Manoj Kumar Kumawat
Department of Mechanical Engineering, Malaviya National Institute of Technology Jaipur, Rajasthan, India
Devesh Kumar
Department of Mechanical Engineering, Malaviya National Institute of Technology Jaipur, Rajasthan, India
Shobha Rathore
Department of Mechanical Engineering, Malaviya National Institute of Technology Jaipur, Rajasthan, India
Sachin Kumar Mangla
Operations Management and Decision Making, Digital Circular Economy for Sustainable Development Goals (DCE-SDG), Jindal Global Business School, O P Jindal Global University, Haryana, India
Prof. R P Mohanty
Former President IIIE and AICTE Distinguished Professor, New Delhi, India
Abstract
The intelligent maintenance system designed for the Flexible Manufacturing System (FMS) aims to achieve high operational efficiency and reduced downtime in line with Industrial 4.0 objectives. Conventional maintenance methods enable the realization of these goals, but they fall short in addressing unexpected equipment failures. To enhance efficiency and decrease downtime, it is essential to move beyond traditional practices like run-to-failure and preventative maintenance. This approach implements an Exponential Weighted Moving Average (EWMA) control chart combined with machine learning for real-time fault identification within a hybrid innovative framework. The main goal of this study was to develop a deep learning model capable of accurately predicting the air pressure of the FMS system. Subsequently, the process was monitored over time using an EWMA control chart, which was constructed from the model’s residuals, allowing for real-time anomaly detection in the FMS process. Lastly, the study compared the long-short-term memory (LSTM) algorithm with other methods to determine the most suitable model for this research. A case study involving the FMS system with variable air pressure data was discussed to demonstrate the proposed framework. Utilising this framework, the FMS can pinpoint anomalies with 95% accuracy, ultimately leading to reduced downtime and improved system performance.
Keywords- Flexible manufacturing system, Deep learning, Long-short term memory, Exponential Weighted Moving Average control chart, Predictive Maintenance.