Optimizing Predictive Maintenance in Industrial IoT Cloud Using Dragonfly Algorithm

Cloud computing enables users to access and utilize various computing resources over the internet, including servers, storage, databases, and analytics. This paradigm offers flexibility, scalability, and cost efficiency, making it a critical technology for numerous applications. In the realm of the Internet of Things (IoT), cloud computing provides a scalable and flexible infrastructure for managing the vast amount of data generated by IoT devices. Specifically, in Industrial IoT applications (IIoT), predictive maintenance has become a key focus, leveraging advanced technologies to forecast equipment failures and minimize downtime. However, achieving high accuracy in fault prediction remains a challenge. To address this, we propose a novel approach called Brokenstick Regression-based multiobjective dragonfly predictive optimization (BR-MDPO). This method aims to optimize predictive maintenance with enhanced accuracy and execution time (ET). The process begins with IoT devices collecting data, such as vibration, temperature, speed, torque, and operational hours, from industrial machinery. This data is then sent to centralized cloud data centers for predictive analysis. The BR-MDPO technique utilizes the Multiobjective Dragonfly Optimization algorithm, a metaheuristic inspired by the natural behavior of dragonflies, to solve multiobjective optimization problems. Brokenstick regression analyzes the data to optimize various objective functions. The technique identifies potential failures, facilitating proactive maintenance and informed decision-making to ensure continuous productivity. The proposed method shows a significant improvement in accuracy, precision, and recall by 7%, 5%, and 6%, respectively. The observed results reveal a 6%, 4%, and 5% enhancement in the accuracy, precision, and recall. Furthermore, the proposed technique realizes a substantial reduction in error rate by 68%, 15%, and 13% reduction in ET as well as latency compared to conventional methods.