Rate-dependent plasticity and fracture of five DP-steels: static and dynamic high-throughput experiments and neural network modeling

A novel high-throughput split Hopkinson tensile bar system equipped with a load inversion device is developed and used to test five dual-phase steels (DP590, DP780, DP1000, DP1180, and DP1470) at strain rates up to 1000 s⁻¹, complemented by quasi-static (and intermediate-rate experiments ranging from 0.001 s⁻¹ to 1 s⁻¹. In total, >500 experiments are performed using uniaxial tension, notched tension, central-hole tension, and shear geometries, covering a broad range of stress states. The proposed form-fit clamping concept enables reliable testing of specimens up to 3.5 mm thickness and strengths exceeding 2000 MPa, reducing test cycle times to less than one minute. The experiments are simulated using a non-associated quadratic plasticity framework combined with a neural-network-enhanced deformation resistance to capture coupled effects of anisotropy, strain rate, and temperature. Fully thermo-mechanically coupled finite element simulations are conducted for all tests. Ductile fracture initiation is predicted using a Hosford–Coulomb model augmented by a neural network function that captures the effects of strain rate and temperature. All steels exhibit positive effect of the strain-rate on strain hardening, while the strain-rate sensitivity of fracture varies strongly between materials.