Assisting asset model development with evolutionary augmentation

In this talk, we explore how Genetic Programming can assist and augment the expert-driven process of developing data-driven models. In our use case, modellers must develop hundreds of models that represent individual properties of parts, components, assets, systems and meta-systems like power plants. Each of these models is developed with an objective in mind, like estimating the useful remaining life or detecting anomalies. As such, the modeller uses their expert judgment, as well as available data to select the most appropriate method. In this initial paper, we examine the most basic example of when the experts select a kind of regression modelling approach and develop models from data. We then use that captured domain knowledge from their processes, as well as end models to determine if Genetic Programming can augment, assist and improve their final results. We show that while Genetic Programming can indeed find improved solutions according to an error metric, it is much harder for Genetic Programming to find models that do not increase complexity. Also, we find that one approach in particular shows promise as a way to incorporate domain knowledge.