实验室硕士生李之颖论文被振动领域TOP期刊JSV录用
祝贺LASIS实验室硕士生李之颖的论文《Rule Learning-Based Mode Switching Control Method for Active Suspension Systems》被振动领域TOP期刊《Journal of Sound and Vibration》(JSV)录用。
Abstract: Active suspension can improve vehicle ride comfort and handling stability, but it has the problem of high energy consumption. Semi-active suspension features low energy consumption and a simple structure, yet its control capability is limited under complex operating conditions. To balance suspension performance and energy consumption under different driving conditions, this paper proposes a rule learning-based mode switching control method for active suspension systems. First, a cost function-based mode switching (CFMS) strategy evaluates full-active and semi-active model predictive control (MPC) responses offline to generate a suspension mode dataset balancing performance and energy consumption. Then, multi-state clustering and the repeated incremental pruning to produce error reduction (RIPPER) algorithm convert labeled vehicle states into prioritized explicit rules. Finally, a rule learning-based mode switching (RLMS) strategy combines rule matching with temporal-consistency evaluation to enable adaptive switching between the two modes. Results show that under the worldwide harmonized light vehicles test cycle (WLTC) with Class C road driving condition, the control effect of the RLMS strategy is comparable to that of the CFMS strategy. Under the new European driving cycle (NEDC), which was not included in the training set, the RLMS strategy still exhibits excellent control performance and energy consumption advantages. It reduces the average computation time per step by 55.44%, demonstrating potential for real-time applications. Compared with the full-active suspension, the RLMS -based full-active suspension significantly reduces energy consumed by 59.09% with only a slight sacrifice in vehicle ride comfort. It still maintains stable pitch and roll responses, demonstrating good generalization performance.
Keywords: Active suspension; Mode switching; Rule learning; Repeated incremental pruning to produce error reduction algorithm; Model predictive control.