top of page


Mr. John Smith

Job title



Dr.-Ing. Merten Stender, Hamburg University of Technology, GERMANY

Ms. Nadine Jendrysik, Hamburg University of Technology, GERMANY

Mr. Daniel Schoepflin, Hamburg University of Technology, GERMANY

Prof. Dr. David Spieler, University of Applied Sciences Munich, GERMANY

Dr.-Ing. Merten Tiedemann, Audi AG, GERMANY

Prof. Dr. Norbert Hoffmann, Hamburg University of Technology, GERMANY

Tonal high-frequency and high-intensity friction -induced vibrations, such as brake squeal, remain one of the biggest engineering challenges for the braking industry. Particularly, numerical models have not yet reached predictive power, such that most of the design efforts are spent during extensive testing on dynamometers and in vehicles. During these tests, several loading conditions are monitored to allow for sensitivity studies and to support the design of countermeasures.

To reduce testing costs and time, virtual development strategies are of major interest. To this end, we illustrate how a digital twin for the NVH behaviour of a brake system can be set up using deep learning. The operational conditions, such as brake line pressure and velocity, are used as inputs to predict the system’s oscillatory behaviour. Specialised neural networks are used to respect instantaneous dynamical interactions of the multi-physic loadings acting during braking. Further, the class imbalance problem resulting from a few number of squeals compared to the overall number of brakings is considered explicitely.

Results indicate that it is possible to predict the NVH behaviour from the operational loading conditions, which therefore proves determinism in the input-output behaviour of such complex mechanical systems. Furthermore, the trained network, i.e. the digital twin, can be used to assess the brake noise behaviour of a vehicle in a virtual environment using synthesised loading parameters or historical recordings. Interpretation of the trained network allows to learn about sensitivity regimes of a particular brake system with respect to dynamic loading spectra.

The authors NH, MT and MS have been active in the field of brake squeal for many years. Recently, they are interested in data-driven approaches to existing challenges in structural vibrations and complex machine dynamics. Following their contribution to EuroBrake 2019, this work is a direct continuation of the main goal to leverage the potential of artificial intelligence for designing better braking systems.

Brake Squeal Prediction Using Deep Learning

EB2020-STP-003 • Video • EuroBrake 2021 • NVHV


Sign up or login to the ICC to download this item and access the entire FISITA library.

Upgrade your ICC subscription to access all Library items.

Congratulations! Your ICC subscription gives you complete access to the FISITA Library.


Retrieving info...

Available for purchase on the FISITA Store


bottom of page