Leveraging Recurrent Neural Networks in Virtual Reality-based Digital Twin Models for Enhanced Predictive Maintenance of Industrial Machines

Digital twin technology has given rise to smart manufacturing in Industry 4.0. Especially with the help of virtual reality, digital twin technology aims to provide an immersive experience by integrating the physical and cyberspace worlds synchronously in real time. One of its key features is the ability to perform predictive maintenance, to help prevent a product from any possible faults in the future. The existing literature has explored the traditional predictive maintenance strategies for forecasting the future state of machines such as simple regression and ARIMA models. However, the potential of machine learning, especially recurrent neural networks (RNNs) is not yet fully investigated for predictive maintenance in digital twin models. The complex nature of industrial operations gives rise to non-linearities in its modeling which can be addressed with the help of neural networks. Therefore, this research aims to investigate the potential of RNNs in the predictive maintenance of an industrial machine use case by performing comparative analysis with simple regression and ARIMA Models. It also highlights the significant improvement RNNs make over the aforementioned strategies. This research also proposes an extension of the existing virtual reality-based digital twin architecture to incorporate automated predictive maintenance of the machine. Moreover, the proposed digital twin architecture acts as a basis for automated predictive maintenance of any product.