High Friction Surface Treatment on Bridge Safety Brian Porter/Rebecca Szymkowski- WisDOT Andrea Bill- UW-Madison TOPS Lab
Objectives Weather in WI can be harsh Bridges can be problematic in inclement weather 2012: 24,439 SIW Crashes (22%) 96 deaths in SIW crashes (16%) 8,062 injuries in SIW crashes (20%) Past projects on bridge decks to improve friction as part of HSIP There is a need to identify and rank bridges throughout the state for future improvement based on crash analysis Identify predictors for weather related crashes to understand why crashes are happening and what bridges would be best served by countermeasures
HFST Reduce crashes, injuries and fatalities Relatively low cost compared to geometric improvements Negligible environmental impacts and minimal traffic impacts
Nebraska Ranking Project Rank bridges to receive countermeasures Benefit-Cost Method Cost Effectiveness Utility Index Composite Programming Preferred Method
WisDOT STN dataset and FWHA NBI dataset Issues: Data Collection Bridge Data No. A&ribute Reason Duplicates: merging 2 lists, start/end pts 1 Construc)on Year Condi)on could affect opera)ons 2 Ra)ng Condi)on could affect opera)ons 3 Deck Material Materials could be affected by weather differently 4 Deck Width Size could impact pavement condi)ons 5 Deck Length Size could impact pavement condi)ons 6 Deck Area Size could impact pavement condi)ons 7 Number of Travel Direc)ons Traffic pagerns could affect crash type/severity 8 Lanes on Structure Traffic pagerns could affect crash type/severity 9 Over Roadway/Railroad/Water Surrounding area could impact pavement condi)ons 10 Lanes Under Structure Surrounding area could impact pavement condi)ons 11 Average Daily Traffic Traffic pagerns could affect crash type/severity 12 Bridge Railing Condi)on could affect crash type/severity 13 Transi)on Railing Condi)on could affect crash type/severity 14 Approach Railing Condi)on could affect crash type/severity 15 Approach Railing End Condi)on could affect crash type/severity
Data Collection Crash Data 5 years of crash data from WisTransPortal SIW crashes vs weather crashes Linked to bridges in ArcGIS 250 buffer zone on either side of bridge
Dataset
Wisconsin s Ranking Project Based on NDOR Benefit-Cost Method Assumed 80% reduction in weather related crashes Assumed $40/square yard to implement (HFST) Based on small bridge area Benefit- Cost Ra?o Number of Bridges >200 2 40-200 1 30-40 2 20-30 10 10-20 7 5-10 10 2-5 69 1-2 139 0.5-1 201 0-0.5 1884 0 3038
Final Dataset 4,537 bridges were evaluated Duplicates were removed Crash locations identified Either removed or added together Bridges with incorrect data 0 for ADT 0 for directions of travel carried 21,667 crashes & 11,112 weather crashes ~1,700 bridges with no crashes ~1,900 with no weather crashes
Directions of travel 1: 43% 2: 57% Facility Spanned Water: 58% Highway (Road): 36% Railroad: 6% Railings: None: 20% 1: 11% 2: 7% 3: 62% Final Dataset
Final Dataset Mean Median Mode Standard Devia?on Min Max Bridge Deck Width 47.33 43 43 19.28 12 373 Bridge Length (P.) 182.96 132 33 240.45 20 3256 Bridge Sufficiency Ra?ng 86.94 90.60 98 12.52 6 100 Bridge Ra?ng Score 93.40 95.00 100 7.63 25 100 Lanes on the Structure 2.37 2.00 2.88 1 9 Average Daily Traffic (ADT) 13565 7900 1100 16762.44 35 153180 Total Crashes 4.78 1.00 0 11.24 0 185 Total Weather Crashes 2.45 1.00 0 5.37 0 101 Total SIW Crashes 1.69 0.00 0 3.61 0 75 Total KA Crashes 0.13 0.00 0 0.42 0 7 Total Weather KA Crashes 0.05 0.00 0 0.25 0 3 Total SIW KA Crashes 0.04 0.00 0 0.23 0 3
Model Selection Poisson Model over dispersed Dispersion parameter 3.54 Zero-Inflated not a disproportionate amount of bridges with no crashes Negative Binomial Good for large dataset with large variance Backward Selection Method: Full Model Forward Selection Method: Null Model, Likelihood Ratios
Forward Selection Null Model Covariates added based on Likelihood Ratio Test Itera?on Covariate Likelihood Ra?o 1 Log (ADT) 1557.46 2 Over Road/Water 541.46 3 Log (Bridge Length) 317.29 4 Direc)ons of Travel 37.01 5 Interac)on of Log(ADT) and Direc)ons of Travel 49.18 6 Construc)on Year 30.79 7 Lanes Under Bridge 15.96 8 Bridge Ra)ng 8.57 Iteration
Final Model C=Exp[ b 0 + b 1 x 1 + b 2 x 2 + b 3 x 3 + b 4 x 4 + b 5 x 5 + b 6 x 6 + b 7 x 7 + b 8 Confidence Interval Covariate Es?mate Pr(>LR) x 8 + b 9 x 9 ] 2.5% 97.5% (Intercept) 8.7971 3.896 13.684 - - log(adt) (x 1 ) 0.6150 0.552 0.678 <2e- 16 *** overroadn (x 2 ) - 0.5749-0.716-0.434 1.305e- 14 *** overroadr (x 3 ) - 0.3821-0.564-0.200 1.305e- 14 *** log(length) (x 4 ) 0.4211 0.372 0.471 <2.2e- 16 *** Dir2 (x 5 ) - 3.5904-4.418-2.765 <2e- 16 *** Year (x 6 ) - 0.0082-0.011-0.006 1.367e- 10 *** laneund (x 7 ) 0.0693 0.036 0.103 3.922e- 05 *** RTG (x 8 ) 0.0061 0.002 0.010 0.003322 ** log(adt):dir2 (x 9 ) 0.3482 0.262 0.434 1.373e- 15 ***
Top 25 Bridges
Ranking Order Predicted number of weather related crashes for 5 years Percent Error not meaningful Many bridges with no crashes Crash numbers on bridges vary Not as accurate for interchanges
Applications Use ranking list to select bridges for countermeasure Combine with understanding of Human Factors to improve safety Affect design of new bridges and maintenance/ construction on existing bridges
Combine with B-C Ratio List Find bridges with high predicted number of crashes, and select based on B-C Ratio Bridges on Top 100 Predicted Crashes with B-C > 0.5 No. Bridge ID Predicted Crashes Benefit- Cost Ra?o 1 B130263 25.069 0.954 2 B510023 19.686 1.282 3 B130008 19.204 0.951 4 B510022 17.138 0.589 5 B400223 14.462 1.391 6 B400224 14.453 1.393 7 B300023 14.246 0.961 8 B670052 13.305 0.741 9 B670053 13.083 0.618 10 B050093 13.038 1.045 11 B050092 13.038 0.885
Combine with B-C Ratio List Find bridges with high B-C Ratio, and select based on high predicted number No. Bridge ID of crashes Benefit- Cost Ra?o Predicted Crashes 1 B130607 32.40 1.54 2 B320150 27.77 1.77 3 B320149 27.77 1.77 4 B440020 27.65 3.21 5 B440021 27.65 3.21 6 B690018 26.81 2.02 7 B690017 26.81 1.93 8 B640024 24.48 1.85 9 B640023 24.48 1.85 10 B510026 19.69 8.60 11 B510025 19.31 7.51 12 B530064 18.93 3.08 13 B530063 17.16 2.97 14 B130463 10.78 6.43 15 B660001 10.75 8.95 16 B170051 9.41 1.20 17 B660017 8.95 0.94 18 B110044 6.72 1.16 19 B110015 5.66 4.52 20 B110016 5.51 5.01 21 B670058 5.50 5.67 22 B170018 5.19 4.56 Wisconsin Traffic Operations and Safety 23 Laboratory B170019 Department 5.16 of Civil and Environmental 4.62 Engineering 24 B410059 5.10 1.59 Bridges with B-C > 5 and at least 1 Predicted Crash
Conclusions Overall, model can accurately predict the number of weather related crashes on bridges Not as accurate for interchange ramps ADT, directions of travel, facility spanned and number of lanes, construction year/condition, and length No overlap with existing ranking methodology to be expected
Future Research Additional characteristics Speed Heavy Vehicles Bridge Geometry Interchange/Functional Classification Weather patterns linked to bridges Comparison with other rankings Weather related crash rates per million vehicles Proportion of weather related crashes to total crashes
QUESTIONS?