Remaining Useful Life Prediction of High-power DC Contactors Based on BiLSTM

DC contactors are key electrical components widely used in aerospace, electric vehicles, photovoltaic energy storage, and other fields. Predicting the Remaining Useful Life (RUL) of DC contactors can enhance the reliability and safety of the systems. Currently, most RUL prediction models focus on AC contactors and are limited to capturing features during the releasing process. However, when predicting the RUL of high-power DC contactors, challenges arise due to large parameter ranges and difficulties in data acquisition. In this paper, a RUL prediction model based on monitoring the features during the closing process for 270V/ 200A high-power DC contactors is proposed, using a Bidirectional Long Short-Term Memory (BiLSTM) network. After conducting sequential experiments on lifespan testing and parameter acquisition, feature extraction and selection, dataset partitioning and preprocessing, and model training and optimization, the performance of the proposed model is evaluated and analyzed using only closing process features, only releasing process features, and a combination of both. The results demonstrate that the RUL prediction model based on closing process features not only exhibits high accuracy but also reduces the difficulty of parameter acquisition. In addition, the wear rate of the contactor exhibits a three-stage 'fast-slow-fast' variation. By detecting the slope of the RUL curve and setting thresholds, it is possible to effectively implement proactive maintenance in practical engineering applications. It offers a practical solution contributing to the advancement of RUL prediction methodologies of high-power DC contactors.