A machine learning-based methodology for identification of the plastic flow in aluminum sheets during incremental sheet forming processes

This study presents a calibration method based on machine learning techniques to identify parameters of hardening law of aluminum alloy sheets in complex manufacturing processes. A V-shape test is designed to characterize material behavior during an incremental sheet forming (ISF) process. A series of virtual materials is first generated using three physical features observed in a standard uniaxial tensile test: initial yield stress, maximum uniform plastic strain and yield-to-strength ratio. These virtual materials are then employed in simulating the designed V-shape tests to numerically collect the material responses, such as forming forces, displacements, or their combinations. Several feed-forward neural networks (FFNNs) are developed and trained to relate the collected material responses to the relevant virtual materials. Then, the trained FFNNs were used to estimate the flow curve of AA5052-H32 sheets up to a plastic strain value of 1.0 using the experimentally measured response from the V-shape test during ISF. The FFNN-based flow curves appear to capture well the stress–strain data obtained from the uniaxial tensile test with the coefficients of determination up to 0.98. The identified flow curves are also employed to simulate the uniaxial tensile and ISF truncated cone tests. The simulated uniaxial tensile forces are in good agreement with the experimentally measured results. A maximum difference of 5% is observed in the comparison between the simulated and measured ISF loading forces in the steady-state deformations. The comparisons demonstrate the efficiency of the proposed method to characterize the plastic flow of metal.