Mapping functions: A physics-guided, data-driven and algorithm-agnostic machine learning approach to discover causal and descriptive expressions of engineering phenomena

Abstract

This paper presents; mapping functions, a machine learning (ML) and simulation-free approach to enable physics-guided and data-driven derivation of expressions that describe engineering phenomena. In this approach, a series of ML models are first developed to examine a given phenomenon, and insights from their analysis, together with those obtained from physics principles, are then used to identify key features governing the noted phenomenon while satisfying the Law of Parsimony of Occam’s Razor. The identified features are subsequently explored via a search space to map the causality of the problem on hand into compact descriptive expressions which can be applied directly to examine such phenomenon, thereby negating the need for subsequent modeling. The proposed approach overcomes some limitations associated with traditional means of arriving at descriptive expressions as examined against structural and fire engineering problems. This approach offers an alternative method that is cognitive, instantaneous, and affordable.