Objective and Subjective Classification of Creep Groan Noise

Jurij Prezelj*, Severin Huemer-Kals, Karl Häsler

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review


Stick-Slip effects occur at the contact surface between the brake pads and the brake disc. The Creep Groan phenomenon is associated with stick-slip effects in the frequency range up to 500 Hz. It induces vibrations on the entire brake system and influences the structural vibrations of the vehicle. Vibrations are transmitted to large surfaces which produce audible noise. Creep Groan Noise is unpleasant and can be considered as a defect, especially for drivers of luxury cars. Harshness and annoyance caused by Creep Groan noise depends on many factors. For this reason, vehicle tests often include a subjective analysis of the Creep Groan Noise. Unfortunately, subjective tests suffer from the inherent uncertainties of subjective evaluation. A model for the assessment of Creep Groan noise harshness is necessary to reduce these uncertainties and to improve the Creep Groan assessment. For this purpose, experimental tests with over 1000 measurements were carried out. Many different features from the measurement data were extracted and analyzed. Selected features (quasi acoustic emission, simulated sound pressure level and vibration level) were used to classify Creep Groan into eight annoyance classes using Kohonen's Self Organizing Maps and a K-Means algorithm. The results of both unsupervised classification algorithms were correlated with a subjective quantification ranging from 1 (very poor) to 10 (excellent). After the right interpretation of the unsupervised algorithmic classification, performed with the K-Means algorithm, the correlation between the algorithmic classification and the subjective quantification of the creep groan harshness is obtained. It is shown that Kohonen's self-organizing map, in combination with selected features of vibroacoustic signals, provides readily usable results without any need for subjective evaluations.
Original languageEnglish
Title of host publicationProceedings of Eurobrake 2020
PublisherFédération Internationale des Sociétés d'Ingénieurs des Techniques de l'Automobile FISITA
Number of pages8
Publication statusPublished - 2020
EventEuroBrake 2020 - virtuell, Spain
Duration: 2 Jun 20204 Jun 2020


ConferenceEuroBrake 2020


  • Kohonen Neural Network
  • Self Organizing Maps
  • K-Means
  • Unsupervised Classification
  • Creep Groan
  • NVH
  • Noise Event Classification
  • Vibrations
  • Acoustic Emission

Fields of Expertise

  • Mobility & Production

Cite this