Life Time Prediction of an Electromagnet Relay using Clustering based Principal Component Analysis with Hybrid Deep Learning Model

Abstract

The estimation of the electromagnet relay remaining useful life is highly crucial to maintain reliability and avoid unscheduled breakdowns in various applications. The objective of this research work will be to design a model with much higher precision and efficiency utilizing PCA coupled with a hybrid deep learning architecture of Bi-LSTM along with Bi-GRU. The C-MAPSS dataset was of reduced dimensionality, since PCA has been applied to eliminate data redundancy while retaining crucial characteristics, and then K-means clustering is applied to classify the data; afterwards, the Bi-LSTM and Bi-GRU models are implemented for RUL relay prediction. The proposed method in comparison with typical deep learning models has a Mean Absolute Error of 0.021 and an R² of 0.996. Results developed reflect how the model can produce some very powerful prediction, however; what it really shows is great potential for this approach with respect to predictive maintenance of electromagnet relays. PCA may well amalgamate with Bi-LSTM and Bi-GRU models to achieve great scalability according to the maintenance engineering, which offers practical applications in improving the lifetime of the electromagnet relays.

Keywords

• Electromagnet Relay
• Remaining Useful Life
• Bidirectional Long Short Memory with Bidirectional Gated Recurrent Unit
• Principal Component Analysis