Towards simulation-based digital twins: sensor placement studies for structural health monitoring of railway switches by implementing numerical simulations with calibrated track models

Pillai, Nikhil Rajesh ORCID: 0000-0002-1075-3933 (2024). Towards simulation-based digital twins: sensor placement studies for structural health monitoring of railway switches by implementing numerical simulations with calibrated track models. University of Birmingham. Ph.D.

Preview

Pillai2024PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (15MB) | Preview

Abstract

This thesis contributes to the knowledge of how train-switch interaction models can be developed and calibrated against reference assets or models in order to support the predictive maintenance of switches. The modelling approach that has been suggested within this thesis could be further developed for implementation as part of a live Digital Twin. The key research that will be demonstrated through this thesis includes: (i) the selection, development and evaluation of numerical simulation approaches to address a range of different research questions; (ii) calibration of the track dynamics for the developed FE model against the reference model, (iii) rolling contact simulations for the wheel and the switch, as well as the validation of the model against the reference model, and (iv) an approach to determine sensor placement in order to implement effective fault detection.

At first, the existing examples of S&C modelling to simulate train-turnout interactions, wheel-rail contact and damage prediction were reviewed. A number of the results from this evaluation have been implemented in order to determine which modelling approaches will be suitable to determine the appropriate locations in switches that are susceptible to surface rail damage, the placement of sensors for fault detection and the testing of the potential for condition monitoring.

Based on the results from the evaluation of the numerical simulation approaches, a combined MBS-FE numerical approach has been implemented for modelling the switch. A 3D solid FE model for a railway switch assembly has been calibrated for representing the dynamic properties for a reference track model that is based on the properties shared in the S&C benchmark project, where 20 research institutions achieved consensus for the results from the dynamic vehicle-turnout interaction simulations whilst implementing different software packages. Furthermore, numerical simulations have been carried out after introducing the rail and substructure defects such as wear, Rolling Contact Fatigue (RCF) and voiding.

The results from the numerical simulations have been used to assess the best locations on the switch rail for the placement of strain sensors for monitoring the health of the switch. The characteristics (range, data acquisition rate, etc.) have also been discussed. Strain sensitivity analysis has been carried out for determining the direction of the strain measurement that is sensitive to the defects that have been introduced.
The correlation analysis between the rail strain and wheel/rail impact forces has been carried out and it has been discovered that a good correlation between the strains and the impact forces is conditional on the sensor placement location. The measurement of the vertical strains on the rail web is more sensitive to the lateral position of the wheel than the measurement of longitudinal strains on the rail foot. The measurement of the strains in the sleeper spacing has shown potential for the detection of the specific defects that have been introduced into the model. The surface defects have been introduced on sleeper-supported rail as well as the section of the rail in the sleeper spacing. The ability of the sensor to detect the faults whilst being installed at the proposed locations has also been numerically verified.

To conclude, an approach has been demonstrated for taking decisions on structural health monitoring of switches through this research. The requirements for sensor specification and the best sensor locations for detecting faults can be assessed digitally, which reduces the need for field experimentation. The knowledge from this research can be taken further for developing Digital Twins of track infrastructure. Real-time data from the measurement can be integrated with the numerical data developed in the present research to obtain a better insight of the switch dynamic behaviour as well as damage prediction.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):