NEXTSENSE GmbH
Predictive Wheelset Management: From Measurement to Demand Planning
This article first appeared in the Railway-News magazine Issue 2 2022.
Predictive wheelset management can be considered a key innovation for the railroad industry, as the requirements are mainly met with digitalisation and technology developments (e.g. transmission speed and storage capacity). This article will present a methodology for optimised demand planning of wheelsets.
TMH International is an internationally active manufacturer and maintenance provider of rail vehicles. Maintenance planning is controlled at headquarters and carried out at the appropriate time, preferably locally near the customer. In our example that component is a particularly important one: the wheelset, a safety-critical wearing part to which certain standards apply to ensure safety. Its maintenance planning is also worthwhile from an economic perspective – by replacing them at an optimal, rather than scheduled time.
The aim is for the entire process – from performing measurements, to monitoring values and planning and executing the maintenance – to be as digitalised as possible.
Predictive, Data-Driven Maintenance
Predictive wheelset management is a preventive maintenance strategy. A single variable (often time or mileage) is assigned a high correlation to the component’s wear curve, and experience and/or engineering knowledge is used to determine an interval after which the component must be serviced/ replaced/maintained. This rule is fragile because it is a static solution for a dynamic system. Data-driven wheelset management adds certain influencing factors to the solution (or model) by digitalising, evaluating and collecting the factors in the model. In addition to domain and engineering expertise, data science supports the accuracy of predictions with stochastics and other statistical/ mathematical methods. The quality of the prediction is determined by the underlying data quality. What parameters are recorded and how accurately this is done is key.
