1 / 28

Chanyoung Park Raphael T. Haftka

Paper Helicopter Project. Chanyoung Park Raphael T. Haftka. Problem1: Conservative estimate of the fall time. Estimating the 5 th percentile of the fall time of one helicopter Estimating the 5 th percentile to compensate the variability in the fall time ( aleatory uncertainty)

liseli
Download Presentation

Chanyoung Park Raphael T. Haftka

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Paper Helicopter Project Chanyoung Park Raphael T. Haftka

  2. Problem1: Conservative estimate of the fall time • Estimating the 5th percentile of the fall time of one helicopter • Estimating the 5th percentile to compensate the variability in the fall time(aleatory uncertainty) • The sampling error (epistemic uncertainty) • Estimating the sampling uncertaintyin the mean and the STD • Obtaining a distribution of the 5th percentile • Taking the 5th percentile of the 5thpercentile distribution to compensatethe sampling error mt,P st,P Sampling tP Sampling t0.05,P

  3. Problem1: Conservative estimate of the fall time • Estimating the 5th percentile of the fall time of first helicopter (mean 3.78, std 0.37) • 100,000 5th percentiles of fall time • Helicopter 1 of the dataset 3 • Height: 148.5 in • 2.88 (sec) is the 5th percentile of the histogram(a conservative estimate ofthe 5th percentile of the fall time for 95% confidence)

  4. Problem2: Predicted variability using prior • Calculating predicted variability in the fall time • We assume that the variability in the fall time is caused by the variability in the CD • The variability in the fall time is predicted using the computational model (quadratic model) and the distribution of the CD • The prior distribution represents our initial guess for the distribution of the CD Height at time t where Steady state speed

  5. Problem2: Comparing predicted variability and observed variability using prior • Area metric with the prior • Data set 3

  6. Problem3: Calibration: Posterior distribution of mean and standard deviation • Estimating parameters of the CD distribution • We assume that CD of each helicopter follows the normal distribution • The parameters, CD and σtest are estimated using 10 data • Posterior distribution is obtained based on 10 fall time data • Non informative distribution is used for the standard deviation After 1 update After 5 updates After 10 updates

  7. Problem3: Comparing predicted variability using posterior and observed variability • Comparing MLE and sampling statistics • MCMC sampling • 10,000 pairs of the CD and the STD of CD are generated using Metropolis-Hastings algorithm • An independent bivariate normal distribution is used as a proposal distribution • MLE of the posterior distribution is used as a starting point

  8. Problem3: Comparing predicted variability using posterior and observed variability • Handling the epistemic uncertainty due to finite sample • How to handle epistemic uncertainty in the CD and the test standard deviation estimates? • Comparing the posterior predictive distribution of the fall time and the empirical CDF of tests (combining epistemic and aleatory uncertainties) • Using p-box with 95% confidence interval of epistemic uncertainty (separating epistemic and alreatory uncertainties)

  9. Problem3: Comparing predicted variability using posterior and observed variability • Area metric of the posterior predictive distribution of CD

  10. Problem3: Comparing predicted variability using posterior and observed variability • Area metric of the posterior predictive distribution of CD

  11. Problem4: Predictive validation for the same height and different weight • Area metric of the posterior predictive distribution of CD

  12. Problem4: Predictive validation for the same height and different weight • Area metric of the distribution of CD with p-box

  13. Problem6: Predictive validation for different height and the same weight • Area metric of the posterior predictive distribution of CD

  14. Problem6: Predictive validation for different height and the same weight • Area metric of the distribution of CD with p-box

  15. Problem5: Linear model • Area metric with the prior • Data set 3

  16. Comparison to predictive validation • Area metric of the posterior predictive distribution of CD • The predictive validation with the linear model is not as successful as that with the quadratic model • Area metric with p-box • Area metric with p-box tries to capture the extreme discrepancy between the predicted variability and the observed variability

  17. Problem5: Linear model • Area metric of the posterior predictive distribution of CD

  18. Problem5: Linear model • Area metric of the distribution of CD with p-box

  19. Problem5: Linear model with one clip • Area metric of the posterior predictive distribution of CD

  20. Problem5: Linear model with one clip • Area metric of the distribution of CD with p-box

  21. Comparison between quadratic and linear models • Area metric of the posterior predictive distribution of CD • Area metric of the distribution of CD with p-box

  22. Concluding remarks • Predictive validation for both quadratic and linear models • The predictive validation for different mass is a partially success • The predictive validation for different height is a success but the assumption of constant CD is not clearly proven • Comparison between models • Cannot conclude • Overall • Reason for the differences in the area metric is not clear • The effect of the manufacturing uncertainty is significant (i.e. very different area metrics for the same test condition)

  23. Kaitlin Harris, VVUQ Fall 2013 • Comments: • Chose to maintain uniform distribution for calibration parameter based on histogram results (vs normal) • Conclusions: • Best models: calibrated quadratic at both heights and calibrated linear with 2 paper clips • Worst models: un-calibrated linear with 2 paper clips and un-calibrated linear with 1 clip for data set 1

  24. Validation of analytical model used to predict fall time for Paper HelicopterBy Nikhil Londhe *Calibrated Analytical Model is validated to represent experimental data *Quadratic dependence is valid assumption between drag force and speed *For given difference in fall height, Cd can be assumed to be constant

  25. Course Project: Validation of Drag Coefficient -Yiming Zhang Validation based on 1 set of data: Validation based on 3 set of data: Quadratic Dependence 2 clips Prior VS. Posterior Dist. Prior Area Metric:0.4920 Prior Area Metric:0.4042 Post Area Metric:0.2 Post Area Metric:0.1695 95% Confidence 0.77 95% Confidence 0.83 Quadratic Dependence 1 clip Predictive Validation. Validation Area Metric:0.1267 Validation Area Metric:0.1923 If just use this set to calculate posterior.95% Confidence 0.86 If just use this set to calculate posterior.95% Confidence 0.82 Linear Dependence 2 clips & 1clip Prior VS. Posterior Dist. Summary: 2 clips: 1 clip: 2 clips: Summary: 1 clip: Posterior Area Metric:0.1987 Validation Area Metric:0.1960 Validation Area Metric: 0.1253 Posterior Area Metric:0.1729 Seems reasonable, but not accurate Two confi. interval of Cd don’t coincide. Linear dependence seems inaccurate. 95% Confidence 0.8 95% Confidence 0.88 95% Confidence 0.8 95% Confidence 0.91 Quadratic Dependence 2 clips Different height. Validation Area Metric:0.1394 Validation Area Metric:0.0970 If just use this set to calculate posterior.95% Confidence 0.76 If just use this set to calculate posterior.95% Confidence 0.79 Summary: (1) Quadratic dependence seems accurate using one set; Quadratic and linear dependence both don’t match well using 3 sets; (2) SRQ is required to be fall time. If use Cd as SRQ, comparison could be more consistent and clear; (3) 0.8 seems a reasonable estimation of Cd. This estimation would be more accurate while introducing more sets of data.

  26. Backup Slides

  27. Problems • Problem1: Conservative estimate of the fall time • Problem2: Comparing predicted variability and observed variability using prior • Problem3: Comparing predicted variability and observed variability using posterior • Problem4: Predictive validation for the same height and different weight • Problem5: Comparing the quadratic and linear models • Problem6: Predictive validation for different height and the same weight (proving the assumption of constant CD)

  28. Problem1: Conservative estimate of the fall time • Estimating the 5th percentile of the fall time of one helicopter • Since fall time follows a normal distribution, estimating the 5th percentile is based on estimating the mean and standard deviation (STD) of the fall time distribution • The mean and STD are estimated based on 10 samples • There is epistemic uncertainty in the estimated mean and STD due to a finite number of samples • To compensate the epistemic uncertainty, a conservative measure to compensate the epistemic uncertainty is required • Estimating the 5th percentile with 95% confidence level

More Related