Machine Learning Regressors in Forecasting Mechanical Properties in Advanced Manufacturing Processes

This paper presents the advances in forecasting mechanical properties by applying machine learning regressors in advanced manufacturing processes. Computational intelligence is advancing by leaps and bounds in artificial intelligence and data science, which is how data is currently considered the new gold. In addition to advances in information technology, manufacturing processes have also evolved. In that context, additive manufacturing has several advantages over conventional manufacturing processes. On the one hand, it offers the possibility of processing any material: metals, ceramics, polymers, and composite materials. On the other hand, these technologies make it possible to optimize the use of the material, reducing its impact on the environment. Due to its versatility, it is possible to improve design to save material and reduce assembly complexity. This work presents a series of cases of applying machine learning regressors in forecasting mechanical properties in advanced manufacturing processes, from friction welding, laser welding, selective laser melting, wire arc additive manufacturing, and fused filament fabrication. The results show that the gradient-boosting regressors have greater precision in predicting several mechanical properties than conventional artificial neural networks. These algorithms are more effective than using the traditional design of experiments, reducing costs and allowing information about the physical phenomenon involved in each manufacturing process.