Gravanis Georgios (Phd Candidate)

Thesis title: Fault Detection in industrial / production processes with Deep Learning methods
Supervisor: Diamantaras Konstantinos
Advisory Committee Members:
Simira Papadopoulou, Professor IHU
Michail Salampasis, Professor IHU

The scope of this Ph.D. thesis is to examine the possibility of using deep neural networks with architectures, such as those of Time Delay Neural Networks (TDNN) and Long Short-Term Memory (LSTM) for early fault detection in industrial and production processes.  With this thesis, deep learning architectures that will make the produced models capable of use in real production time, will be developed. Also, the use of appropriate metrics for proper result evaluation, as well as feature extraction methods for effective network training, will be investigated. Finally, unsupervised and reinforcement learning algorithms will be utilized for providing solutions in real-life large-scale applications.