Data-Driven Predictive Maintenance Strategies for Light Rail Vehicles : Applying Machine Learning and IoT Technologies to Enhance Operational Efficiency and Reliability
Savolainen, Antti (2024)
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024060219828
https://urn.fi/URN:NBN:fi:amk-2024060219828
Tiivistelmä
The evolution of electric transportation as a pivotal solution to climate change has seen significant growth in the past decades. The shift towards battery-powered public transit systems, notably the investment in electric LRVs (Light Rail Vehicles) like the Škoda ForCity Smart Artic X34 tram, underscores the urgency and commitment to sustainable urban mobility.
This thesis explored the integration of predictive maintenance and machine learning techniques to enhance the reliability and efficiency of light rail vehicle (LRV) operations, with a specific focus on lead-acid batteries. Advanced data acquisition technologies and machine learning algorithms, particularly Gaussian Process Regression (GPR), were utilized to develop robust predictive models for battery health monitoring and maintenance optimization.
The research employed a comprehensive case study approach, focusing on the Škoda ForCity Smart Artic X34 tram. Key findings demonstrated a moderate positive correlation between battery temperature and voltage, highlighting temperature's significant influence on battery performance. The key_value metric, derived from voltage, temperature, and duration data, was effective in assessing battery health, providing a unit-specific measure of battery response to load changes.
The GPR model, configured with a Rational Quadratic kernel and a White Kernel, achieved high predictive accuracy with a mean squared error (MSE) of 0.00114, accurately predicting battery capacity changes. These predictive capabilities enabled proactive maintenance scheduling, reducing unplanned downtime and extending battery life, thus improving operational efficiency and cost-effectiveness.
The thesis provided valuable insights into the practical applications of IoT technologies and machine learning in predictive maintenance, offering substantial benefits in terms of operational reliability, cost efficiency, and sustainability for urban transport systems.
Recommendations for future implementations included enhancing data integration, exploring advanced machine learning techniques, developing real-time monitoring systems, and conducting comparative analyses of different battery types. Economic impact assessments and sustainability evaluations were also suggested to further optimize predictive maintenance strategies.
This thesis explored the integration of predictive maintenance and machine learning techniques to enhance the reliability and efficiency of light rail vehicle (LRV) operations, with a specific focus on lead-acid batteries. Advanced data acquisition technologies and machine learning algorithms, particularly Gaussian Process Regression (GPR), were utilized to develop robust predictive models for battery health monitoring and maintenance optimization.
The research employed a comprehensive case study approach, focusing on the Škoda ForCity Smart Artic X34 tram. Key findings demonstrated a moderate positive correlation between battery temperature and voltage, highlighting temperature's significant influence on battery performance. The key_value metric, derived from voltage, temperature, and duration data, was effective in assessing battery health, providing a unit-specific measure of battery response to load changes.
The GPR model, configured with a Rational Quadratic kernel and a White Kernel, achieved high predictive accuracy with a mean squared error (MSE) of 0.00114, accurately predicting battery capacity changes. These predictive capabilities enabled proactive maintenance scheduling, reducing unplanned downtime and extending battery life, thus improving operational efficiency and cost-effectiveness.
The thesis provided valuable insights into the practical applications of IoT technologies and machine learning in predictive maintenance, offering substantial benefits in terms of operational reliability, cost efficiency, and sustainability for urban transport systems.
Recommendations for future implementations included enhancing data integration, exploring advanced machine learning techniques, developing real-time monitoring systems, and conducting comparative analyses of different battery types. Economic impact assessments and sustainability evaluations were also suggested to further optimize predictive maintenance strategies.