An edge computing based anomaly detection method in IoT industrial sustainability
Xiang Yu,Xianfei Yang,Qingji Tan,Chun Shan,Zhihan Lv
DOI: https://doi.org/10.1016/j.asoc.2022.109486
IF: 8.7
2022-10-01
Applied Soft Computing
Abstract:In recent years, the evolving Internet of Things (IoT) technology has been widely used in various industrial scenarios, whereby massive sensor data involving both normal data and anomalous data are generated. However, the anomalies involved might have a great impact on the industrial sustainability. Actually, in most scenarios, it is of vital importance to detect the anomalies and processed accordingly in real time. In previous work, some methods, which analyze the collected industrial sensor data on cloud computing center to detect anomalies, casually increase both the transmission pressure of bandwidth and the computing pressure on cloud center. By contrast, some other methods based on edge computing detect anomalies on edge nodes, but they only take the anomalies of sensor data from single-source time series into account and ignore the correlation anomalies between the data from multi-source time series. In this paper, we first present an anomaly detection model in distributed edge computing. Then, the Edge Computing based Anomaly Detection Algorithm (ECADA), which can detect the anomalies from both single-source time series or multi-source time series is proposed. Finally, we conduct a series comparison experiments in order to demonstrate the effectiveness of the algorithm proposed in the paper. And the experimental results demonstrate that both the anomalies from single-source time series and the anomalies from multi-source time series can be accurately detected by the proposed algorithm. Specifically, it performs more efficient and effective when there exist the anomalies of new type.
computer science, artificial intelligence, interdisciplinary applications