NTPC Transactions on Energy Research

In modern industrial automation, DCS (Distributed control System) and PLC (Programmable logic Controller) are widely used for controlling processes and ensuring the safe operation of complex systems. The Distributed Processing Unit (DPU), a very important component in both DCS and PLC architectures, is responsible for real-time data acquisition and control. However, due to its continuous operation in challenging environments, DPUs are susceptible to failures, leading to unplanned downtimes and substantial financial losses. This paper proposes a predictive maintenance framework using Artificial Neural Networks (ANN) to predict DPU failures by monitoring key operational parameters. The model got trained & validated using the “Application failure prediction” dataset from Kaggle, which includes sensor readings and other self-diagnostic parameters and failure logs that align well with the operational characteristics of DPUs. The ANN model demonstrated an accuracy of 92.8% on the validation dataset, providing a reliable solution for early failure prediction. By implementing this ANN-based predictive maintenance framework, industries can proactively predict and address DPU failures, reducing unplanned downtime and minimizing maintenance costs, thereby enhancing overall operational efficiency.