| Summary: | Approximately 70% of electronics package failures occurred due to solder joint failure. Thus, the strength of the solder joint becomes essential in determining the solder joint’s life. Stress–strain curves from various aging conditions, temperatures, and strain rates are necessary to predict the non-linear behavior of solder joints under loading. The stress-strain curves of solder alloys can be perfectly fit using a hyperbolic tangent empirical model. Accurately forecasting these constants model is crucial in properly determining the inelastic model parameters for solder alloy deformation. This paper presents an artificial neural network (ANN) model for predicting the stress–strain curves of 96.5Sn-3.0Ag-0.5Cu (SAC305) solder alloy with various aging conditions, temperatures, and strain rates. Each stress–strain curve from published literature is fitted to the hyperbolic tangent model. The model parameters obtained from curve fitting are trained to the ANN model. The input of the ANN model are aging periods (0, 24, 120, 480, 1440, 2880, 4320, 6480, and 8640 h), temperatures (25 °C, 50 °C, 75 °C, 100 °C and 125 °C), and strain rates (0.00001, 0.0001 and 0.001 s−1). This study uses the 48 neurons in the hidden layer. Based on this ANN architecture, the mean squared error (MSE) and the coefficient of determination (R2) are 0.578 and 0.9995, respectively. In conclusion, the developed ANN model can accurately predict the stress–strain curves of SAC305 solder alloys. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
|
|---|