Logo: to the web site of Uppsala University

Carbon-fibre-reinforced polymers are in great demand in aerospace, civil engineering, and many other fields. During its manufacturing process, they undergo a process called curing. This process is carried out by using an effective microwave heating system. The system uses Beambook, a collection of microwave source settings and a Beambook algorithm to control the temperature and their resulting heating pattern.

This thesis mainly focuses on the pipeline that maps microwave heating system parameters with carbon fiber’s heating patterns for controlling the heat within the microwave heating system. The first part of the pipeline includes analyzing the heating patterns and reducing their dimensions using principal component analysis (PCA). Following PCA, the deep neural network technique is implemented to predict the heat patterns for the input system parameters. Thus investigating how a black-box method would work on our dataset by comparing their result with a white-box model.

The experiments performed with different machine learning techniques andthe neural network have proven that the pipeline built works for the Beambook data and is suitable for future analysis.