A key limitation of finite element analysis is accurate modelling of material damage. While additional material models exist that improve correlations between simulated damage and experimental data, these models often require additional parameters that are difficult to estimate. In this work we show that Bayesian optimisation, a machine learning technique, can be used to identify material model parameters. We show that Bayesian derived material model parameters result in simulated output with less than 2 % error compared to experimental data. The framework detailed here is fully autonomous, requiring only basic information that can be derived from a simple tensile test. We have successfully applied this framework to three datasets of P91 material tested at ambient (20 ◦C) and higher (500 ◦C) temperatures.