Working with the OEM

Hexagon has been working on in-service digital twins with a number of different manufacturers of ground vehicles, extending across different noncompeting industries. This paper is based on a collaboration with one company in particular, a world-renowned manufacturer of ground vehicles. For reference, the vehicles are ICE driven and the gearboxes have several discrete ratios.

This project has gathered a wealth of data, all of it derived from the operation of these vehicles and processed by Hexagon using its digital twin. To protect the interests of its client (the OEM), minimal data is displayed. Nonetheless, the narrative, insight, conclusions and ambitions from the study are the same as for the OEM, whose principal engineers have read and approved the text of this paper prior to its submission.

Building the Cloud-Based Digital Twin

Whilst the DTP is not the focus of this paper, the technical methods used owe their origins from the DTP. During the design phase a designer must carry out calculations to confirm that the gearbox is fit for purpose, and in this respect durability/reliability is the most important consideration. The prediction of gearbox reliability does not start from a clean sheet. Gearbox fatigue has been a subject of mathematical methods that have been developed over the years and implemented in standards (ISO 6336 for gears, ISO 16281 for bearings [Ref. 2]) that have become universally implemented.

Additionally, the digital twin has included many of the refinements that have been proven to be essential in the prediction of gear and bearing performance over recent decades. Housing stiffness has been shown to impact gear and bearing misalignment and hence life (Ref. 3), and this is included; likewise, gear micro-geometry is introduced to accommodate misalignment from system deflections, impacting gear stress and fatigue; finally, bearing internal load-sharing, pre-load, and misalignment are also included. All these influences are part of the commercial software package, Romax Enduro, which has been marketed under various Romax names since its release in 1994 and which was acquired by Hexagon in 2020. Romax Enduro is principally used during the design phase of the gearbox, i.e., as a DTP.

Both ISO 16281 and ISO 6336 output component fatigue damage. This does not predict failure. Rather, 100% damage for the L10 life of bearings indicates 10 percent failure, whereas 100 percent damage (a safety factor of 1.0) for gears indicates 1% failure.

Being a long-term user of Romax as a DTP, the OEM had a confirmed model of the gearbox which could be used for the digital twin study. This model had been used in the design of the gearbox in question (i.e., a design twin), and a design duty cycle had been established by the OEM with the intention of representing the usage pattern that the vehicles would see during an anticipated working life.

However, the OEM has strong ambitions for using data to learn more about their applications and to derive a competitive advantage. In recent years, the OEM has installed on its standard production vehicles the instrumentation that is necessary to transmit to the Cloud significant quantities of controlled area network (CAN) data. It is important to note that this setup was established to understand the vehicle as a whole, not just the gearbox—this gearbox digital twin work simply worked with data that was already being downloaded. Of this CAN data being downloaded, the following data were available at a frequency of 1Hz:

Engine speed

Engine torque

Selected ratio

Gearbox oil temperature

It is obvious how this data can be used as an import to define an “in-service duty cycle,” against which component fatigue calculations can be carried out. Indeed, Romax has been able to import time domain data for fatigue calculations for over 15 years. However, this project provided additional challenges.

The first was to set everything on the cloud and to have it operating without human intervention or human interaction. This is required because, in the end, the digital twin would need to process the data from many thousands of vehicles. The digital twin was set up such that the data was transferred and processed daily for each vehicle. Discussions did take place regarding having more (and less) frequent data transfer; however, it was decided that daily processing would provide the best balance between cost/complexity and insight.

Data integration, the ability to pull data across the cloud and integrate different solutions, was also required. In this respect, Hexagon has assembled a team that spans across its different offerings, combining Xalt Integration with Romax. This now handled by Nexus, Hexagon’s open digital reality platform for manufacturing that is developed to provide connectivity and interoperability across all aspects of design, manufacturing, metrology and in-service operation for all Hexagon’s client industries.

Hexagon successfully built the digital twin and as of the end of December 2022, it had successfully processed eight months’ worth of data from 10 different test vehicles, with more being processed each day. The fact of the digital twin’s successful operation is noteworthy. However, of greater interest is what the acquired data has already revealed in terms of how to approach the reliability engineering of gearboxes. This is to be the focus of the rest of the paper.

Challenges in Predicting Bearing Reliability

As has been stated, it is possible to calculate the fatigue damage for the gears (contact and bending) according to ISO 6336 and for the bearings according to ISO 16128. 100 percent damage pertains to a 1 percent failure rate for the gears and a 10 percent failure rate for the bearings. The conversion from fatigue damage to reliability for bearings was discussed at length in a paper (Ref. 4) by the same author at the CTI International Congress and Expo in Novi, MI, on 23–25 May 2023.

The paper identifies that the Digital Twin took the most ‘official’ recommendation for the reliability characteristic of the rolling element bearings, i.e., the reliability factor A1 from ISO. Fundamental to this is a Weibull shape parameter (