Constitutive modelling of the TiB₂–B₄C composite by experiments, simulation and neutral network

The JH-2 constitutive relations of the TiB₂–B₄C composite were studied by the experiments, simulations and feedforward neural network. The flattened Brazilian disc experiments were designed to obtain the strength model of intact ceramic. The strength and strain rates sensitivity of the TiB₂–B₄C composite were studied by the static and dynamic tests. The Hugoniot elastic limit and equation of state of the composite were analyzed by the plate impact tests and Doppler pins test system. The confined compression tests for ceramic particles were designed to model the normalized fracture strength relations. Then the feedforward neural network model was established to train samples and analyze the damage constants of the TiB₂–B₄C composite. The samples of the input layer and output layer were derived from the experiments and simulation of the ceramic/metal composite armor impacted by high speed ball projectile. The Bayesian regularization algorithm was used to avoid converging on local minimum value in the neural network. Lastly, the constitutive model of the TiB₂–B₄C composite was verified by the comparison between the experimental and simulation results which were the depth of penetration, radial and circular crack of ceramic plate.