Comparative analysis and optimization of nonlinear hysteresis models for a magnetorheological fluid dual-clutch of an electric vehicle transmission

Abstract Electric vehicle (EV) drivetrains have witnessed remarkable progress, prompting intensified research into advanced transmission systems. Magnetorheological fluids (MRF) clutches offer precise modulation of input currents, enabling swift and seamless torque delivery for EV transmission systems, owing to their exceptional performance. The transmission of an EV requires MRF-based clutches to deliver a precise and rapid torque transfer during gear shifting. In these scenarios, the inherent current rate-dependent hysteresis of the MRF-based clutches between the output torque and input current poses a significant challenge in accurately regulating output torque. Therefore, an accurate clutch model of the MRF-based clutches that can describe the rate-dependent hysteresis is crucial to achieve precise control of the output torque. This study investigates the nonlinear hysteresis phenomena using a prototyped MRF dual-clutch (MRFDC) for the transmission system of EVs, followed by a comprehensive analysis of three widely used hysteresis models: two parametric models, including the Bouc-Wen (BW) model and algebraic model (AM), and a non-parametric model, the NARX model. Accuracy, fitting time, and stack size are selected as the main indicators to evaluate the three models comprehensively. Results indicate that the NARX model has exceptional accuracy compared to the others, while it has a much higher memory requirement. The algebraic model shows a great advantage in computational efficiency because it has a straightforward expression. The BW model is in the middle position for all three indicators. To optimize the classic BW model (CBW), a fractional-order modified BW model (FOMBW) is proposed based on the polynomial input function and fractional-order derivatives. The proposed FOMBW model demonstrates superior capability in capturing asymmetric and rate-dependent characteristics compared to the CBW model. These findings provide the basis for choosing an appropriate model to effectively capture nonlinear current hysteresis phenomena within MRFDC with the requirement for precise torque control during gear shifting.