Experimental investigation of drag loss behavior of dip-lubricated wet clutches for building a data-driven prediction model

Abstract

AbstractFundamental knowledge of wet clutches’ drag loss behavior is essential for designing low-loss clutch systems. In contrast to the widely investigated injection lubrication, more comprehensive knowledge is needed on the drag loss behavior of dip-lubricated wet clutches. In the development phase, data-driven models allow drag loss predictions with low computational effort and, at the same time, sufficient accuracy. Therefore, this study aimed to deepen and expand knowledge of the drag loss behavior of dip-lubricated wet clutches based on experimental investigations. Moreover, the investigations were designed and conducted so that the generated data and findings can be used in further research for building data-driven prediction models. The investigations were conducted on two clutch systems from automotive and industrial applications. The practice-relevant parameters of clearance, oil level, oil viscosity, and plate shape were investigated based on a mixed-level full factorial design. The evaluation shows that a reduction in drag loss can be achieved primarily by increasing the clearance, reducing the oil viscosity, and choosing waved plates. The obtained drag loss behavior can be traced back to the form of oil displacement from the gaps. The displacement process, in turn, is influenced by the operating and geometry parameters. Although the flow in the gaps develops differently for dip and injection lubrication over differential speed, the study shows comparable integral effects of the influencing parameters for both types of lubrication. The generated datasets contain the investigated parameters as features and characteristic drag loss values as targets. The findings can support the selection and configuration of the machine learning algorithm and the validation of the trained models. The described procedure can serve as a template for generating and analyzing datasets for data-driven modeling of wet clutches’ drag losses.