Wheel-rail contact force reconstruction using trackside measurements and inverse models

Railway condition monitoring is a fundamental aspect that influences operational safety, costs and efficiency. Wheel-rail contact forces, specifically vertical wheel-rail contact forces provide crucial information that can be further analysed and used by decision makers of railway operators to optimize planning and operation. Current methods of vertical force measurement involve onboard measurement systems, which are costly, or trackside measurements many of which cannot isolate contact forces to individual wheels. Few studies have demonstrated trackside load reconstruction methods capable of attributing forces to individual wheels. This research implements an inverse load reconstruction approach employed in bridge load monitoring applications. The forward model of the coupled track-train system is developed and validated using a two-dimensional multibody simulation. The inverse problem is formulated through a linear state-space model of the track and solved using Tikhonov regularization with various regularization matrices. Rail velocity and displacement responses are used as inputs to reconstruct contact loads. The method is evaluated through twelve case studies involving a 40 mm wheel flat, exploring different sensor configurations and regularization strategies for 2 loads. The framework accurately reconstructs individual wheel-rail contact forces with a 2.7 % RMSE and a Pearson correlation coefficient of 0.787 when using four rail-mounted velocity sensors and an identity regularization matrix. The case studies examined load reconstruction for two loads and was extended to four to demonstrate scalability. The results demonstrate the feasibility of trackside load reconstruction for per-wheel force estimation, laying the groundwork for integration into future railway asset-monitoring systems.

Grantor dc:publisher

Stellenbosch : Stellenbosch University

Year dc:date.issued

2026

Author dc:creator

• Perumal, Michael Savio

Advisors dc:contributor.advisor

• Nickerson, Brendon
• Bekker, Anriette

Language dc:language.iso

en

Repository record dc:identifier.uri

https://scholar.sun.ac.za/handle/10019.1/135959

OAI identifier oai:identifier

oai:scholar.sun.ac.za:10019.1/135959