Prediction of forming temperature in electrically-assisted double-sided incremental forming using a neural network

Electrically-assisted double-sided incremental forming (EA-DSIF) is a flexible forming method suitable for processing hard-to-form materials and complex-shaped parts. A challenge in EA-DSIF experiments is temperature measurement. Since the localized forming zone is blocked by the tools, it is not possible to measure the actual forming temperature distribution in the forming zone. To address this issue, we propose an artificial neural network (ANN) framework for predicting the forming temperature using measurements of the surrounding temperature and toolpath features. The ANN model was trained using the temperature outputs of finite element models. A simplified EA-DSIF simulation model was developed for computational efficiency needed for synthetic data generation. Model simplifications were justified in multiple cases and validated with experimental data by comparing the temperatures from positions that is visible to an infrared camera. The feasibility of applying the developed ANN model to untrained geometries and in practical applications was demonstrated. The findings generated from this study are crucial for selecting optimum process parameters, estimating the forming force, and predicting microstructure evolution during EA-DSIF.