Anomaly Detection with a LSTM Autoencoder Using InfluxDB

Abstract

In the manufacturing industry, anomalies are an unfortunate but inevitable reality. If left unaddressed, they can lead to costly production defects and halted production lines. However, with the rise of Industry 4.0, many industrial machines are now equipped with sensors that can be used to detect anomalous behaviors, allowing for early identification and prevention of defects. Therefore, this study presents a solution using a Long Short-Term Memory (LSTM) autoencoder to detect abnormal behavior in an industrial machine temperature sensor dataset. The algorithm is compared with conventional methods, further demonstrating its capabilities in anomaly detection. Additionally, an implementation architecture is proposed using InfluxDB and Telegraf software, providing a simulated real-world application of the proposed solution.