Advanced Predictive Maintenance of Phoenix Contact Relays: A Digital Twin and Machine Learning Approach

Predictive maintenance (PdM) leveraging digital twin technology offers a transformative approach to maintaining industrial equipment, minimizing downtime, and reducing maintenance costs. We focus on the predictive maintenance of Phoenix Contact relays using a comprehensive dataset of relay sensor readings (the Phoenix Contact Relay dataset). By implementing advanced machine learning models, including a bagging and boosting model, different types of LSTM, and an Autoencoder with a fully connected neural network, we aim to predict relay failures accurately. Our approach involves data preprocessing, feature extraction, and feature augmentation to create the dataset that improves the predictive capabilities of these models. To provide a dynamic, real-time representation of the relay. We use a digital twin framework that continuously simulates its behavior and predicts potential failures. This integration tries to improve the accuracy but also provides insights into the underlying causes of relay degradation. In industrial applications, we try to demonstrate that using digital twin technology alongside machine learning and deep learning models can result in highly efficient and reliable predictive maintenance.