Development of Leakage Detection Model and Its Application for Water Distribution Networks Using RNN-LSTM

With the advent of 4th Industrial Revolution, advanced measurement infrastructure and utilization technologies are being noticeably introduced into water supply system to store utilize data. From this perspective, leak detection technology in networks is becoming increasingly vital sustainable resource management clean worldwide. In particular, leakage buried pipelines rated as a very challenging research topic given current level technology. However, underground Therefore, data-driven model was developed through study using deep learning based on inflow meter Multiple threshold-based models were applied reduce RNN-LSTM (Recurrent Neural Networks–Long Short-Term Memory models) false prediction range, which programmed conjunction with Python language Google Colaboratory (a big data analysis tool). The consists flow pattern shape extraction, RNN-LSTM-based prediction, threshold setting modules. actual accident data, followed by performance evaluation. As result, recognized at most points immediately after accident. evaluated Confusion matrix showed more than 90% accuracy all except singularities. can be used critical software proactively identify various present smart introduced. addition, highly scalable it consider operational situations expert system, also efficiently reflect results pipe network across different scenarios.