Descriptive modelling of turnout settlements: A step toward data-driven decision support in infrastructure management

This paper presents a descriptive, data-driven model based on track settlement rates to support turnout design decisions. Given the wide range of turnout configurations, infrastructure managers must select designs that exhibit favourable long-term behaviour. Descriptive models can support this process by combining empirical in-field data with objective mathematical metrics. Key challenges arise from non-normally distributed settlement rates and strong interdependencies between turnout features, which are addressed using non-parametric methods, scaling, and targeted subsetting. The approach is demonstrated using settlement rates derived from 25 years of longitudinal level measurements recorded by regular track recording vehicles, covering 446 turnouts on the Austrian railway network. Driving direction is shown to have no influence on overall settlement behaviour. Sleeper type, by contrast, exhibits a clear and statistically robust effect: concrete sleepers equipped with under-sleeper pads show about halved settlement rates compared to conventional concrete sleepers and reduced by 40% compared to wooden sleepers. The effects of turnout radius and turnout type are physically plausible and consistent across subsets, but not always statistically significant due to the high variability inherent in field data. Despite remaining limitations, the proposed model provides a transparent and scalable basis for comparing turnout designs within a national network and across infrastructure managers. From an engineering perspective, it enables the identification of design configurations associated with reduced settlement behaviour and maintenance demand. When combined with life-cycle cost models, it represents a step toward data-driven decision support for more informed and economically efficient turnout design decisions.