Real-World Anomaly Detection by Using Digital Twin Systems and Weakly Supervised Learning

The continuously growing amount of monitored data in the Industry 4.0 context requires strong and reliable anomaly detection techniques. advancement Digital Twin technologies allows for realistic simulations complex machinery; therefore, it is ideally suited to generate synthetic datasets use approaches when compared actual measurement data. In this article, we present novel weakly supervised industrial settings. make a training dataset, which simulates normal operation machinery, along with small set labeled anomalous from real machinery. particular, introduce clustering-based approach, called cluster centers (CC), neural architecture based on Siamese Autoencoders (SAE), are tailored settings very few samples. performance proposed methods against various state-of-the-art algorithms an application real-world dataset facility monitoring system, by using multitude measures. Also, influence hyperparameters related feature extraction network investigated. We find that SAE-based solutions outperform robustly many different hyperparameter all