Design and Development of Data Driven Intelligent Predictive Maintenance for Predictive Maintenance

Abstract

Industrialists generally believe that maintenance is a necessary evil. A significant portion of all manufacturing systems' overall operating costs are related to maintenance. Industrial manufacturing businesses are severely impacted by the loss of production time and product quality brought on by an ineffective maintenance strategy. With the introduction of Industry 4.0, IoT-based data-driven automated remote-controlled operations and digitalised maintenance methods can be implemented by utilising cyber-physical systems methodologies. Maintenance scheduling and shop floor job allocation can be made simple using an IoT-based intelligent decision support system for machinery health management. One reliable way to maintain the health of machines and the quality of products is through predictive maintenance. Predictive maintenance uses machinery condition monitoring data, such as component vibrations, temperature, acoustic emissions, etc., to gain insights into the actual operating condition of the manufacturing system rather than depending on in-plant average life statistics or industrial field failure data. Real-time equipment status monitoring data collection and data acquisition from any location are made possible by developments in sensor technology and Internet of Things connectivity. Predictive maintenance of machine tools is becoming increasingly popular in the industrial sector as a way to increase production rates and improve tolerance of machined products. However, industrialists are discouraged from using predictive maintenance due to the high installation costs of the additional instruments and the complexity of computational tools. AI algorithms and other data-driven prognostics methods require little technical understanding of how machinery works and how failures occur. A promising AI computational approach for RUL estimation and machinery health prognostics is deep learning. However, there are numerous difficulties in putting deep learning algorithms for equipment health prognostics into practice.