Master's thesis project focusing on unsupervised anomaly detection in process data using deep learning. This project was published in MDPI Journal: Sensors. This project presents a novel approach for anomaly detection in industrial processes. The system solely relies on unlabeled data and employs a 1D-convolutional neural network-based deep autoencoder architecture. As a core novelty, we split the autoencoder latent space in discriminative and reconstructive latent features and introduce an auxiliary loss based on k-means clustering for the discriminatory latent variables. We employ a Top-K clustering objective for separating the latent space, selecting the most discriminative features from the latent space. We use the approach to the benchmark Tennessee Eastman data set to prove its applicability. We provide different ablation studies and analyze the method concerning various downstream tasks, including anomaly detection, binary and multi-class classification. The obtained results show the potential of the approach to improve downstream tasks compared to standard autoencoder architectures.

Performed exploration and feature engineering on Tennessee Eastman Process data Trained a convolutional autoencoder with K-Means clustering algorithm incorporated with the network criterion Extracted the encoder to train a new model in a semi-supervised manner Developed different algorithms for binary and multi-class classification Tuned the hyperparameters and visualized the results using Tensorboard