Enhanced Condition Monitoring of Low Voltage Motors in Harsh Environments with Ladybug Beetle Optimization and Context-Aware Graph Convolutional Networks

Low voltage motors are essential in many applications in contemporary industrial settings, as they power machines and guarantee efficient operations. However, these motors often operate in harsh environments where they are exposed to factors such as extreme temperatures, moisture, dust, and vibrations, which can lead to premature wear and failure. This paper proposes a hybrid method for low voltage motor condition monitoring in challenging industrial environments. The proposed method is the combined execution of Ladybug Beetle Optimization algorithm (LBOA) and Heterogeneous Context-Aware Graph Convolutional Network (HCAGCN). Hence it is named as LBOA-HCAGCN technique. Initially, the input data is collected from Triboelectric vibration sensors. Then, the collected data is fed into preprocessing utilizing Generalized Multi-kernel Maximum Correntropy Kalman Filter (GMKMCKF). This filter normalizes and cleans the data by effectively removing noise. Recursive Hilbert Transform (RHT) is used to extract time domain features such as mean, Root Mean Square (RMS), Peak-to-Peak Value, Skewness, and Kurtosis. Then the extracted features are given to HCAGCN for classify the motor health conditions such as normal, broken-bar, bowed-bar, bowed-rotor, faulty bearing, and voltage imbalance. In general, HCAGCN does not express adapting optimization strategies to determine optimal parameters to classify motor conditions. Hence, the LBOA is used to optimize the weight parameter of the HCAGCN which accurately classify motor conditions. The proposed LBOA-HCAGCN is implemented in MATLAB. Performance indicators such as Mean Absolute Error (MAE), Precision, and Accuracy were used to analyze the effectiveness of the proposed approach. Comparing the proposed LBOA-HCAGCN methodology to other existing techniques like Deep Residual Neural Network (DRNN), Convolutional Neural Network (CNN), and Recursive Neural Network (RNN), it achieves 25.8%, 26.4%, and 24.7% greater accuracy, 26.7%, 29.4%, and 21.7% precision.