Introduction
In the realm of predictive maintenance and reliability engineering, condition indicators play a pivotal role in assessing the health of track infrastructure. Vibration signals, collected from the wheels of the train as they travel along the track, emerge as a symphony of information, offering a unique insight into the condition of the track. Developing condition indicators from these signals demands a strategic blend of data analytics and signal processing. In this blog, Antti Laiho explains the methodical approach to developing condition indicators for SmartVision Track Condition Monitoring, shedding light on the key stages and considerations involved.
Understanding Condition Indicators
Condition indicators are quantitative measures that reflect the state of an asset. They are derived from various sources of data such as sensor readings, performance metrics, and historical records. The goal is to transform this data into meaningful actionable information that helps maintenance teams make informed data-driven decisions about when and how to perform maintenance activities. Vibration signals are like a secret language spoken by machinery, revealing clues about its health, alignment, and potential issues. Monitoring these signals allows for the early detection of anomalies, paving the way for predictive maintenance strategies.
The Process
1. Define Objectives
At the outset of developing condition indicators, it’s imperative to chart a clear course by precisely outlining the objectives of the condition monitoring program. Are you aiming to predict impending failures, optimize maintenance schedules, or extend the operational life of your assets? These objectives provide the framework upon which the condition indicators will be built. Simultaneously, it is essential to identify the critical parameters that wield direct influence over the asset’s performance. Whether it’s temperature, vibration, pressure, or other key metrics, understanding these parameters lays the groundwork for selecting the most relevant data sources and refining the scope of your condition monitoring initiative. In essence, a well-defined set of objectives and critical parameters sets the stage for a targeted and effective development journey.
2. Data acquisition
To ensure comprehensive coverage and accurate monitoring, you must instrument a carefully selected number of trains that regularly travel on the lines to be monitored. The optimal number of installations is determined by various factors, including train schedules and the types of faults typically observed on the lines. By strategically deploying the system on these trains, you can achieve maximum effectiveness in detecting and analysing track conditions, allowing you to capture a wide range of data and gain valuable insights into the health of the track infrastructure.
3.Noise reduction
Any vibration sensor mounted on a train’s bogie results in a very complex signal characteristic containing signatures from both the train (wheel, bearings, etc) and the track (joints, points, sleepers, track defects etc). Our recent article about SmartVision Track Condition Monitoring explains some of the background of this https://www.eke-electronics.com/smartvision-track-condition-monitoring-article-published-in-metro-rail-news/. Employing time-domain techniques and multi-channel measurement locations can mitigate signal noise effectively. This preprocessing step prepares the signal for further analysis using time-frequency methods, facilitating the derivation of condition indicators.
4. Feature Extraction
Feature extraction is a pivotal process in defect identification, involving the detection of patterns indicative of specific types of defects. Utilise statistical, frequency domain, and time-series analysis techniques for defect identification. Machine learning, particularly unsupervised methods, aids in clustering unlabelled data, leading to pattern identification. Refinement through historical data training enhances the model’s ability to discern intricate patterns.
5. Normalisation
The speed of a train significantly influences the vibration signal, necessitating efforts to normalize the data relative to speed. Speed normalization entails employing both modelling and statistical analysis techniques for measurement data. Wheel responses, influenced by various rail elements such as switches and insulated joints, may exhibit diverse speed dependencies contingent upon their unique characteristics and conditions. It is important to note that speed-dependence can also manifest in strongly nonlinear ways. In such cases, statistical analysis techniques to address linear and nonlinear speed dependencies should be employed, leveraging measurement data spanning a relatively large speed range to accurately capture these complexities.
6. Validation
The validation of condition indicators stands out as perhaps the most challenging phase in the process. Replicating real-life conditions in the lab, especially considering the extreme temperatures and varied landscapes that trains encounter, poses significant difficulties, particularly in regions like Finland. Given these challenges, comparing detections with results from field inspections becomes paramount. This allows for crucial feedback on the development of condition indicators, ensuring their effectiveness in practical scenarios.
7. Implementation
Integrate validated condition indicators into the Track Condition Monitoring solution, ensuring continuous monitoring to minimise false negatives and false positives.
Conclusion
In the dynamic landscape of railway operations, decoding the vibrations of in-service trains unlocks a realm of possibilities for proactive track maintenance. By translating the vibrations of in-service trains into actionable insights, railway maintainers/operators can ensure the smooth, safe, and reliable operation of their networks. Developing condition indicators is a dynamic and iterative process that combines domain expertise, data analysis, and technology. By investing in robust condition monitoring programs, organisations can enhance asset reliability, optimise maintenance strategies, and ultimately achieve greater operational efficiency. As technology continues to advance, the fusion of vibration analysis and data analytics promises to enhance our ability to decipher the language of the rails, evolving even more sophisticated and accurate indicators, moving organisations towards a predictive maintenance capability.
By Antti Laiho
Condition Monitoring Expert
EKE-Electronics
More about Antti Laiho
Antti has worked at EKE-Electronics for 5 years developing condition indicators for train and track assets. He has a PhD in electromechanics and has published over 20 journal articles related to vibration analysis on various assets such as electric machines and rotating machinery. Since 2019 he has worked on developing vibration-based condition indicators which are integrated in the SmartVision Track Condition Monitoring software by EKE-Electronics. These condition indicators have been validated and tested in collaboration with FITA.
If you would like to find out more about how Antti’s condition indicators help improve track reliability, safety, and efficiency please contact us to arrange a demonstration.