Health Monitoring Technology for Hydraulic Engineering Gates Based on Multi-Sensor Data Fusion

Abstract

An essential part of the hydraulic engineering systems, hydraulic gates regulate water flow in dams and other structures. To avoid structural problems, they must ensured to be consistent. Multi-sensor data fusion has become a key tool for health monitoring with advances in machine learning (ML) and sensor technologies. This study aims to provide a novel multi-sensor data fusion health monitoring system for hydraulic engineering gates. To provide more effective and precise real-time monitoring of hydraulic engineering gates, a unique Dynamic Satin Bowerbird search-driven Bidirectional Long Short-Term Memory (DSB-BiLSTM) model were proposed. The hydraulic monitoring dataset was collected from Kaggle website, comprising information and some attributes obtained from hydraulic engineering gates. To improve data quality, noise reduction, and data normalization are used. Time-series analysis techniques extract important information such as vibration patterns and pressure variations. The DSB-BiLSTM model leverages bidirectional LSTM to progress time-dependent data, and the Satin Bowerbird search optimizes sensor selection. The method shows better accuracy and efficiency in identifying potential faults compared to traditional models. When compared to four attributes such as pump, cooling, valve and accumulator with the metrics outcomes of 0.9901, 0.9925, 0.9950, and 0.9890 in accuracy, 0.9930, 0.9960, 0.9902, and 0.9889 in precision, 0.9920, 0.9960, 0.9911, and 0.9890 in recall, and 0.9915, 0.9955, 0.9905, and 0.9889 in F1-score, the DSB-BiLSTM model performs better than all other approaches that were examined. This suggests better overall hydraulic engineering gate health monitoring performance. The proposed DSB-BiLSTM model is a show potential solution for real-time health monitoring of hydraulic gates, provided that it enhances accuracy and reduces computational complexity.