Clutch Housing Temperature Prediction during Repeat Restart Test Using Machine Learning

Abstract

<div class="section abstract"><div class="htmlview paragraph">Product validation time reduction and limit number of physical testing is major challenge all over the world OEMs are facing and they are trying to use latest technologies to fill gap between design parameters, simulated results, and physical validation results. Automotive industry is going through a major transformation with use of artificial intelligence and machine learning and especially in the area of transmission system design and development where lot of data is available from physical testing. Clutch is still being used in internal combustion engines vehicles. Clutch is an important part in transmission system in vehicle, which transmits power generated from engine to transmission and changes the gears at different speed. Design and validation of clutch is a critical and laborious task. Clutch failure occurs due to excessive rise in temperature. The motivation behind this work is to reduce clutch design and selection cycle time and iteration, since physical testing and CAE iteration are a time consuming and costly process for any automotive OEMs. Physical testing requires manpower and test setup. If clutch design engineer knows temperature inside clutch housing before design finalization then corrective actions can be taken to avoid failures during field trials by providing early feedback from results predicted by machine learning model. This paper focuses on prediction of clutch housing temperature during physical validation of a clutch using machine learning. Historical data of different vehicles is used for development of machine learning model. The current study focuses on the machine learning approach for prediction of clutch housing temperature. The machine learning methodology and results correlation between machine learning predicted results and physical test results for different types of commercial vehicles. The developed solution using machine learning helps clutch design engineer in selection of critical clutch parameters so that clutch design can be improved at early vehicle design stage.</div></div>