Developing Data-driven deformation models for tin using symbolic regression and genetic programming

Tin (Sn) exhibits complex deformation behavior characterized by significant dependence of strength on temperature and strain rate, which complicates predicting its deformation using traditional strength models. This work addresses this challenge by training a data-driven model on a set of compression tests at various strain rates and temperatures using genetic programming to perform symbolic regression. The strength model developed in this work showed increased accuracy compared to traditional strength models. Furthermore, the developed strength model adequately predicted independent experimental data (i.e., data that was not used to train the model). Results demonstrate that genetic programming successfully established a valid analytical function that adequately characterizes the temperature and strain rate dependent strength behavior of tin. Lastly, the analytical nature of the resultant model enables the possibility to perform high strain-rate simulations given that it can be incorporated into high-fidelity simulation codes.