1 / 33

Implementation of WIM Data Analysis and Simulated Scenarios

A systematic approach to identify calibration issues in WIM systems through statistical analysis and simulated scenarios. Overcome limitations and improve reliability of WIM data.

kennethc
Download Presentation

Implementation of WIM Data Analysis and Simulated Scenarios

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. Implementation of WIM Data Analysis and Simulated Scenarios 12/15/2014

  2. Objective • Limitations of current practice • Fails to identify the time point when WIM system went out of calibration • Reliability of WIM data is questionable • Limited resources to perform frequent on-site calibration checks • Goal • Systematic approach to identify time point when WIM system went out of calibration • Estimate the bias in WIM sensor

  3. WIM : Sample Class 9 GVW

  4. EM Fitting of a 3-Component Mixture Model WIM station #37, lane 1, on 05/15/2012

  5. WIM# 37 Lane 1, Vehicle Class 9 EM Estimated Daily Average Fully Loaded GVW

  6. Statistical Process Control Technique • Original CUSUM Approach (Page, 1954) • Decision based Interval (DI) CUSUM(Hawkins,1998) • Upper CUSUM • Lower CUSUM

  7. DI based CUSUM • Upward shift: • Downward shift: • Algorithm parameters: decision interval reference value • Choice of h & k: false positive tolerance

  8. Limitations • Independent assumption • Autocorrelation function (ACF) plot from station# 29 • Presence of autocorrelation results in higher number of false alarms.

  9. WIM Sensor Drift Analysis: Case Study • Data from station#29, Lane 1 Average daily EM estimates of fully loaded class 9 vehicles • Data partitioned into 2 subsets: • Learning Set • Testing Set

  10. GVW for Average Daily Fully Loaded Trucks

  11. Fitting Learning Sample AR(1) Model: where, ~N(0,) Estimated Parameters:

  12. Prediction on Testing Sample AR(1) Residuals: Upward CUSUM on standardized residuals: Estimate of shift in Mean level:

  13. Prediction on Testing Sample AR(1) Residuals: Upward CUSUM on standardized residuals: Estimate of shift in Mean level:

  14. CUSUM based DI Estimated shift, =5.3 kips (6% of average daily GVW) Time of the shift: 05/10/2011 Validation: MnDOT’s test run on 06/08/2011: shift by 6 kips

  15. Implementation Guidelines

  16. Software Implementation • Windows 7 & .NET framework 4 • Microsoft .NET based application • Integrated with R statistical software through R.NET control (open source)

  17. Flowchart of Data Analyst Software 1. EM Analysis EM Analysis Set Path Select Station, Lane, Year, Month Load GVW9 Data Process EM Output EM Data to File EM Files CalibrationFiles 2. CUSUM Analysis Select Station, Lane, Date Period Load Calibration Log File Set Path Load Processed EM Data CUSUM Analysis Model Testing Sample Model Learning Sample CUSUM Analysis Estimate Sensor Drift Result Output

  18. WIM Data Analyst Interface

  19. WIM Data Analyst Interface

  20. WIM Data Analyst Interface

  21. WIM Data Analyst Interface

  22. Instability in WIM Data Final calibration: 5%

  23. Instability in WIM Data Final calibration change= 3% Final calibration change= - 4%

  24. Using Simulated Sample

  25. Simulated Scenario #1 Sensor Shift ( =5 kips T= 150; Learning Sample, Tl =60 Stationarity test: Data is covariance-stationary

  26. Splitting the Testing Sample Data is splitted into 30 days (T1=30) Output from R: No upward shift found in WIM sensor No downward shift found in WIM sensor

  27. 2nd part of Testing Sample Output from R: No upward shift found in WIM sensor WIM sensor shifted by -3.94 kips (~5%) on 93

  28. Consistency of WIM Sensor Shift Updated Mean level: Output from R: WIM sensor bias of -3.94 kips (~5%) at index# 93 is consistent

  29. Simulated Scenario #2 T= 150; Learning Sample, Tl =60 Stationarity test: Data is covariance-stationary

  30. Splitting the Testing Sample Output from R: No upward shift found in WIM sensor No downward shift found in WIM sensor Data is splitted into 30 days (T1=30)

  31. 2nd part of Testing Sample Output from R: No upward shift found in WIM sensor WIM sensor shifted by -3.79 (kips) at index= 93

  32. Consistency of WIM Sensor Shift Updated Mean level: Output from R: Warning message: Estimated WIM sensor bias is not consistent

  33. Next Step • Implementation stage • Incorporate the WIM sensor bias consistency check • Test WIM analysis for new data • Task 6 – Adjust WIM data • Task 7 – Draft final report & user manual • Task 8 – Final report

More Related