Implementing Automatic Generation of Primary Equipment Status Awareness and Maintenance Priority Using a Hybrid GCN-RNN Architecture

To address issues in industrial production, such as low monitoring accuracy of primary equipment, poor adaptability to complex electromagnetic testing and industrial operating environments, and unreasonable maintenance priority allocation, this paper proposes a framework integrating GCN and RNN for intelligent equipment condition awareness and automatic maintenance priority generation. First, multi-source monitoring data, including vibration, temperature, pressure, current, and operating load, from primary equipment are cleaned, standardized, and subjected to feature engineering preprocessing to construct an operational status dataset. Second, a GCN module is designed to extract spatial correlation features among monitoring points, while an RNN subnetwork based on LSTM captures temporal evolution patterns of equipment status. This establishes a hybrid model integrating dual spatial-temporal features, enabling precise identification of equipment health levels, including normal, mild anomaly, moderate anomaly, and severe anomaly. Then, based on the equipment status perception results, an evaluation system for maintenance priority is established by integrating indicators such as equipment importance, fault impact scope, and repair costs. The Entropy Weighting-TOPSIS method is employed to achieve quantitative prioritization, providing technical support for maintenance decision-making in industrial systems involving electromagnetic measurement platforms, antenna testing equipment, and radio-frequency monitoring devices.