Evaluating Machine Learning Algorithms for Anomaly Detection in Industrial Engines on Edge Devices
Year: 2025
Type: conference-paper
Abstract: Industrial automation and the Internet of Things (IoT) are rapidly evolving, highlighting the need for efficient anomaly detection systems to prevent failures and reduce maintenance and operational costs. This study evaluates the feasibility of running machine learning-based anomaly detection models on edge devices like the Raspberry Pi. Using real sensor data from industrial engines with induced faults, we tested three unsupervised algorithms: LOF, OCSVM, and IF. Metrics such as sensitivity, specificity, inference time, and resource usage were analyzed under different loads. The results show that accurate and low-latency detection is possible on resource-constrained devices, supporting the development of cost-effective and autonomous monitoring systems for industrial environments.
Authors: Lucas Araújo, Sérgio Chevtchenko, Danilo Araújo, Ermeson Andrade
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4416740526