A survey on anomaly detection for technical systems using LSTM networks

Anomalies represent deviations from the intended system operation and can lead to decreased efficiency as well partial or complete failure. As causes of anomalies are often unknown due complex dynamics, efficient anomaly detection is necessary. Conventional approaches rely on statistical time-invariant methods that fail address dynamic nature anomalies. With advances in artificial intelligence increasing importance for prevention various domains, neural network enable more types while considering temporal contextual characteristics. In this article, a survey state-of-the-art using deep especially long short-term memory networks conducted. The investigated evaluated based application scenario, data further metrics. To highlight potential upcoming techniques, graph-based transfer learning also included survey, enabling analysis heterogeneous compensating its shortage improving handling processes.