Reliability modelling – data-driven and physics-based methods

Reliability modelling is widely used in industrial applications, ranging from structural and mechanical engineering to material design. Regardless of the application, reliability is defined as the probability that an intended function is satisfied in the presence of uncertainties. Reliability modelling imposes challenges in integrating knowledge from multiple disciplines such as material modelling, manufacturing, optimization and uncertainty quantification. In this lecture, both data-driven and physics-based methods will be discussed. Data-driven methods include advanced statistical analysis and probabilistic techniques as well as machine-learning. Physics-based methods focus on the development of stochastic constitutive material models at different material length-scales. This presentation will summarize recent research activities and highlight challenges and opportunities within reliability modelling.