Abstract
While numerical models provide indispensable tools for solving advanced physics problems, reducing their high computational costs is a highly active research topic. Artificial intelligence (AI) is often used for this purpose by defining data-driven and high-performance models that learn to mimic the numerical models and eventually replace them with machine learning (ML). In this study, we review these so-called surrogate models, but with a focus on advanced articular cartilage (AC) modeling. AC is a low-friction soft tissue with excellent load-bearing capacities that covers and protects articulating bones, but given the high prevalence of cartilage damage due to biomechanical factors, surrogate models are used to efficiently study the multi-physics, particularly the biomechanics, of AC. To that end, we familiarize the readers with the key biological, numerical, and learning aspects of AC models. In particular, the implicit FE modeling as a well-founded numerical method is briefly explained in order to clarify its benefits and complexity at the same time. Next, we give a detailed overview of the relevant ML algorithms, and it is shown that while the general-purpose ML models can be used as a surrogate for the AC FE simulation, they potentially require large and expensive numerical datasets. This can be handled by hybrid surrogates, which are based on the application of simplified numerical models in the ML surrogates. We conclude this chapter by discussing future directions.