Usage of Digital Twins for Predictive Maintenance

This presentation was made at the 2019 NAFEMS World Congress in Quebec Canada

Resource Abstract

Drivetrains with dynamic loads require maintenance effort for the whole system and its single components because of possible fatigue failures. In a worst case scenario, a single and preventable component failure can lead to the shutdown of a machine or even an entire production line, if the damaged part is mandatory for its use. This scenario can e.g. lead to a significant economic damage, especially when spare parts are not available.



Today maintenance is often performed either in case of failure (Reactive Maintenance) or in a preventive way (Preventive Maintenance). The latter is an improvement of the first and commonly used but can neglect the maximum utilization of single components in an uneconomic and unsustainable way, caused by a conservative choice of fixed maintenance periods. That is why crucial drivetrain components such as gears, bearings or shafts can be monitored by condition monitoring systems (CMS) to prevent instant failures to happen without any indication. CMS can furthermore schedule maintenance periods to maximize the utilization of components, but it is limited to measurable signals and to the characteristics of the past operation behavior. The number of installable sensors and their position are mostly restricted in practice, which is a disadvantage in understanding the dynamic characteristics of today’s complex drivetrains.



A current research project has the objective to investigate the possibility of fixing these disadvantages by combining CMS with a real-time digital twin of the system at the example of a complex gearbox. A validated digital twin can provide a virtual conditioning monitoring system (VCMS) to monitor any state of the system at any time, at any resolution and allows e.g. the prediction of the system behavior. It provides virtual loads for components where sensors are not applicable in practice. Digital twins can additionally calculate arbitrary load cases to investigate the systems characteristics of future operation conditions before they are finally applied to the real system. The comparison of virtual signals with real sensors in frequency and time domain ensures the accuracy of the models and their suitability for prediction.



With such a combination of CMS and a validated digital twin, an individual maintenance strategy is possible (Predictive Maintenance). Within this, the CMS monitors the gearbox for failures and provides data for validation, whereas the digital twin allows a signal evaluation at any component and a prediction of the possible system behavior for a short time period in real-time. Moreover, it provides loads to reevaluate the maximum utilization of single components and is the basis to set up a predictive maintenance strategy. It relies on the individual system behavior and is therefore more flexible as conservatively chosen, fixed maintenance periods.

Document Details

ReferenceNWC_19_344
AuthorSchulz. C
LanguageEnglish
TypePresentation
Date 18th June 2019
OrganisationHochschule Anhalt
RegionWorld

Download


Back to Search Results