Analysis and Anomaly Detection in DWLR Logs using Deep Learning

Project Code :TCMAPY2142

Objective

This project focuses on the analysis and anomaly detection in Digital Water Level Recorder (DWLR) log data using machine learning and deep learning techniques. Synthetic data is generated to simulate accurate water monitoring conditions, including normal behavior and different anomaly types such as spike, drift, flatline, and outlier. Five key input features—water level, temperature, rainfall, flow rate, and dissolved oxygen—are used for classification. Three models, Random Forest, Artificial Neural Network (ANN), and Long Short-Term Memory (LSTM), are implemented and evaluated to identify anomalies effectively. Random Forest captures feature-based patterns, ANN learns complex non-linear relationships, and LSTM handles temporal dependencies in sequential data. A web-based application is developed using Flask for backend processing and HTML, CSS, and JavaScript for the frontend, including user authentication and result display. The project demonstrates an automated, accurate, and scalable approach for DWLR log analysis and supports reliable water data monitoring.

Abstract

Digital Water Level Recorder (DWLR) systems generate continuous log data that reflects changes in water-related parameters. Analyzing this data is important for identifying unusual behavior that may indicate sensor issues or abnormal conditions. This project focuses on the analysis and anomaly detection in DWLR logs using deep learning and machine learning methods. Synthetic data is used to simulate different patterns and irregularities in water level records. The input features considered in this study include water level, temperature, rainfall, flow rate, and dissolved oxygen. The output is classified into five categories: normal condition, spike, drift, flatline, and outlier. Three algorithms are implemented for classification: Random Forest, Artificial Neural Network (ANN), and Long Short-Term Memory (LSTM). Each model is trained and evaluated to detect anomalies effectively based on the given features. A web-based system is developed using Flask as the backend framework and HTML, CSS, and JavaScript for the frontend. The system includes user authentication and a classification module to display prediction results. This project demonstrates how data-driven models can support accurate anomaly identification in DWLR logs and provides a structured platform for analysis and comparison of different algorithms.

Keywords: DWLR, Anomaly Detection, Synthetic Data, Random Forest, ANN, LSTM, Classification, Deep Learning, Water Data, Flask

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.