Porous Metal Properties Analysis: A Machine Learning Approach

In the past few decades, computer-aided techniques (i.e., numerical simulation) have complemented the research and development process (R&D) in material sciences. This approach is usually paired to experimental testing. Yet, both techniques have shown cost-efficiency disadvantages and are time consuming. Optimization algorithms like the ones used in machine learning have proven to be an alternative tool when dealing with lots of data and finding a solution. While the use of machine learning is a well-established technique in other research fields, its application in material science is relatively new. Material informatics provide a new approach to analyse materials such as porous metals by employing previous data sets. This article aims to study reliability to predict permeability and Forchheimers coefficient of lost carbonate sintering open-cell porous metal. The key features selected as predictors are porosity, pore size, and coordination number. A comparison among multiple linear regression, polynomial regression, random forest regressor and artificial neural network is revised