Predictive maintenance plays an important role in reducing long-term maintenance costs, unplanned downtime, and improving the lifetime of industrial machines. A common trait of machines is that they produce heat while working, resulting in a temperature pattern. Temperature can be a key parameter for monitoring the condition of machines, further aiding the diagnostics of problems. This paper presents an Internet of Things (IoT) system that monitors and detects thermal anomalies in industrial machines using deep neural networks (DNNs). The proposed system enables the DNN to run and make predictions inside a microcontroller, reducing the amount of data that needs to be transmitted to any external server. Furthermore, this system uses a platform that centralizes multiple sensors with the option of communicating with a server that runs two additional neural networks that are specialized in highlighting zones of interest in the thermal image and monitoring the temperature behavior over time. The system was tested in a laboratory and two industrial environments. Overall, the system performed well and can detect machine anomalies while also drastically reducing the amount of data needed to be transmitted. The system also presented high adaptability to different environments.
CITATION STYLE
Oliveira, V. M., & Moreira, A. H. J. (2022). Edge AI System Using a Thermal Camera for Industrial Anomaly Detection. In Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST (Vol. 442 LNICST, pp. 172–187). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-06371-8_12
Mendeley helps you to discover research relevant for your work.