Research On Leakage Detection Algorithms Of Water Supply Pipeline Based On One-dimensional Convolutional Neural Network

Water plays a vital role in people’s production and life.However,due to the influence of natural or non-natural factors such as changes in the surrounding environment and long service life of water supply pipelines and man-made destruction,pipeline damage and corrosion problems emerge in endlessly.If the pipeline can not be detected in time,serious economic and property losses and potential safety hazards will be caused,as well as water resources.A lot of waste.Traditional leak detection methods are time-consuming and labor-intensive,and rely heavily on human experience.With the development of artificial intelligence,machine learning and deep learning are gradually applied to leak detection.Previous researchers use machine learning algorithms such as SVM and artificial neural network to detect leakage signals.Such methods need to extract features manually and carry out a large number of experiments.Different feature combinations have a great impact on accuracy.In order to solve this problem,a one-dimensional convolution neural network is proposed to classify the leakage data.The feature is extracted automatically by the algorithm,and then the extracted feature is connected to each other through the full connection layer.Finally,the predicted value is output through the sigmoid activation function.Because in-depth learning relies heavily on the amount of data,and the actual data collected can not meet the needs of model training,this paper proposes a data enhancement technology of offset sampling for one-dimensional time series signals.In order to enhance the classification accuracy of noisy signals,a 1D-CNN classification algorithm model based on BN is proposed.The optimal number of hidden layers and activation functions of convolutional neural network are selected through comparative experiments,and the highest classification accuracy reaches 100%.The classification accuracy of 1D-CNN based on BN under different SNR conditions is analyzed and verified.