Self-Adaptation Strainer Block Detection for Heating Ventilation and Air Conditioning System of Urban Metro Transportation

Along with the rapid development of urban metro transportation, the operating mode of heating ventilation and air conditioning system becomes diversified and self-adaptive, which results in significant differences in health status of strainers assembled in separated units. So far, the maintenance of strainers has been proceeded periodically regardless of confirming the actual status of individual parts, as a result, the maintenance strategy intends to purify or replace [1] the ones which are not too dirty to degrade the air conditioning performance [2]. On the other hand, the strainers are more likely to be dirty in summer than other seasons [3], and the suggestion of periodic maintenance remains the same after rolling stock delivered, hence the over maintenance of strainers is a common issue for urban metro transportation. Many researches and efforts have devoted to this field, the most popular solution is to install pressure transducers [4] to monitor the difference between two sides of the strainer. It does help to improve the efficiency of maintenance, however strict limitations, such as refurbishment of existed rolling stocks, unstable detection of sensors, unpredictable changes of real-time thresholds, remains not solved which makes the solution difficult to widely applied. In the aim to reduce the over maintenance and waste of spare parts without installing new sensors or devices, the essay introduces a comprehensive model to estimate the health level of strainers. The model is trained based on ten-dimensional real-time train-borne data, including three-phase currents of motors, mode of dampers, etc., and K-means is adapted to improve the accuracy of strainers block detection. The model converges to four levels of blocking risks: none, low, medium and high, the simulation results generated by model matches the experiment data. Finally, the essay discusses the benefits of wide application and suggests rolling-stock maintenance from periodic mode to status-driven mode.