Scholarlink

This study evaluates the effectiveness of a model that combines simulation with a long short-term memory (LSTM) network trained on simulation data to predict key vehicle parameters. Using the BMW X3 xDrive as a case study, simulations performed with MATLAB Simulink are compared with empirical data from a road-to-rig (R2R) test rig using the standard Worldwide Harmonized Light Vehicles Test Procedure (WLTP) driving cycle. While simulation is a valuable tool for vehicle analysis, achieving accurate real-world data through simulation remains challenging due to various physical and environmental factors. The proposed approach shows moderate fit between simulation and test rig data, highlighting the need to integrate real-world elements to simulation model accuracy. Further improvements could be achieved by analyzing the behavior of components such as transmission and powertrain efficiency to gain deeper insights into vehicle performance. The study also demonstrates the potential of advanced machine learning techniques, with the LSTM model predicting acceleration from simulated torque data with an R2 score of 0.67. This predicted data is then compared with actual test rig data. The simulation model is also applied to analyze slip effects and driving resistance under varying road conditions, showcasing its flexibility for broader testing scenarios. This work highlights the potential of combining simulation with predictive models to extend testing capabilities and adapt to different approaches based on specific requirements.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving Vehicle Dynamics in Road-to-Rig Testing by Integrating Model-Based Simulation and LSTM Predictions

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Akash Mangaluru Ramananda,
  • Timo König,
  • Simon Zimmermann,
  • Philane Tshabalala,
  • Rangith Baby Kuriakose,
  • Steffen Schwarzer,
  • Markus Kley

摘要

This study evaluates the effectiveness of a model that combines simulation with a long short-term memory (LSTM) network trained on simulation data to predict key vehicle parameters. Using the BMW X3 xDrive as a case study, simulations performed with MATLAB Simulink are compared with empirical data from a road-to-rig (R2R) test rig using the standard Worldwide Harmonized Light Vehicles Test Procedure (WLTP) driving cycle. While simulation is a valuable tool for vehicle analysis, achieving accurate real-world data through simulation remains challenging due to various physical and environmental factors. The proposed approach shows moderate fit between simulation and test rig data, highlighting the need to integrate real-world elements to simulation model accuracy. Further improvements could be achieved by analyzing the behavior of components such as transmission and powertrain efficiency to gain deeper insights into vehicle performance. The study also demonstrates the potential of advanced machine learning techniques, with the LSTM model predicting acceleration from simulated torque data with an R2 score of 0.67. This predicted data is then compared with actual test rig data. The simulation model is also applied to analyze slip effects and driving resistance under varying road conditions, showcasing its flexibility for broader testing scenarios. This work highlights the potential of combining simulation with predictive models to extend testing capabilities and adapt to different approaches based on specific requirements.