REAL-TIME PROGNOSTICS AND HEALTH MANAGEMENT WITHOUT RUN-TO-FAILURE DATA ON RAILWAY ASSETS

Abstract

Predictive maintenance is fundamental for working on the dependability and execution of assorted parts
and frameworks. In any case, the shortfall of open rush to-disappointment information every now and again blocks
the making of exact prognostic models. This study handles this trouble by presenting a creative prognostic strategy
explicitly intended for reasonable railroad support arranging, with an accentuation on entryway frameworks. The
significant objective is to give a prognostic methodology fit for determining the leftover valuable existence of railroad
entryway frameworks without relying upon race to-disappointment information. The strategy tries to work with
productive prescient upkeep arranging by assessing shortcoming seriousness and computing the time left until basic
issue limits are reached. The proposed approach utilizes engine current signs to deliver a disintegration marker for rail
line entryway frameworks. “Dynamic time warping (DTW)” is used to assess the closeness among typical and
blemished conduct, though the K-means procedure is applied to decide shortcoming seriousness. A delegate time
assessment is performed for every seriousness level, empowering the estimate of residual time until basic shortcoming
levels are accomplished. This strategy doesn't require rush to-disappointment information. The proposed strategy,
through preliminary and examination, is valuable in prescient support making arrangements for railroad entryway
frameworks. The strategy offers valuable experiences for opportune support intercessions by definitively assessing
shortcoming seriousness and guaging the leftover time until significant flaws happen. K-means bunching has been
refined, and Random Forest along with a Stacking Classifier (LGBM+RF+DT) has been integrated into the
undertaking to foresee predisposition type, with a great exactness pace of 99.5.