Quality prediction and control of manufacturing process is a particularly challenging task. This is due to the presence of stastistical uncertainties as well as difficulty to assess the influence of each process parameter on the mechanical properties and life of the component. In the Reliability Modelling and Optimization (RMO) Lab, we develop analytical probabilistic models, assessing the contribution of process parameter uncertainties to the experimentally observed variation in product quality. Our research is extended to adopt a probabilistic approach in an Industry 4.0 framework as well as to assess the interaction between human and machine as an integral part of Industry 5.0. The research effort consists of combining data from mechanical simulation performed on digital twins and data from physical experiments, to train machine learning models. The aim is for these models is to be control the manufacturing process as well as predict the resulting component quality and longetivity.

Defects in additive manufacturing (AM) occur due to a combination of material properties, process parameters, and environmental conditions. These defects can arise from improper laser power, scanning speed, or layer thickness. Common types of defects include porosity, caused by trapped gas or lack of fusion (LOF); residual stresses, which develop due to rapid cooling and can cause warping or cracking; surface roughness, resulting from uneven powder distribution or inadequate layer bonding; and microstructural inhomogeneities, where variations in cooling rates lead to inconsistent mechanical properties. In our research we are particularly focusing on the influence of differenct post-processing techniques, including preloading and Hot Isostatic Pressing (HIP), to help mitigate these defects and improve the quality of AM parts.