Development and evaluation of health monitoring techniques for railway point machines

A more predictive maintenance approach for railway point machines can lead to fewer delays, lower operating costs, and improved safety for the railway industry. However, achieving this improvement in reliability, availability, and maintainability, necessitates the development of a predictive monitoring system for the fleet of point machines. This monitoring system would include acquiring the appropriate signals, the use of advanced pattern recognition and statistical methods for finding the early symptoms of point machine degradation, and a user interface for displaying and reporting the health results. This research effort focuses on the development of the statistical and pattern recognition tools for point machine health monitoring. The health monitoring algorithm development was performed for an MET electromechanical point machine, in which a test-bed was constructed for the data collection and algorithm development and validation. The introduction of seeded faults and degraded conditions were used on the test-bed to determine the level of accuracy, false alarm rate, and sensitivity of the health-monitoring algorithm to various failure modes. A series of algorithm processing steps, which include segmenting the point machine movements, statistical feature extraction, and multivariate health assessment tools such as principal component analysis and self-organizing maps were used in this study. The results showed significant promise, and it appears that the health condition assessment is quite good for various levels of friction degradation, obstacle detection, and detecting the higher levels of misalignment. From a technical perspective, future work could include a more robust and general approach for segmentation, along with an improved method for threshold setting. From an implementation perspective, there is still additional work needed for selecting the appropriate hardware and software platform for deploying this health monitoring system for point machines.