Abstract
Predictive maintenance has emerged as a transformative solution in industrial operations, reducing unexpected machine failures and minimizing downtime. This study explores the application of machine learning in vibration analysis to detect imbalances in pump-motor units, a critical issue that accelerates equipment wear and increases maintenance costs. Traditional maintenance methods, such as scheduled inspections, often fail to identify faults early, leading to significant operational disruptions, particularly in utility, oil and gas sectors. By leveraging real-time vibration data and advanced machine learning techniques, this research aims to develop an efficient imbalance detection model that enhances maintenance efficiency and reliability. The study employs a structured methodology based on the CRISP-DM framework, covering data collection, preprocessing, modeling, and evaluation. The dataset includes vibration readings from multiple points on pump-motor units. Extensive preprocessing techniques were applied, including handling missing values, detecting outliers using Mahalanobis distance, and employing Principal Component Analysis (PCA) for dimensionality reduction. Feature engineering and selection methods, such as decision trees and statistical tests, were used to optimize the dataset for model training. Five machine learning algorithms—Logistic Regression, Linear Support Vector Machine (LSVM), Neural Networks, Decision Trees, and Random Forest—were tested across multiple scenarios to identify the most effective model. The dataset was split into three conditions: full dataset, dataset without outliers, and dataset with only outliers. The models were evaluated based on accuracy, recall, precision, and AUC (Area Under the Curve). Among the tested models, the Random Forest algorithm demonstrated the highest accuracy (99.3%) and recall (97.6%), making it the most reliable choice for detecting imbalances in pump-motor units. The results confirm that machine learning significantly improves imbalance detection, providing a proactive maintenance approach that reduces unexpected downtime, optimizes maintenance schedules, and extends equipment lifespan. The study also highlights the importance of data balancing techniques to address class imbalances and prevent bias in predictive models. Through advanced feature selection and sensitivity analysis, the most influential variables contributing to imbalance detection were identified, ensuring model interpretability and efficiency. The findings of this research contribute to the field of predictive maintenance by demonstrating the effectiveness of machine learning in real-world industrial applications. The integration of IoT-based vibration monitoring with machine learning can revolutionize equipment maintenance strategies, offering cost-effective and data-driven decision-making tools. Future work could explore the application of deep learning techniques and real-time deployment of the model for continuous monitoring in industrial settings.
Library of Congress Subject Headings
Pumping machinery--Performance--Forecasting--Data processing; Plant maintenance--Management--Data processing; Machinery--Maintenance and repair; Vibration--Testing; Reliability (Engineering); Machine learning
Publication Date
2-27-2025
Document Type
Thesis
Student Type
Graduate
Degree Name
Professional Studies (MS)
Department, Program, or Center
Graduate Programs & Research
Advisor
Sanjay Modak
Advisor/Committee Member
Hammou Messatfa
Recommended Citation
Albuflaseh, Abdulrahman, "Machinery Imbalance Failure Prediction Based on Pump – Motor Vibration Readings" (2025). Thesis. Rochester Institute of Technology. Accessed from
https://repository.rit.edu/theses/12056
Campus
RIT Dubai
Plan Codes
PROFST-MS
