A novel parameter-optimised LSTM model fused by convolutional block attention module for water pipeline leakage detection

The detection of water supply pipeline leakage is crucial to maintaining the seamless operation of the water supply system. While deep learning has shown promise in pipeline leakage detection (PLD), the complexity and uncertainty of pipeline data present challenges for accurate leak identification. To address this, a novel Parameter-optimised Long Short-Term Memory (PLSTM) model with Convolutional Block Attention Module (CBAM) is proposed. The article employs an improved Quadratic Interpolation Optimization (IQIO) algorithm to identify the optimal state of the model, thereby enhancing its learning capability and convergence rate. The CBAM enhances the capability of models to identify specific features in a more effective manner. The experimental results illustrate the efficacy of the proposed approach, with a recognition accuracy rate of 98.33% achieved. This methodology offers precise and reliable pipeline leakage detection in transportation systems.