Knowledge-based Engineering

In the course of digitization, there are new arising opportunities to improve the methods for designing vehicle components.

For instance, today a transmission is designed for a certain lifetime or mileage, taking into account extreme load cases. However, only a small proportion of drivers (about 2-3%) cause such extreme load cases.

With the help of smart-big-data methods and innovative approaches for real-time prediction of the remaining lifespan, vehicle components can be developed more easily and cost-effectively in the future. An important aspect is the knowledge and consideration of the actual user behavior. Based on measured driving profiles, test cycles are synthesized which better map the load spectum of real users than existing standard cycles. We are pursuing these and other ideas at the IMS as part of Vision Vehicle 5.0. This unites the research projects of several institutes of TU Darmstadt, which deal with the implementation of the future, knowledge-based vehicle.

Current Projects Related to this Key Topic:

CRISP TP3.2 – Part of CRISP

Period: 2017 – 2020

Cooperation: ATHENE

Support: see right

The project aims to develop methods for the collection, aggregation and evaluation of vehicle operating data, which at the same time allow protection of individual privacy. For this purpose, three exemplary applications are being investigated: “software-based lifetime monitoring” in the area of lightweight construction, the synthesis of representative driving driving cycles in the field of emissions/efficiency and geodata acquisition for the transmission of local dangers in the field of autonomy. All applications will be assessed on the data they need to enable the implementation based on aggregated-data technical and how sensitive information about individuals can be protected.

The focus of IMS research in the project is on methods that enable a stochastic synthesis of short and representative driving cycles based on aggregated fleet data. The methods work multidimensionally and can consider the interdependency of multiple signals in realistic test cycles. For example, not only the speed specification plays a decisive role for the consumption assessment of vehicles (which is what classic driving cycles limit), but also the slope profile and the outside temperature, among other things.

The user and his / her privacy are an essential aspect. Therefore, possibilities for the aggregation of the operating data before the storage are examined, which maintain technical information and at the same time suppress privacy-critical data. The goal is to enable technical applications with the least possible restriction of privacy.

Contakct: Arved Eßer

Further informationen you can find on the Project homepage.

Selected publications:Enabling a Privacy-Preserving Synthesis of Representative Driving Cycles from Fleet Data using Data Aggregation

Period: 2018 – 2021

Cooperation: ATHENE

Support: see right

The IMS is researching on adaptive, self-learning longitudinal control as an application of resilient IoT systems. The use of machine learning methods in combination with the consideration of aggregated driving data and vehicle-to-vehicle communication (V2V) is being investigated. Thereby, the longitudinal dynamics of the vehicle can be optimised, e.g. to achieve energy efficient or comfortable driving characteristics. At the same time, the functional safety has to be assured as a part of a resilient technical system.

Contact: Tobias Eichenlaub

Period: 2018 – 2020

Cooperation: Daimler AG

Novel electric drives offer great potential in many ways compared to conventional powertrains. The additional degrees of freedom associated with this allow, among other things, the development of new comfort functions in order to increase the acceptance of this new technology and to strengthen the connection between driver and vehicle. In cooperation with Daimler AG, the IMS is pursuing the vision of an intelligent comfort function that learns while driving and thus adapts individually to the needs of the driver. For this purpose, current research approaches from machine learning are picked up and combined with classic engineering approaches. The aim is to exploit synergies such that the robustness of classical methods can be combined with the agility of machine learning in vehicle functionality. In addition, it will be examined to what extent the intelligent integration of the cloud can be used to improve the overall system.

Contact: Philippe Jardin

Period: 2018-2019

Cooperation: COMPREDICT und Universität Tongji

Support: Getrag China

As part of the project, a method is being developed to monitor the operating status of the wet disc clutch. For this purpose, the functions, such as the monitoring of the load condition of the disc plates and the temperature calculation while driving are provided. After the development phase, this method is merged with an existing gear state monitoring algorithm of COMPREDICT, and then investigations are to be made in a simulation environment of Tongji University. Such fundamental investigations and the simultaneous optimization of the method should ensure that a continuous gain of knowledge takes place with the exclusion of systematic errors. Finally, the entire monitoring algorithm is implemented in a test vehicle and tested in the real driving environment. The recorded measurement data can be used to develop and improve the method.

The project initially serves as a feasibility study for a condition monitoring of wet disc clutches. Although there are several publications in this field, these are largely theoretical studies or methods obtained by test benches. This project is characterized by the fact that the method is developed on a real system. Here, the real-time functionality under limited computing power represents the biggest challenge.

Contact: Ping He