudo-Wigner vile distributions were also used to identify and verify the feature vectors. A combination of statistical and frequency based techniques were used to get the variation of different feature vectors and identify the patterns of the feature vectors with the change in the temperature levels. These feature vectors are used to predict the RUL of the packages.