Publication: Development of real-time machine learning monitoring web server for pump diagnostics
Loading...
Date
2023-07-14
Authors
Mohamad Faizal bin Rashidi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This research project focuses on the development of a real-time machine learning monitoring web server for pump diagnostics. To enable early and accurate diagnosis of a running centrifgal pump, the project utilizes machine learning
algorithms to analyze vibration and acoustic data. By collecting data from sensors and leveraging MATLAB and Microsoft Azure Machine Learning, an implementable machine learning model is developed. The main objective of the project is to construct a web server-based system that allows engineers to rapidly identify and diagnose pump problems through real-time machine learning implementation. The key features of the project include the implementation of a machine learning algorithm for detecting obstructions in centrifugal pumps by analyzing vibration and acoustic signals in the time and frequency domainas. A web server with a user interface is developed to connect to the machine learning model, providing real-time diagnostic status reports accessible from any device with an internet connection. Additionally, a system comprising a Raspberry Pi and the MPU6050 sensor is implemented to transmit telemetry data to an Azure IoT Central application, enabling real-time visualization. The project scope involves data collection from an experimental rig in, which incorporates sensors for vibration, acoustic signals, and rotational speed measurements. The collected data is preprocessed and transformed into the frequency domain using fast Fourier transform (FFT). Feature extraction is performed in both the time and frequency domains, and a supervised machine learning algorithm, specifically the Support Vector Machine, is trained using MATLAB.The trained machine learning model is exported for the development of a user interface using Visual Code Studio. Additionally, the MPU6050 sensor connected to a Raspberry Pi facilitates the transmission of telemetry data to an Azure IoT Central application for real-time visualization. In conclusion, this research project successfully develops a real-time machine learning monitoring web server for pump diagnostics. By providing
engineers with real-time diagnostic information and prognostic insights, the system contributes to the field of pump diagnostics by enhancing the efficiency of maintenance planning and reducing equipment failure rates. The project's main
achievements lie in the development of an accurate machine learning model for pump fault detection, the construction of a user-friendly web server interface, and the integration of a Raspberry Pi-based system for real-time data transmission.