Proposed Modifications to AASHTO Culvert Load Rating Specifications (2019)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 C H A P T E R 2 Research Approach The objective was accomplished by  Surveying the DOTs for current culvert practices and issues related to culvert rating.  Contacting states via telephone who agreed to talk to us.  Developing and executing full scale field tests on seven culverts with the aid of four different state agencies (Maryland, Massachusetts, Ohio, and Pennsylvania).  Developing and executing a full analysis of the subject culverts in both 3D and 2D analysis.  Identifying areas of the MBE and LRFD Specifications for improvement and recommendations for updates to those specifications based on the field testing and analysis results. The literature survey was conducted in the first phase of the project to ascertain the current state of the culvert rating specifications. This survey included an online survey with an optional follow-up interview at the agency’s request. Some of the results yielded by the survey and interviews were:  Concrete culverts are used extensively, followed by steel corrugated culverts. Only one state indicated that they rated thermoplastic culverts and they are not reporting rating issues for those types of structures.  Reinforced concrete culverts are not rating well but don’t show physical signs of distress.  Studies that included input files for culvert analysis that could be utilized by this study.  Other research efforts including those from California, Ohio, and Pennsylvania which were utilized in this research. Based on the literature survey and subsequent interaction with the project panel, a sample testing plan and target field testing matrix was developed. Upon approval by the project panel, the effort to locate target culverts was conducted by contacting the state DOTs. This process yielded a field testing plan that was developed in tandem with an analytical plan that would be used to evaluate the current processes and specifications used for culvert rating and evaluation. In all, seven culverts were field tested and analyzed in 3D using a FEM software package and in 2D using the CANDE (Culvert ANalysis and Design) software updated under NCRHP Project 15-28 (Mlynarski, M., M. G. Katona, and T. J. McGrath. NCHRP Report 619: Modernize and Upgrade CANDE for Analysis and LRFD Design of Buried Structures. Transportation Research Board of the National Academies, Washington, D.C., 2008.). The culverts selected and the process for selection are described in subsequent sections of this chapter. In addition to the field testing plan, an analysis testing plan was developed to model the selected field testing results and to review the current specifications to determine where improvements could be made to the rating process. This process for the analysis and the tools used is described in detail in later sections of this chapter. The results of those recommendations are part of this research and include modifications to the MBE and LRFD Specifications. The process for the specification review and recommendations is provided in more detail in Chapter 3. The recommendations for the specification changes are provided in Chapter 4 and include:  Recommendations for culvert live load distribution.  Recommendations for non-rectangular culverts.  A study on the effects of pavement in culvert rating and recommendations for the use of an analytical tool developed on this project (CANDE Tool Box and CANDE) for analyzing culverts with pavement elements. Recommendations were also made for implementing the effects of pavement in other culvert programs.  Changes to the shear capacity calculations for reinforced concrete box culverts.  Changes to the live load surcharge loads (now described as approaching wheel load).

4  The effects of fill depth on live load.  The effects of haunches on reinforced concrete culvert analysis.  Sections for concrete, metal, and plastic culverts. Development of Field Test Program The process for selecting the field tested culverts is described in this section. It includes: – The culverts selected. – The selection process. – A description of the culverts selected. Culverts Selected This section provides a description of the culvert selection process, some information about the culverts selected, and a description of the models that were prepared (both 3D and 2D). The culverts were selected from four states: Pennsylvania, Ohio, Massachusetts, and Maryland. The matrix of culvert types was developed early in the process and presented to the panel for review with each culvert type having multiple candidates. The final list of models for analysis and field testing is provided in Table 1. A more detailed description of the culverts is provided in subsequent sections of this document. The detailed field testing data is provided in Appendix F. The following sections provide a brief description of the selection process. Review and Selection Process  Several states agreed to cooperate with the field testing portion of the research effort. These are Pennsylvania, Maryland, Illinois, Ohio, and Massachusetts. Ultimately, seven culverts were chosen for field testing from Pennsylvania (2 state, 2 county), Ohio (1), Maryland (1), and Massachusetts (1). Selection Criteria During the selection process of the culverts that would ultimately be used for field testing and analysis, the following criteria was considered:  Relatively low ADT to minimize disruption to traffic and to simplify access and instrumentation.  Close proximity to the members of the research team that are performing the instrumentation. This reduced some of the travel cost and set up for the testing.  Availability of detailed drawings – unlike with non-culvert structures, DOTs tend to retain less detailed plan information on culverts. Furthermore, DOT design plans tend to be less detailed for culvert structures than for other types of bridges because some of the design and detail development is left for the fabrication stage. As such, for most structure types included in this research effort, both design and shop drawings of the candidate structures was required to have sufficient detail for analytical modeling and field testing. This was a significant limiting factor on the availability of suitable candidates for the research.  Review of potential paving schedules (if available) for reviewing the effects of existing pavement versus new pavement. Vendors Representatives from the vendors BigR Bridge and CONTECH were contacted to assist in the identification of suitable candidate structures. This was important particularly because of the previously mentioned requirement that shop drawings of the candidate structures be obtained and in many cases these drawings were found to be absent from the DOT files. Furthermore, for some of the selected structures,

5 CONTECH and BigR Bridge were able to provide design calculation and structure modeling information helpful to the research effort. Pennsylvania (DOT and County) PennDOT provided an inventory of all culverts in the state, which was then reviewed for culverts that meet the criteria of the matrix provided in Table 1 and the selection criteria provided in the previous section. The culverts selected from Pennsylvania were:  Model 1-Candidate 1, Single Cell Reinforced Concrete Box Culvert – longer span  Model 3-Candidate 1, Precast Single Cell Reinforced Concrete Box Culvert – shorter span  Model 5- Candidate 1, Metal Corrugated Arch  Model 6-Candidate 2, Metal Corrugated Box Arch Maryland DOT Multiple structure types were solicited in Maryland including concrete arch and box structures. However, after reviewing the availability of structures provided by the DOT and with consideration of the structures already identified in other states, only concrete boxes in Maryland were selected as test candidates. The culvert selected from Maryland was:  Model 2-Candidate 1 – Twin Cell Reinforced Concrete Box Culvert Ohio DOT For Ohio, CONTECH helped the research team identify several suitable test candidates within the state of Ohio. Ultimately one field test was scheduled for a concrete arch culvert in Ohio  Model 4-Candidate 1-Reinforced concrete arch culvert Massachusetts DOT MASSDOT and BigR Bridge provided information that led to the effort of being able to field test a longer span metal corrugated culvert. This structure, in Attleboro, Massachusetts, was selected to broaden the range of structure types being included in analytical and field testing and being new construction, allowed for load testing both before and after paving. The model selected was:  Model 7-Candidate 1 – Deep Corrugate Metal Arch Culvert Illinois DOT While Illinois DOT offered to provide support for field testing culverts, the above states provided a good selection of culverts that met the criteria of the developed testing matrix but were closer geographically and thus more economical for the research team.

6 Table 1 – Field Testing Matrix (Selected culverts are highlighted) Model Structure Type Desired Features Span (ft) Depth (ft) Candidate 1/ State/ Candidate 2/ State 1 Reinforced concrete box - single cell Precast 16-20 Field Test Depth + 2 additional PA BRKEY: 20274,25’x7’-6” Juniata County PA PA BRKEY 46704, 18’ Span 2 Reinforced concrete box - multiple cell Cast-in- place 10-12/cell Field Test Depth + 2 additional MD 329500: 10x7 2-cell 15degree skew, 40’ width, 1’ cover MD 301700: 12’x12’ 2-cell 29 degree skew, 40’ width, min cover 3 Reinforced concrete box New – precast >10 Field Test Depth + 2 additional PA BRKEY 48389 Single 14’-cell, 80 degree skew PA BRKEY 9649 2-16’ cells 90 degree skew 4 Three- sided arch top concrete CONSPAN type 24-36 0-2 36’x12’ (2’-6 cover) Perry County OH 24’x10’ (2’ cover) New Albany, OH 5 Metal arch 6x2 corrugation 30 -35 AASHTO min. or less 31’-7x10’ (3’ min cover) Willshire Estates Dauphin Co, PA Toledo OH Arch 30’x9’-6 (3’ cover) 6 Metal box culvert 25 1.4 Kiwanas Rd, 20’-1”x6’-6” (1’- 4”cover) Orange Twp, Dauphin Co, PA Sleepy Hollow (Shermans’ Dale Perry County PA), 19’0x6’-1 (2’- 6”cover) 7 Deep corrugated metal culvert 5.5 to 6 in. deep corrugation 40 Minimum available 56’6” or 40’ New Constr Box, Attleboro, MA

7 Executing the Field Testing Program The field testing program drafted in the first phase and finalized in the second phase of this project was implemented. The following sections provide details of the field testing program. The detailed testing plans are presented in Appendix F of this report. Table 2 provides a summary table of the structures that were tested, the dates tested, the state, and the geographical latitude and longitude. Subsequent sections provide information regarding the loading, load lines, and general information regarding the testing. Table 2 – Field Testing Table Culvert Model Date Tested State/Agency Latitude, Longitude (click on link to open in a browser) Model 1 – Reinforced Concrete Box (Single Cell) 8/30/2017 Pennsylvania DOT, District 2-0 40.3538, - 77.6488 Model 2 – Reinforced Concrete Box (Twin Cell) 12/14/2017 Maryland, MDOT 39.411056, - 76.409278 Model 3 – Reinforced Concrete Box – (Single Cell-Precast) 11/7/20017 Pennsylvania DOT, District 9-0 40.0513, ‐78.9943   Model 4 – Reinforced Concreted Arch 4/17/2018 Ohio DOT, District 6 39.79083, ‐ 82.25111  Model 5 – Metal Arch 6/12/2018 Pennsylvania, Lower Paxton Township 40.300437, - 76.8018666 Model 6 – Metal Arch 5/3/2018 Pennsylvania, Carroll Township 40.359593, ‐ 77.142020  Model 7 – Metal Arch (Long span) – Deep Corrugation 5/1/2017 (no pavement) 6/1/2017 (with pavement) Massachusetts, Mass DOT 41.962574, ‐ 71.299294  Culvert Instrumentation The research team worked with state DOTs and municipalities to facilitate the instrumentation and load testing of the selected candidate culverts. In general, the cooperating agencies provided access to the culverts, loaded trucks for testing and traffic control during the load tests. The agencies and/or the research team also coordinated with the state police who weighed the load trucks with mobile scales. The research team developed the customized testing plans (see Appendix F) to perform the tests, installed the instrumentation gauges tied to a data acquisition system, set up the test loading positions and executed the tests. Typically, the instrumentation was set up a day or two before the load test took place and measures were taken to protect the installed gauges from the elements. A sample of the instrumentation for Model 1

8 is shown in Figure 1. In the case of Model 7 where testing was performed in two phases (before and after paving), there was a significant gap between the two phases as roadway construction above the culvert was completed. All load tests were performed without any other traffic on the culvert. Typically, multiple strain gauges were installed at each point of interest for redundancy. Details on the test procedures including instrumentation locations, loading positions and data collected can be found in Appendix F. Load lines were painted on the roadway surface to correspond to the gage locations. See Figure 2 for a sample used on Model 1 (M1C1). Following the load testing, a cursory review of the collected data was performed prior to releasing the loaded truck driver and traffic control. The collected data strain and deflection data was then brought into the office for post processing to prepare it for comparison to the FEM models and calibration. Figure 1 – M1C1 Culvert Instrumentation

9 Figure 2 – M1C1 Load Line on Roadway

10 Model 1 – Field Test - Reinforced Concrete Box – Single Cell (M1C1) This section summarizes general information related to the field testing for Model 1, a 25 foot single cell reinforced concrete box culvert with less than 1 foot of cover. The truck axle weight and configuration are shown in Figure 3. The culvert plan and typical section are provided in Figure 4. A sample load testing photograph is shown in Figure 5. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Pennsylvania Route: 3020 Latitude-Longitude: 40.3538, -77.6488 Figure 3 – Wheel Loads/Configurations for Model 1-Candidate 1 (M1C1)

11 Figure 4 – Model 1, Candidate 1 (M1C1) – Plan and Typical Section

12 Figure 5 – M1C1 Load Testing

13 Model 2 – Field Test - Reinforced Concrete Box –Twin Cell (M2C1) This section summarizes general information related to the field testing for Model 2, a twin cell (10 foot each) reinforced concrete box culvert with 1 foot of fill. The truck axle weight and configuration are shown in Figure 6. The culvert plan and typical section are provided in Figure 7 and instrumentation and load testing photographs are shown in Figure 8 and Figure 9. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Maryland DOT Route: SR 7 (Philadelphia RD) Latitude-Longitude: 39.411056, -76.409278 Figure 6 – Wheel Loads/Configurations/Testing Times for Model 2-Candidate 1 (M2C1)

14 Figure 7 – Model 2, Candidate 1 (M2C1) – Plan and Typical Section

15 Figure 8 – M2C1 Culvert Instrumentation Figure 9 – M2C1 Load Testing

16 Model 3 – Field Test - Reinforced Concrete Box – Single Cell – Precast (M3C1) This section summarizes general information related to the field testing for Model 3, a 12’ single cell, precast reinforced concrete box culvert with 1’ of fill. The truck axle weight and configuration are shown in Figure 10. The culvert plan and typical section are provided in Figure 11 and instrumentation and load testing photographs are shown in Figure 12 and Figure 13. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Pennsylvania Route: SR 281 Latitude-Longitude: 40.0513, -78.9943 Figure 10 – Wheel Loads/Configurations/Testing Times for Model 3-Candidate 1 (M3C1)

17 Figure 11 – Model 3, Candidate 1 (M3C1) – Plan and Typical Section

18 Figure 12 – M3C1 Culvert Instrumentation Figure 13 – M3C1 Load Testing

19 Model 4 – Field Test - Reinforced Concrete Arch (M4C1) This section summarizes general information related to the field testing for Model 4, a 36 foot precast reinforced concrete arch with 1 foot of fill. The truck axle weight and configuration are shown in Figure 14. The culvert plan and typical section are provided in Figure 15 and instrumentation and load testing photographs are shown in Figure 16 and Figure 17. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Ohio DOT Route: SR 669 Latitude-Longitude: 39.79083, -82.25111 Figure 14 – Wheel Loads/Configurations/Testing Times for Model 4-Candidate 1 (M4C1)

20 Figure 15 – Model 4, Candidate 1 (M4C1) – Plan and Typical Section

21 Figure 16 – M4C1 Culvert Instrumentation/ Load Line

22 Figure 17 – M4C1 Load Testing

23 Model 5 – Field Test - Metal Arch (M5C1) This section summarizes general information related to the field testing for Model 5, a 23 foot metal corrugated arch. The truck axle weight and configuration are shown in Figure 18. The culvert plan and typical section are provided in Figure 19 and instrumentation and load testing photographs are shown in Figure 20 and Figure 21. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Pennsylvania Route: Latitude-Longitude: 40.300437, -76.8018666 Figure 18 – Wheel Loads/Configurations for Model 5-Candidate 1 (M5C1)

24 Figure 19 – Model 5, Candidate 1 (M5C1) – Plan and Typical Section

25 Figure 20 – M5C1 Culvert Instrumentation/ Load Line

26 Figure 21 – M5C1 Load Testing

27 Model 6 – Field Test - Metal Box Arch (M6C2) This section summarizes general information related to the field testing for Model 6, a 19 foot metal corrugated box arch. The truck axle weight and configuration are shown in Figure 22. The culvert plan and typical section are provided in Figure 23 and instrumentation and load testing photographs are shown in Figure 24 and Figure 25. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Location: State: Pennsylvania, Carroll Twp Route: Sleepy Hollow Rd. Latitude-Longitude: 40.359593, -77.142020 Figure 22 – Wheel Loads/Configurations for Model 6-Candidate 1 (M6C2)

28 Figure 23 – Model 6, Candidate 1 (M6C1) – Plan and Typical Section

29 Figure 24 – M6C2 Culvert Instrumentation

30 Figure 25 – M6C2 Load Testing/Load Line

31 Model 7 – Field Test - Metal Box Arch (M7C1) This section summarizes general information related to the field testing for Model 7, a 56.5 foot long span metal corrugated box arch with 2 foot of fill. The truck axle weight and configuration are shown in Figure 26. The culvert plan and typical section are provided in Figure 27 and instrumentation and load testing photographs are shown in Figure 28 and Figure 29. Additional information regarding the testing such as the overall test plan and data regarding the loading are provided in Appendix F. Since this culvert was under construction, it was tested in two phases; once without pavement and once with pavement. Location: State: Massachusetts Route: I-95 over North Avenue Latitude-Longitude: 41.962574, -71.299294 Figure 26 – Wheel Loads/Configurations for Model 7-Candidate 1 (M7C1)

32 Figure 27 – Model 7, Candidate 1 (M7C1) – Plan and Typical Section

33 Figure 28 – M7C1 Culvert Instrumentation

34 Figure 29 – M7C1 Load Testing/Load Line

35 Development of Analysis Testing Plan The analysis testing plan for this research project was developed concurrently with the field testing plan for this project. As mentioned in the field testing section, the ability to analyze the proposed structures was necessary to compare the analytical results with those of the field testing. As such, choosing structures with available plans, shop drawings, and/or existing models was a criteria for choosing the models. Software Used for the Analysis/Data Gathering Changes to software tools were performed on this research project to enable help with the analysis of the selected culverts, to refine models, and to determine the effects of proposed changes to the specifications. The CANDE software was originally developed in the 1980’s and was updated to include new analytical capabilities and a modern user interface under NCHRP Project 15-28 (Mlynarski, M., M. G. Katona, and T. J. McGrath. NCHRP Report 619: Modernize and Upgrade CANDE for Analysis and LRFD Design of Buried Structures. Transportation Research Board of the National Academies, Washington, D.C., 2008.). The AASHTOWare Bridge ® Design and Rating (BrDR) software contains a reinforced concrete culvert component that was utilized to not only test how the current rating specifications perform but also to compare those results with the proposed specifications that were a result of this research. This section provides a description of the software modifications and the functions of the changes with respect to this research project and the use for engineers beyond this project. These revisions include:  The CANDE Tool Box – This utility provides options to modify CANDE models and includes items as creating a refined model, modeling pavement elements, and rating culverts. A full description of the CANDE Tool Box Manual is provided in Appendix C. The revised source code will be a delivery with this project.  CANDE interface – changes were made to the CANDE interface to apply some rudimentary features to refine existing CANDE models. The revised source code will be a delivery with this project.  BrDR changes – regression data mining – with AASHTO’s permission, changes to BrDR regression tool were made to allow for additional data mining of the regression data produced by the software. Changes to the specification articles were also made to compare the current specifications to the proposed specification revisions. This source code will be available for future releases of the BrDR software. CANDE Tool Box/Development of the 2D CANDE Models A pre/post processor tool (CANDE Tool Box) was created for this project and includes the options listed below that helped build and analyze the models need for this project. A user manual for this software is included in Appendix C of this report. The five available options for the CANDE Tool Box are:  Option 1 – This option will develop a full level 3 mesh from a level 2 input file. Currently level 2 models in CANDE only produce a half mesh. The advantage of using a level 2 model is that minimal input is required to generate the model (e.g. basic model geometry). The full level 3 mesh (full finite element model) facilitates the live loading of the model. Option 2 – This option provides the automated insertion of a surface pavement on any FEM mesh. This is applicable to any existing level 3 input file that has a continuous horizontal soil surface. The added pavement layer is characterized by linear-elastic beam elements whose thickness and material properties are input by the user. The following is the input that is provided by the user. The input required is: – Pavement uniform thickness (default = 8.0 inches) – Young’s modulus (default = 200,000 psi) – Poisson ratio (default = 0.3) – Weight density (default = 140 lbs/ft3)

36 Option 3 – This option creates the boundary conditions to simulate a moving live load. It is applicable to any existing level 3 input file that has a continuous horizontal soil surface above the culvert wherein the soil surface may be paved or unpaved. Live loads are simulated by point-like strip forces applied as boundary conditions specified for specific nodes and load steps. An option is provided to proportionately distribute the live load to adjacent nodes when a live load axle falls between two nodes. The user specifies the desired truck type: HL93-design, HL93-tandem, or User-defined with up to ten axles and provides yes/no answers to a series of sub-options as well as information on the vehicle travel path. The final output of option 3 is a completely new input file with the same name as the original input file except with the preface “live”. The live load is distributed through RSL (Reduce Surface Load) or CLS (Continuous Load Spreading). These are defined in the CANDE Tool Box User Manual (See Appendix C). Option 4 – This option provides a permanent bandwidth minimization and is applicable to any existing level 3 input file wherein node numbering is not optimum, thereby creating long solution times or exceeding system storage capacity. The procedure to permanently minimize bandwidth is to redefine the nodal numbering, starting with node "1" in lower left corner of element located in the lower left corner of the mesh. Option 5 – This option automatically computes the load ratings factor (RF) from an existing CANDE output file and writes the final RF values along with all supporting information at the end of the CANDE output report. This option is applicable to any existing level 2 or level 3 output file dealing with live load analysis based on LRFD methodology. The above options are described in detail in the CANDE Tool Box Manual described in Appendix C of this report. Changes to the CANDE GUI and Steps for Creating CANDE Models The changes described in the CANDE Tool Box, along with some additional mesh generating capabilities, were used to generate more complex CANDE models from simple level 2 input. A brief description of the process is provided below for Model 1 – Candidate 1. A similar process was performed on other models with the results for pavement vs. no pavement discussed in Chapter 3 of this report.  Step 1– Generate the level 2 model Level 2 models in CANDE can be generated with minimal input. For a full definition of level 2 models, refer to the CANDE User Manual. In general, level 2 models in CANDE are a mirrored image of the model (i.e., only half the mesh is generated). They are models that can be easily generated with relatively few input commands by the user. See Figure 30.   Figure 30 – Level 2 Plot of Model 1 - Candidate 1

37 Step 2 – Generate the level 3 model from the level 2 model To facilitate a useful model for this project, the CANDE Tool Box converts a level 2 model into a level 3 model so that it can be analyzed for live load. For a full definition of level 3 models, refer to the CANDE User Manual. In general, level 3 models in CANDE are more complex, fully defined FEM models. They are more difficult to generate (many often with an external mesh generator), so providing an option to generate them from a simpler level 2 input is a helpful enhancement for the user. The level 3 model generated can then be used for non-symmetrical loading (such as live loading). See Figure 31 below. Figure 31 – CANDE Level 2 Model After Conversion to a Level 3 Model Step 3 – Create a more refined mesh using the CANDE UI tools Some tools have been added to the CANDE UI that facilitate the generation of CANDE mesh elements for dividing the generated CANDE mesh from Step 2, into a less-coarse mesh. The new options are shown in Figure 32. These options along with the CANDE Tool Box are project deliverables.

38 Figure 32 – CANDE Mesh Generation Options The options shown in Figure 32 allow for the subdivision of CANDE elements of the model generated in Step 2, to a more refined mesh. The outcome is shown in Figure 33. Figure 33 – Level 3 Converted to a Model with a More Refined Mesh Step 4 – Simulate moving a live load vehicle across the structure to produce a rating Using the CANDE Tool Box, the live load is simulated by moving a series of axle loads (which may be combined with an optional, user input lane loading) across the structure (see Figure 34). The stacked ‘circles’ shown in the figure are the CANDE graphics engine method of displaying boundary conditions. In this case, the circles at the top of the culvert are the live loads being placed (and later removed) from the

39 model. Once the analysis is performed, the results can be post-processed using the CANDE Tool Box to produce a rating factor. A sample of the CANDE rating factor table output is shown in Figure 35. Full rating tables for each member of the culvert are also available. Figure 34 – Simulating a Live Load Vehicle in CANDE Figure 35 – Sample CANDE Tool Box Rating Output Step 5 – Check the same model for varying fill depths The same model can be used with varying fill depths. The current model has a fill depth of 0.97 feet. Additional models of 2’, 5’, and 10’ were also generated. A model with the 5’ fill depth is shown in Figure 36.

40 Figure 36 – Model with 5’ Fill Depth Step 6 – Modify the model to include pavement elements The CANDE Tool Box has the capability to add pavement elements to the model, so that the same model can be analyzed with and without the effects of the pavement. Each model was tested using this option. In general, this process was used to help develop the CANDE models that represent the culverts that were field tested for this project. A more detailed description of the models and the backup for the development is provided in Chapter 3 and Appendix D of this report. BrDR Regression Testing and Data Mining The BrDR software provides the capability to produce data in a form that can be used for regression testing. For the purposes of this research, that regression testing was used primarily to view data for the same culvert with respect to fill depth changes and with respect to changes in the specification. Each regression test utility file (RTU) produced by BrDR contains data in a form that can be imported in a relational database and analyzed with respect to a revised subsequent version of the software. This allowed the RT to compare the current specification articles with the articles proposed as a result of this research. In the BrDR RTU files, report IDs are used to identify each piece of output produced by the BrDR engine (spec output and analysis output). This is based on the NCHRP 12-50 process (Michael Baker Jr., Inc., Bridgetech, Inc., Modjeski and Masters, Inc., and Paul D. Thompson. NCHRP Report 485: Bridge Software—Validation Guidelines and Examples. Transportation Research Board of the National Academies, Washington, D.C., 2003.). BrDR report IDs used in the BrDR RTU were examined and additional report IDs were added. In addition, some inconsistencies in labeling the RTU records for culverts were discovered and reported. These are were addressed for the next release of BrDR (6.8.3). The regression database was reviewed and a data mining option was added to the regression testing tool (see Figure 37). Using the regression database, critical rating factors for shear and moment were queried and results for other report IDs (capacity, loads, spec results, etc.) can be obtained at the same locations. This enabled the RT to extract critical data for multiple culverts at varying fill depths at critical moment and shear rating locations. The results of the data mining are depicted graphically in Figure 38.

41 Figure 37 – Getting Critical Rating Information from the BrDR Regression Data In addition, extraction of the RTU data allowed for the review of the effects on specifications before and after recommendations. A simple data mining function was added to the regression test utility that allows for data to be presented in different forms (e.g. find the minimum rating factor and corresponding values related to that factor; DL, LL, Capacity). A rudimentary interface for the data mining tool is shown in. An example of the graphs that can be produced with the tool is provided in Figure 38. The figure displays the rating results of one of the Caltrans culvert models at different fill depths (right vertical axis), against the rating parameters (Capacity, DL, LL) on the left vertical axis. The bottom axis is the fill depth. Using the regression test utility and the regression data generated by BrDR, these graphs can be generated and exported to a Word document for quick review.

42 Figure 38 – Data Mining Using the BrDR Regression Tool – Getting Critical Data and Corresponding Values The RT used BrR models provided by Caltrans during the survey portion of this research and obtained a final set of the culvert models that were used to regression test the specification changes described in Chapters 3 and 4. The regression tool with the data mining option was during this research for helping to identify areas of the specification for potential review. Samples for data mining performed on the Caltrans Double Box culverts are provided in Appendix J. Development of the 3D Models A document was developed to establish the parameters and approach to the 3D modeling effort and to achieve agreement among the research team on key aspects of the 3D modeling (See Appendix M). This document was used to develop each of the 3D culvert models summarized in Appendix E. Also included in Appendix E is sample output for the completed 3D models.