Predictive Vehicle Diagnostics through Machine Learning
- PDF / 689,629 Bytes
- 6 Pages / 595.276 x 790.866 pts Page_size
- 109 Downloads / 257 Views
AUTHORS
Dipl.-Ing. Mathias Kohlhase is Doctoral Candidate at the Institute of Automotive Engineering (IAE) at the Technical University of Braunschweig (Germany).
Predictive Vehicle Diagnostics through Machine Learning Wear predictions based on vehicle diagnosis data collected online offer a high potential for better planning vehicle maintenance in the future. At the Technical University of Braunschweig in cooperation with Volkswagen Commercial Vehicles, a methodology has been developed that makes it possible to diagnose the wear condition of com ponents and predict a defect before it becomes a breakdown. The first component in the research focus is a mechanical component, the V-ribbed belt.
Prof. Dr.-Ing. Ferit Küçükay is Director of the Institute of Automotive Engineering (IAE) at the Technical University of Braunschweig (Germany).
Prof. Dr.-Ing. Roman Henze is Head of Vehicle Dynamics and Active Systems at the Institute of Automotive Engineering (IAE) at the Technical University of Braunschweig (Germany).
Dr.-Ing. Can Yilmaz is Engineer in the field of Decarbonization at Volkswagen Commercial Vehicles in Hannover (Germany).
74
© IAE
At the Institute of Automotive Engineering (IAE), sensory anal yses were therefore carried out on the test rig using the example of the V-ribbed belt and methods for predicting wear based on machine learning were investigated.
1 MOTIVATION 2 STATE OF THE ART 3 RIBBED BELT TEST BENCH 4 MACHINE LE ARNING ME THODS 5 TEST SCHEDULE 6 DATA PRO CESSING AND MODELING
2 STATE OF THE ART
7 E VALUATION 8 C ONCLUSION
1 MOTIVATION
The current maintenance system used in the automotive industry is based on a preventive approach. The maintenance intervals at which a vehicle must be inspected are prescribed by the car manu facturer. It is not uncommon for components to be overlooked whose service life has already expired by 90 %, or for components to be replaced that still have a longer remaining service life. The V-ribbed belt is often subject of both scenarios and in the worst case responsible for a breakdown. In order to avoid the resulting breakdowns and to save maintenance costs, it is advantageous to predict workshop visits.
Previous investigations into the sealing of V-ribbed belts are based on a theoretical calculation of the strength of V-ribbed belts and their design [1, 2]. Accordingly, the service life of the belt depends on two main factors, FIGURE 1: on the one hand on mechanical stresses (bending stresses, circumferential and normal forces), on the other hand on environmental influences (temperature, foreign bodies in the belt drive, FIGURE 2). Apart from these theoretical assumptions, there are no practical research projects that can be used for wear analysis of the V-ribbed belt. The service life is mainly dependent on the abovementioned influences; the greatest forces act during the starting process due to the high accelerations. In the research project, continuous start-up cycles were there fore simulated on the test rig in order to bring about the end of the belt’s
Data Loading...
