Figure 7.

(Left) Online HSR turnout condition monitoring system; (right) defect diagnosis method with model updating approach [70].

updating methods have been investigated dealing with great uncertainties arisen from the online monitoring data for more accurate and efficient damage diagnosis. AE method incorporating Bayesian framework is utilised in an online rail turnout crack monitoring system developed by CNERC-Rail [70]. The method is able to detect defects without training data of damaged rail structure and the monitoring systems have been implemented on Shanghai-Nanjing HSR lines, as shown in left panel of Figure 7. The rail turnout conditions are indicated in a probabilistic manner through a structure health index (SHI), as shown in right panel of Figure 7.
