Description

The thesis will study the development of Machine Learning Techniques to handle process data derived from a Portuguese Plastic Industry Plant, aligned with the recent trends of Digitalization and Industry 4.0. The study embraces a data collection stage, being proposed a low-cost technological solution that can be used in different brand machines operating in the plant, avoiding high investments related to commercial data collection solutions and software licenses commonly requested. Afterwards, with the data being stored and monitored, a detailed analysis of different features derived from different machines and related to the production process (eg. injection time, plasticization time, injection pressure or the cushion), becomes possible.
It is our purpose to use Machine Learning techniques contributing to a feature selection/extraction stage, to enhance maintenance actions (such as a prior identification of conditions that may lead to rejected parts or to avoid interruption periods due to machines malfunctioning conditions), or even enabling an online adjustment of machine parameters yielding to high-quality final products while preserving a competitive production time.