Physically-based models derive from first principles (e.g. physical laws) and rely on known variables and parameters. Because these have physical meaning, they also explain the underlying relationships of the system and are usually transportable from one system to another as a structural entity. They only require model parameters to be updated. Data-driven or regressive techniques involve data mining for modelling and one of the major drawbacks of this is that the functional form describing relationships between variables and the numerical parameters is not transportable to other physical systems as is the case with their classical physically-based counterparts. Aimed at striking a balance, Evolutionary Polynomial Regression (EPR) offers a way to model multi-utility data of asset deterioration in order to render model structures transportable across physical systems. EPR is a recently developed hybrid regression method providing symbolic expressions for models and works with formulae based on pseudo-polynomial expressions, usually in a multi-objective scenario where the best Pareto optimal models (parsimony versus accuracy) are selected from data in a single case study. This article discusses the improvement of EPR in dealing with multi-utility data (multi-case study) where it has been tried to achieve a general model structure for asset deterioration prediction across different water systems.

This content is only available as a PDF.