Artificial Intelligence (AI) and data-driven decisions based on Machine Learning (ML) are making an impact on an increasing number of industries. As these autonomous and self-learning systems become more and more responsible for making decisions that may ultimately affect the safety of personnel, assets, or the environment, the need to ensure the safe use of AI in systems will be crucial to safety management and operations. Machine learning for high-risk and safety-critical applications in particular is challenging, as there is a reduced tolerance for erroneous predictions due to potentially catastrophic consequences. The models which fit the data well are also often opaque, making them less falsifiable and difficult to trust. Moreover, relevant data is usually scarce, and a proper treatment of uncertainty is essential, as we are not only concerned with what is likely to happen, but also with less likely events that may happen. However, there are some positives - namely, that there is often additional causal and physics-based knowledge available.
In this presentation I will introduce how probabilistic machine learning can be applied, in combination with physics-based simulations or experiments. I will introduce how Gaussian processes can be applied to tackle the numerical challenges that come with expensive computer models or experiments, and how phenomenological knowledge can be used as constraints on Gaussian processes for improved performance.