A methodology for data-driven modeling and prediction of the drag losses of wet clutches

• DOI: 10.1007/s10010-023-00661-y
• Publication Date: 2023-04-25T11:15:01Z
• Authors: Lukas Pointner-Ga...; Katharina Voelkel; Hermann Pflaum; Karsten Stahl
• Keywords: Experimental data; Drag equation

Abstract In wet clutches, load-independent drag losses occur in the disengaged state and under differential speed due to fluid shearing. The torque of a clutch can be determined accurately reliably by means costly time-consuming measurements. As an alternative, already precisely calculated early development phase using computing-intensive CFD models. contrast, simple analytical calculation models allow rough but non-time-consuming estimation. Therefore, aim this study was develop methodology that used build data-driven model for prediction clutches with low computational effort and, at same time, sufficient accuracy consideration high number influencing parameters. For building model, we use supervised machine learning algorithms. covers all relevant steps, from data generation validated as well its usage. comprises six main steps. Step 1, is generated on suitable test rig. 2, characteristic values each measurement are evaluated quantify loss behavior. serve target train model. 3, structure quality dataset analyzed subsequently, input parameters defined. 4, relationships between investigated (model input) output) determined. Symbolic regression Gaussian process have both been proven task. Lastly, 5 predict values. Based predictions, predicted function 6, approximation function. allows user-oriented even time.