ANOMALY DETECTION BASED ON DEEP AUTOENCODER FOR MANUFACTURING PROCESSES

I have developed a project called AnomalyDetector, a software written in Python using Tensorflow 2.0 for the detection of anomalies in industrial machinery using deep learning techniques. Anomalies are very rare, so it is necessary for "AnomalyDetector" to be able to recognize them using historical data relating to a few anomalous instances. The high-level architecture is composed of 4 main components: 3 of these implement the functions required by the 3 phases of analysis (features extraction, anomaly score calculation and anomaly discriminator) while a fourth component orchestrates the first 3, for each case study