基于可识别性修正的相关未知模型变量统计模型校准
Fig. 1. Flowchart representing the overall process of the proposed statistical model calibration method
Fig. 2. Probability distributions estimated by the existing OBMC method using only (a) 10 and (b) 30 output test data: both true and estimated distributions represent the same contour levels of the PDF, which are presented at 0.05 intervals from 0.02 to 0.5
Fig. 3. Estimation results from integrating (a) 10 and (b) 30 output test data with 3 input test data; similarly, estimation results from integrating (c) 10 and (d) 30 output test data with 10 input test data: Both true and estimated distributions represent the same contour levels of the PDF, which are presented at 0.05 intervals from 0.02 to 0.5
Fig. 4. Corrected estimation results using only each input test data from the results in Fig. 3
Fig. 5. Estimation results from integrating (a) 10 and (b) 30 output test data with 3 input test data; similarly, estimation results from integrating (c) 10 and (d) 30 output test data with 10 input test data: Both true and estimated distributions represent the same contour levels of the PDF, which are presented at 0.05 intervals from 0.02 to 0.5
Fig. 6. Corrected estimation results using only each input test data from the results in Fig. 5
Fig. 7. Estimation results of adopting the Bayesian method to the identical 10 input test data exhibited in Fig. 6: Maximum likelihood estimation (MLE) results for the unknown model variable (a) X1, (b) X2, and (c) the estimated copula function with the true copula function
Fig. 8. Estimation results using only (a) 10 and (b) 30 input test data that are identical to the quantity of output test data reflected in the previously: Both true and estimated distributions represent the same contour levels of the PDF, which are presented at 0.05 intervals from 0.02 to 0.5
Fig. 9. Estimated distributions of fatigue strength parameter using (a) 10 and (b) 30 output test data with 10 input test data; fatigue ductility parameter using (c) 10 and (d) 30 output test data with 10 input test data: Both true and estimated distributions represent the same contour levels of the PDF
Fig. 10. Corrected estimation results using only each input test data from the results in Fig. 9
Fig. 11. Cross-sectional view and configuration of the pedal feel simulator in the IDB system
Fig. 12. Measured data with input and output test for use in model calibration procedure: (a) A joint probability distribution identified from 21 Mooney-Rivlin material parameters obtained from coupon test; (b) A histogram of strain energies obtained through UTM test for the 100 feeling damper samples
Fig. 13. Estimation results of Mooney-Rivlin material parameters using (a) 10 and (b) 30 output test data with 3 input test data; similarly, estimation results using (c) 10 and (d) 30 output test data with 10 input test data: Both true and estimated distributions represent the same contour levels of the PDF
Fig. 14. Corrected estimation results using only each input test data from the results in Fig. 13
作者信息 | Authors
韩国 HL 万都公司 (HL Mando Co.)
韩国科学技术院 (Korea Advanced Institute of Science & Technology) 机械工程系
韩国科学技术院 (Korea Advanced Institute of Science & Technology) 机械工程系
Email: ikjin.lee@kaist.ac.kr
律梦泽 M.Z. Lyu | 编辑 (Ed)
P.D. Spanos | 审校 (Rev)
陈建兵 J.B. Chen | 审校 (Rev)
彭勇波 Y.B. Peng | 审校 (Rev)
