The Industrial Internet of Things (IIoT) has the potential to boost the growth of industrial intelligence, increase production effectiveness, and lower manufacturing costs. The automatic monitoring and identification of anomalous events, changes, and drifts on the acquired data constitutes one of the Industrial IoT's primary objectives. All methods for identifying data patterns that differ from expected behaviour are categorised as anomaly detection methods. Accurately and promptly detecting anomalies is becoming more and more crucial since Industrial IoT device failures have a significant impact on the production of industrial goods. Also, anomalies which are needed to be identified could be used for better data analysis. So to address these problems, in this paper work is carried out using unsupervised learning with clustering approach in which sensors data from a prototype embedded system are used for findings anomaly in the data and the concept is to recognise clusters for interpretation of sensed data under different working circumstances to evaluate if new inspections fall in to any of these clusters. The approach is to find faulty or normal working conditions based on the anomaly findings methods applied to sensor data for analysis. There are three methods used in this proposed approach include fuzzy c-means-k-means clustering, and density-based clustering. The complete approach is found to be favourable for anomaly finding and situation control using cluster-based methods for industry related equipment’s systems rely on sensor data. The experimental results indicates that proposed method is accurate and effective based on sensor for different operating conditions of industrial scenario and assist in predictive maintenance.

Unsupervised learning IIoT industrial scenario predictive maintenance