Intelligent Rotating Machine Fault Detector

What it does

This system is an intelligent fault detection solution that identifies mechanical issues like imbalance, misalignment, looseness, and bearing defects in rotating machinery using vibration analysis and AI, and sends alerts through a web application.

Your inspiration

The inspiration came from real-world challenges in manufacturing environments, where even minor faults in machinery can lead to unexpected breakdowns and costly downtime. Traditional fault diagnosis methods often rely on manual inspections or reactive maintenance, which can be inefficient and unreliable. This motivated me to develop an intelligent system capable of continuously monitoring machine conditions and detecting faults at an early stage. By enabling predictive maintenance, the system aims to bridge the gap between conventional mechanical diagnostics and modern, technology-driven monitoring solutions.

How it works

The Intelligent Rotating Machine Fault Detector works by acquiring real-time vibration data from a mechanical test rig using NI DAQ hardware and piezoelectric accelerometer sensors. The signals are processed through a digital bandpass filter to isolate fault-related frequency components and further refined using wavelet denoising to reduce background noise. From the filtered signal, time and frequency domain features such as RMS, mean, skewness, kurtosis, and crest factor are extracted. These features are then input into a machine learning model, specifically a Support Vector Machine (SVM), which has been trained on labelled datasets representing various fault types. Once the fault is classified, the system displays the result through a user-friendly MATLAB App GUI and notifies users via a connected web-based dashboard. The system can accurately detect multiple fault conditions and provides real-time insights into the health status of rotating machinery.

Design process

The project began with identifying the need for a smarter approach to machine fault detection. I researched common mechanical faults in rotating equipment and studied how vibration signals can reflect different fault conditions. The initial concept was to develop a system that could classify faults using vibration data and AI. I designed a basic test rig setup with controlled fault simulations such as imbalance, misalignment, looseness, and bearing defects. In the first prototype, I used NI DAQ hardware and accelerometer sensors to acquire vibration signals. The data was processed in MATLAB using a bandpass filter, but initial results were noisy and inconsistent. I then implemented wavelet denoising to improve signal clarity. Feature extraction focused on key time and frequency domain metrics like RMS and kurtosis. Various machine learning models were tested, and SVM was chosen for its accuracy. The MATLAB App GUI was developed to allow user interaction, followed by integrating a web-based dashboard. Each iteration improved usability, accuracy, and system response. Currently, the system is compact, accurate, and capable of real-time fault classification, ready for further industrial testing and IIoT integration.

How it is different

Unlike conventional condition monitoring systems that rely on manual inspections or fixed threshold alarms, the Intelligent Rotating Machine Fault Detector uses artificial intelligence to classify multiple fault types directly from raw vibration data. Traditional systems often miss early-stage faults or detect issues only after significant performance loss. This system integrates real-time data acquisition, signal preprocessing, and machine learning into a compact and scalable solution. Using NI DAQ hardware, vibration signals are filtered and denoised before extracting key time-domain features, which are then classified using a trained model. The system provides instant fault feedback through a MATLAB App GUI and notifies users via a web-based dashboard. Its intelligent, modular design offers higher accuracy, usability, and flexibility, making it suitable for both academic and industrial environments.

Future plans

Moving forward, I plan to expand the system to detect more advanced and compound mechanical faults, such as complex bearing and gear defects. I aim to validate its performance in real industrial environments to ensure accuracy and reliability. The hardware will be miniaturised using embedded systems and IIoT sensors with wireless connectivity, making it compact and scalable. I also plan to integrate a cloud-based dashboard for remote monitoring and intelligent data analytics. Ultimately, my goal is to develop a cost-effective predictive maintenance solution tailored for SMEs.

Awards