Machine learning techniques to improve die casting process

The production of die-castings includes several steps, each of which can lead to a defective casting if it is not properly carried out. Since it can be costly to repair deficient castings, it is important to avoid defects if possible. Therefore, when the production process is already completed, several defects can appear in castings. In this thesis, we introduce the problem of the quality aspects of a production line from an industry “2A” that produces aluminum machined parts for FCA (FIAT Chrysler Automotive) through an innovative high pressure die casting techniques. FCA is requiring from is supplier high-quality standards and 2A is involved in Industry 4.0 project to ensure these standards. We are retrieving data coming from the production machine, and other inputs to analyze which are the machine setting that can be involved in the malfunction of the machined parts using machine learning techniques. Nevertheless, these models would be used as a black box to predict the quality as a function of process parameters.