Performance of Longitudinal Barriers on Curved, Superelevated Roadway Sections (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.

58 Crash Simulation Analysis of Impacts into Longitudinal Barriers on CSRS 5.1 Introduction This research revealed that there has been little testing or analyses of longitudinal barriers used on CSRS for reasons ranging from the difficulty of testing barriers on such road- ways, limited capabilities to employ other approaches, and limited details in crash records that make it hard to isolate incidents involving this specific type of barrier deployment. There have been strides in analyzing barrier effectiveness in varying deployment scenarios using simulation. This research proposed to employ crash simulations based on FE modeling to analyze the performance of longitudinal barriers on CSRS. This approach was offered because crash simulation capabilities have evolved to a point where viable insights can be derived considering the full range of condi- tions associated with impacting a barrier on a CSRS. This chapter describes the FE modeling and crash simulation that was employed. Crash simulation results allow determination of effective performance envelopes that can serve as the basis for enhanc- ing or creating new guidance for highway and barrier design and deployment on CSRS. The effectiveness of simulation tools allows many combinations of features and impact conditions to be investigated economically. Simulation has become a common means to understand barrier perfor- mance without the cost of multiple, full-scale crash tests. Computer simulations also yield significantly more data than can be extracted from the full-scale crash tests. The simulation results include displacements, velocities, and accelerations of every point on the vehicle during impacts with roadside hardware. The deformations and energies absorbed by each component of the vehicle and the roadside hardware under various impact conditions are also computed and provided in the simulation results. Such information is useful for identify- ing critical weaknesses in the design and for providing a bet- ter understanding of the influences of CSRS conditions and placement features on roadside hardware safety performance. 5.2 Background For more than 20 years, the FHWA has promoted the use of crash simulations based on FE models as a means to develop innovative designs and to evaluate their performance. Doing so requires FE models of vehicles and the roadside hardware. FE models have been developed to describe the vehicle and test articles as a collection of elements that reflect the geo- metry of the items, the nature of connections between adja- cent elements, the characteristics of the element materials, and properties associated with the relationships between elements (e.g., joints, fracture mechanics). FE models for vehicles are developed by reverse engineering. For hardware, the geometries of the components are used to define elements. Over the years, the vehicle models have become more detailed and complete (e.g., functional representation of sus- pension systems, interior modeling, and air bag capabilities). This has allowed a broader range of applications. The more recent generation of vehicle models consists of more than 1 mil- lion elements when all the interior components are included. Because all structural components are explicitly modeled, these detailed models can be used to study different impact scenarios including frontal, side, rear, oblique, and roof impacts. The objectives of the simulation efforts were to (1) develop, adapt, and validate FE models for longitudinal barriers typi- cally deployed on CSRS, (2) analyze the effects of curvature, superelevation, shoulder design, and roadside conditions on the MASH performance of the barriers, and (3) use the simu- lation results to formulate guidance of improved practice for the selection and placement of barriers for such situations. The efforts focused on a set of typical types of CSRS. 5.3 FE Modeling and Crash Simulation Analyses Finite element analysis (FEA) involves the use of FE models of vehicles and barriers in crash simulations. The simulations analyze the physics of each discrete element of C H A P T E R 5

59 the models for small increments of time (e.g., microseconds) over the duration of an impact event (e.g., the vehicle hitting the barrier). Because elements of the vehicle and the barrier will contact each other and the forces will cause the elements to deform, move, or fail in accordance with the defined material properties and nature of connections between elements, it is possible to replicate the crash dynamics that provide an indication of a barrier’s performance. While such simulations do take a considerable amount of time to go through all the elements over the duration of the crash event, they are more economical than using crash testing. Further, the details of the elements in the model and principles of physics allow more detailed data to be derived from the simulation, and multiple simulation runs can permit parametric changes to study a broader range of conditions (e.g., impacts at different speeds and angles). The LS-DYNA commercial FE package developed by Livermore Software Technology Corporation (LSTC) is used in the simulations (Hallquist 1997, 2006). It uses an explicit Lagrangian numerical method to solve nonlinear, three- dimensional, dynamic, large displacement problems. It has numerous features that allow for the analysis of several non- linear dynamic engineering problems. It has a large selec- tion of FE types which include one node lumped mass; two node spring; damper and beam elements; three and four node shell elements; and eight node solid and thick shell elements. For each of these element types, a number of element formulations are implemented in the code. As an example, more than 16 different shell formulations are avail- able. These include reduced, fully integrated, and membrane formulations. LS-DYNA has a library of more than 180 constitutive material models. The majority of these models can be used with all element formulations mentioned above. These models cover a wide range of material behaviors includ- ing elasticity, plasticity, thermal effects, and rate dependency. These constitutive models have been successfully used to model several materials including metals, plastics, rubber, soil, concrete, ceramics, composites, foams, and fluids. LS-DYNA has over 20 options for modeling connections including welds, rivets, and joints. Some of these connections incorpo- rate failure. It also has over 50 different methods for modeling initial conditions, boundary conditions, and loadings. Some of these include initial velocity, initial stress, nodal forces, pressure, prescribed accelerations, and fixed nodes. The most advantageous capability of LS-DYNA over other FE codes is its advanced contact algorithm. Over 20 contact inter- faces are available in the code including Nodes-to-Surface, Surface-to-Surface, Single-Surface, and Automatic-General. These allow for solving diverse types of impact problems. The LS-DYNA program has been used by the Research Team in many studies to address transportation safety problems. Several vehicle and roadside hardware models have been developed and used in these studies. Some of these models were used in this research to assess the per- formance of longitudinal barrier when installed on CSRS. These models and associated validations are presented in the following sections. 5.3.1 Vehicle Models The crashworthiness analysis under NCHRP Report 350 and MASH involves different test vehicles. For NCHRP Report 350, the small car is represented by an 820-kg vehicle (820C) and the pickup truck by a 2,000-kg vehicle (2000P). Under MASH, the small car is represented by an 1,100-kg vehicle (1100C) and the pickup truck by a 2,270-kg vehicle (2270P). These reflect the trend in the United States that vehicle sizes and weights are increasing (a primary reason for the new MASH requirements). Computer models rep- resenting these four vehicles are included in the array of vehicle models available to support crash simulation analy- ses (National Crash Analysis Center 2012a). Basic informa- tion about these FE vehicle models can be found in Table 5.1, but additional information can be obtained from references (National Crash Analysis Center 2012b, 2012c, 2012d, 2012e). Even though the research focuses on MASH evaluations, both sets of models were needed to allow validation against avail- able crash tests for the barriers. It should be noted that the test requirements only give generic vehicle features; different vehi- cles are often used in the actual testing. The models are believed to be viable surrogates for each of the weight classes. The fol- lowing sections describe the validation efforts undertaken for the Chevrolet Silverado (representing the 2270P) and Toyota Yaris (representing the 1100C) vehicle models (Marzougui et al. 2012b, 2012c). These reflected the new “extended vali- dation” approach used to create a vehicle model. The other two models (Geo Metro and Chevrolet C2500) were devel- oped earlier and had more usage but less rigorous validation. Because the Silverado and Yaris are the primary models used to assess crashworthiness against the latest criteria, the fol- lowing additional details are provided. 5.3.1.1 Chevrolet Silverado Model (2270P) The Silverado model was developed jointly by FHWA and NHTSA to serve multiple purposes in this research and advancement of vehicle and highway safety research. Reverse engineering methods were used to build the FE model and the attention to detail was critical to making it suitable for application for different crash conditions. The model con- sists of over 950,000 elements including the components of the steering and suspension systems (Marzougui et al. 2009b; National Crash Analysis Center 2009).

60 This model was initially validated following traditional protocols for comparison of the data from the full frontal impact with a vertical wall required under the New Car Assessment Program (NCAP) administered by NHTSA and the simulated results for that test. In addition, the Silverado model was subjected to validation exercises including the following: • Comparisons of actual and simulated inertial properties • Front suspension system component tests • Rear suspension system component tests • Full-scale speed bump and terrain traversal tests Data from these tests was useful in enhancing the model and providing quantitative measures that increased confidence in the predictive capabilities for roadside barrier impacts. The results are believed to indicate that this model will provide a sound basis for many types of crash simulation applications in the future. The FE model of 2270P vehicle is based on the 2007 Chevrolet Silverado 1500 pickup truck. The vehicle used for creating the model was a 4-door crew cab, short box, vehicle with a 4.8L, V8 engine and an automatic 4-speed transmis- sion weighing 2,298 kg. The model was developed through a reverse engineering process. The vehicle was disassembled Description Vehicle Image 1997 Geo Metro (820C) • Weight: 820 kg (1,806 lb) • CG 664 mm (26.14 in.) • Model Parameters: Parts-230, Nodes-200,348, Elements-193,200 • Features: FD, CD, SD • Validations: FF, SP • Original Release: 12/21/2000 2010 Toyota Yaris (1100C) • Weight: 1,100 kg (2,420 lb) • CG 1,004 mm rear, 569 mm high • Model Parameters: Parts-771, Nodes-998,218, Elements-974,348 • Features: FD, CD, SD, IM • Validations: FF, OF, MDB, SI, IP, SP, SC, ST, OT • Release Date: 12/02/2011 1994 Chevrolet C2500 Pickup Truck (2000P) • Weight: 2,000 kg (4,410 lb) • CG 664 mm (26.14 in.) • Model Parameters: Parts-248, Nodes-66,684, Elements-58,400 • Features: FD, CD, SD • Validations: FF, SP, SC, ST • Original Release: 12/12/2000; 11/03/2008 2007 Chevrolet Silverado Pickup Truck (2270P) • Weight: 2,270 kg (5,000 lb) • CG 736 mm (28.8 in.) • Model Parameters: Parts-606, Nodes-261,892, Elements-251,241 • Features: FD, CD, SD, IM • Validations: FF, IP, SP, SC, ST, OT • Original Release: 2/27/2009 Validations Legend • FF: NCAP Full Frontal • OF: Offset Frontal • SI: Side Impact • MDB: Modified Deformable Barrier • IP: Inertial Parameters • SP: Spring Response • SC: Suspension Components • ST: Suspension Tests (full-scale) • OT: Other Features Legend • FD: Fine Detail Version • CD: Coarse Detail Version • SD: Suspension Details • IM: Interior Modeled Table 5.1. Models representing NCHRP Report 350 and MASH test vehicles.

61 and each part was cataloged, scanned, measured, and clas- sified by material type. Each part was meshed to create an accurate computer model representing the data gathered in the disassembly, including geometry and material proper- ties. Material data and properties were obtained through coupon testing. Because the Silverado model is primarily used for roadside hardware testing, component testing and simulations were performed to ensure accurate representation of the suspen- sion systems. Over the years, the model has been validated using several full-scale crash tests [NHTSA, NCAP Frontal Barrier Impact—2007 Chevrolet Silverado, NHTSA Test Report 5877; NHTSA, NCAP Side Impact Test—2007 Chevrolet Silverado, NHTSA Test Report 6185; and Insurance Institute for Highway Safety (IIHS) Test CEF0825]. The tests included automotive crashworthiness tests as well as roadside hardware tests. One sample validation is summarized below. Additional validations are included in Appendix C. One of the tests that was used for the validations is a NCAP test conducted for NHTSA (Test 5877). The vehicle in this test impacted a rigid wall at 35 mph in a full frontal impact configuration (90° angle). The simulation results were com- pared with the test results. The simulation yielded similar vehicle kinematics and deformation, as shown in Figure 5.1. Figure 5.2 compares the left and right rear sill accelerations of the test and the simulation. These graphs indicate good correlation between the test and the simulation. The left and right rear sill velocities were also compared, showing a velocity change of 62 km/h, versus the test, which showed a velocity change of 65 km/h (Figure 5.3). The velocity profiles were similar for both the left and right rear sills, indicating a symmetric response. Figure 5.1. Side view of tested vehicle and model after NCAP crash. Figure 5.2. Left and right rear sill accelerations for test and simulation. Figure 5.3. Left and right rear sill velocities for test and simulation.

62 5.3.1.2 Toyota Yaris Model (1100C) This vehicle model was developed to be used in roadside hardware evaluation, as well as in occupant risk and vehicle compatibility analyses (National Crash Analysis Center 2011). It was selected to conform to the MASH requirements for an 1100C test vehicle. The model was based on a 2010 Toyota Yaris 4-door passen- ger sedan. Similar to the Silverado model, the vehicle was dis- assembled and each part was scanned to define its geo metry, measured for thickness, and classified by material type. Material data for the major structural components was obtained through coupon testing. A total of 160 tensile tests were performed to generate the material properties for 12 different materials. Upon completion of the model development, several auto- motive full-scale crashworthiness tests were used for valida- tions (NHTSA, NCAP Frontal Barrier Impact—2010 Toyota Yaris, NHTSA Test Report 5677; and NHTSA, NCAP Frontal Barrier Impact—2010 Toyota Yaris, NHTSA Test Report 6221). The model has also been used in roadside hardware impacts as the surrogate for the 1100C test vehicle. A sample validation is summarized below. Additional validations are included in Appendix C. One of the impact configurations that was used for the Yaris model validations was an NCAP frontal crash into a rigid barrier at 35 mph. Two full frontal NCAP tests were available for validation of the Toyota Yaris FE model in this configura- tion: Test No. 5677 and Test No. 6221. The overall global defor- mation pattern of the FE model was very similar to that of the NCAP test, as shown in Figure 5.4. Figure 5.5 compares the left and right rear seat accelerations of the test and the simu- lation, also indicating similar vehicle behavior between the test and the simulation. The response of the engine during the crash event was captured through two accelerometers. Both the engine top and bottom accelerations in the simulation closely tracked the engine response in the two tests, as shown in Figure 5.6. 5.3.2 Barrier Models Crash simulation analysis requires FE models of the barriers as well as the impacting vehicles. Three roadside hardware devices were identified as the longitudinal barriers to be studied in this research: • G4(1S) W-beam guardrail with height < 31 in. Identified as the most commonly used longitudinal barrier for both Figure 5.4. Side view of tested vehicle and model after NCAP crash. Figure 5.5. Left and right rear sill acceleration for test and simulation.

63 previously installed and currently being installed barrier on CSRS. • MGS W-beam guardrail with height ≥ 31 in. Identified as the second most commonly used longitudinal barrier currently being installed on CSRS. • NJ concrete barriers with heights ≤ 32 in. Identified as the second most commonly used longitudinal barrier for previously installed barrier on CSRS. FE models that represent the longitudinal barriers listed have been developed and used by the Research Team in pre- vious research (Esfahani et al. 2009; Marzougui et al. 2008b, 2009c, 2010b, 2010c, 2010d, 2011a, 2011b, 2012d). The fea- tures of these barrier models are described in the following sections. 5.3.2.1 G4(1S) W-Beam Guardrail Model This FE model of the G4(1S) was adapted from previ- ous modeling efforts by the Research Team to reflect the specifications for the hardware. This included the specifics for the posts, blockouts, and connectors. The model was based on explicit geometry of all components. Appropriate material and cross sectional properties were assigned to all components to ensure that the correct mass, inertia, and stiffness of the different parts were reflected in the model. The soil was also explicitly modeled using solid elements. The shape of the post was incorporated in the soil mesh to simulate the post/soil interactions. Because the geometry of the bolts was previously found to affect system behavior, the bolts were explicitly incorporated in the model. The model was used in several previous studies and validated against full-scale crash tests. The rails in this system were made up of standard 12-gauge W-beams with lengths of 3.807 m (12.5 ft). The rails were sup- ported using W150 × 12.6 (W6 × 9) steel posts. These posts were 1,830 mm (72 in.) in length and embedded 1,100 mm (43.3 in.) into the ground. Wood blockouts were placed between the posts and the W-beam rails and had dimensions of 150 mm × 200 mm × 360 mm (6 in. × 8 in. × 14 in.). The system level model of the guardrail system was modeled to have a total length of 53.3 m (175 ft) and anchored at both ends using a standard Breakaway Cable Terminal (BCT). The system consisted of 29 posts and 14 W-beam sections. Fig- ure 5.7 depicts some of the details of the model. 5.3.2.2 MGS W-Beam Guardrail Model The MGS guardrail system used in this research was based on the modified G4(1S) design. A similar modeling approach was used in developing this model with few minor differences. The differences between the two models include the following: • Rail height was increased to 31 in. by raising the whole G4(1S) system (except for the soil elements) by 2¼ in. Figure 5.6. Top and bottom engine accelerations for test and simulation. Figure 5.7. G4(1S) strong-post W-beam guardrail model.

64 • Blockouts were changed from 150 mm × 200 mm × 360 mm (6 in. × 8 in. × 14 in.) to 150 mm × 300 mm × 360 mm (6 in. × 12 in. × 14 in.) blocks. • Rail splices were moved from being at a post to being in between two posts. The model was also used by the Team in previous studies and validated against full-scale crash tests. Figure 5.8 shows some of the details of the model. 5.3.2.3 NJ Concrete Barrier Model This FE model of the NJ concrete barrier developed by the Research Team was used for the simulations. The NJ concrete barrier had a height of 32 in. As concrete safety barriers do not deform or deflect even under severe crash conditions, the bar- rier was modeled using rigid shell elements. For the simulations, the length of barriers was extended to over 150 ft to make sure the vehicle did not reach the end of the barrier before the end of the simulation. The barrier model mesh was refined to sizes between 2 in. and 3 in. to ensure optimum contact between the vehicle and barrier without excessive penetrations. Finer mesh was used at the edges of the barrier. The barrier was fixed to prevent any movement or deformation in the barrier during the crash simulation. The model is shown in Figure 5.9. 5.3.3 Barrier Modeling Details To create the FE models of the barriers, several key features were carefully examined and appropriate modeling tech- niques were used to ensure that the model was an accurate representation of the actual system. First, explicit geometry of all components of the systems were incorporated in the model. This included the W-beams, posts, blockouts, and bolts. This ensured the correct mass, inertia, and stiffness of the different parts were reflected in the model. The soil was also explicitly modeled using solid elements. The shape of the post was incorporated in the soil mesh to simulate the post/soil interactions. The geometry of the bolts was found to affect system behavior, so they were explic- itly incorporated in the model. These modeling conventions are described in the following paragraphs. 5.3.3.1 Modeling of Steel and Soil Elements Appropriate material and cross sectional properties were assigned to all components of the barrier systems. Rigid material was assumed for the concrete barrier models. For the W-beam guardrail models, two main LS-DYNA material types were used. The metal components, such as the posts and W-beams, were represented as “piecewise linear plasticity” material in LS-DYNA. This material model has been exten- sively used to represent structural metals, such as steel and aluminum, and it has been fully validated and optimized. The material behavior is isotropic elasto-plastic with strain rate effects and failure. The properties used for these materials were extracted from the literature as well as data from cou- pon tests that were performed on similar steels. The “soil-and- foam” model in LS-DYNA was used to represent soil properties. The properties used for this model were back-calculated from previously conducted tests. These tests consisted of a bogie vehicle impacting wood and steel posts that are embedded in soil similar to what has been used in the full-scale crash test. Simulations with the same test set ups were performed, and the material properties were varied until acceptable compari- sons were achieved between the tests and the simulations. 5.3.3.2 Modeling of W-beam, Post, and Blockouts A detailed FE model of the steel post with wooden blockout is shown in Figure 5.10(a) and the FE model of the W-beam Figure 5.8. MGS strong-post W-beam guardrail model. 19” 6” 32” 7” 10” 24” Figure 5.9. NJ concrete median barrier model.

65 is shown in Figure 5.10(b). For computational purposes, six rails located at the middle of the entire guardrail system were modeled using fine mesh, while the remaining rails were mod- eled using coarser mesh. All post and rails were modeled using quadrilateral shell elements. The material formulation used for the rail and post is the isotropic piecewise linear elastic plastic model. Wooden blockouts were modeled using eight node reduced integration hexahedral solid elements. These elements capture the behavior of the model at much less cost, because they consume much less computer time and memory. 5.3.3.3 Bolts Modeling Eight short bolts were used to connect the W-beams together and a long bolt was used to connect the rails to the wooden blockout and post as shown in Figure 5.11. For the small bolts, the material formulation selected for the bolts and nuts was the rigid material formulation. This assumption was made to reduce the computation time, because small elements are needed to capture the geometry of the bolts. These elements would control the time step and lead to larger computation time. By assuming the rigid material model for the bolts, their element size was no longer critical, because rigid elements did not control the time step. A spring was placed between the bolt head and the nut to represent the stiffness of the bolt. The properties of these springs were determined from the material properties, cross sectional area, and length of the bolt. The long bolts have significant effects on the behavior of the system and had to be modeled in detail. To accurately and efficiently represent these bolts, a special modeling technique was utilized in which the bolt was modeled with beam ele- ments to capture its tensile, bending, and shear behavior. By using beam elements, the time step was not controlled by the cross sectional geometry of the bolt. Hence, a larger simula- tion time step and smaller computation time was needed to reach a solution. An elasto-plastic material model with failure was assigned to the beam elements to simulate the nonlinear and failure behavior of the bolt. The geometry of the bolt is represented by shell elements with “null” material properties. The null shell elements had no effect on the stiffness of the bolts, and their size did not affect the simulation time step. They are used to represent the bolt geometry for only contact purposes. Nodes from shell elements were tied to the beam element nodes to transfer the contact forces. 5.3.3.4 Soil and Soil/Post Model The soil was modeled as a cylindrical block 2.7 m (9 ft) in diameter and 2.02 m (6.5 ft) in length as shown in Figure 5.12. These dimensions were chosen so that the behavior of the soil and post/soil interaction is accurately captured with reasonable computation time. The outer boundaries of the soil model were constrained using the non-reflection boundary constraint option. This option is often used in modeling an infinite domain and prevents the stress wave from reflecting (a) (b) Figure 5.10. FE models of (a) steel post with blockout and (b) W-beam rail. Figure 5.11. FE models of short and long bolts. Figure 5.12. Soil model with post and wooden blockout.

66 at the fixed boundary. The soil block was modeled using eight node hexahedral solid elements. The shape of the post was incorporated into the soil mesh with appropriate flange and web thickness to avoid penetration between post and soil and to have full representation of the post/soil interaction. An automatic single surface sliding interface was defined between the outer faces of the post and inner faces of the soil block to simulate the contact between the post and the soil, and friction between the post and the soil was also included. The material constitutive model used for the soil is the “soil and crushable foam” model. 5.3.4 Crash Simulation Software The crash simulations were performed using the LS-DYNA nonlinear explicit FE code Version MPP971sR6 on an Intel MPI 3.1 Xeon 64 parallel computer platform. The simu- lation run times would be expected to vary for other facilities depending on hardware, LS-DYNA version, and precision used. 5.4 Computer Model Validations Model validation involves simulating a known crash test and comparing the results. A solid validation effort provides confidence that reasonable variations of the model reflect- ing other situations will yield representative results. For this effort, there were multiple validations for each of the barriers selected for analysis. These made use of the best available crash test data existing at the time of the analysis. Table 5.2 lists the crash tests used for the model validations. A rigorous verification and validation (V&V) effort was undertaken to provide confidence that the models for each of the three barriers are viable in replicating crashes into bar- riers on CSRS. The results from the eight comparisons detail the viability or strengths of the validations based on the V&V results. A summary of the validation efforts is provided in Table 5.3, which includes the graphic of vehicle roll, pitch, and yaw angular rotations and change in vehicle velocity along the x-, y-, and z-directions. Additional comparisons from all seven cases, including side-by-side images from test and simulation at different stages of impact and overlay plots are shown in Appendix C. V&V analytic comparisons for all seven validation cases were also undertaken based on NCHRP Web-Only Docu- ment 179 (Ray et al. 2010). Roadside Safety Verification and Validation Program (RSVVP) Tables and Phenom- ena Importance Ranking Tables (PIRTs) were generated. Sample V&V results are included in the next sections. Full V&V reports for each of the seven cases selected are provided in Appendix C. The validity of the models was assessed by analyzing the distribution of energy associated with the crash event. The laws of physics dictate that the total energy be balanced. Typically, an energy balance graph is generated to assess changes in kinetic, internal, sliding, hourglass, and total energy. All of the comparisons were characterized by the following: • Relatively constant energy balances were noted suggesting there are no unusual characterizations in the structure of the model that would be an unrealistic sink (point of dissipation) of energy. • The kinetic energy associated with the motion of the vehicle dropped off as the velocity decreased during the crash. • Internal energy increased as components of the vehicle absorbed energy through deformation. • Sliding energy, which is associated to the friction between the vehicle and barrier, increased as expected during the simulations. All of the V&V criteria for energy balance were met. These aspects led to the conclusion that the model met the funda- mental requirements for crash simulation. Sample metrics derived from the RSVVP procedure in accordance to NCHRP Web-Only Document 179 are included Barrier Vehicle Test Date Place Evaluat ion Ref NJ Concrete 2002 Kia Rio 2214NJ -1 5/28/04 MwRSF MASH Polivka et al. 2006b 2007 Silverado 476460 -1-4 1/10/09 TTI MASH Bullard et al. 2009 G4(1S) 1989 C2500 405421 -1 11/16/95 TTI NCHRP Report 350 Bullard et al. 1996 2002 RAM 2214WB -2 4/08/05 MwRSF MASH Polivka et al. 2006a MGS 2002 Kia Rio 2214MG -3 11/08/04 MwRSF MASH Polivka et al. 2006c 1994 Geo Metro NPG-1 6/29/01 MwRSF NCHRP Report 350 Polivka et al. 2004 2002 D odge Ram 2214MG -2 10/06/04 MwRSF MASH Polivka et al. 2006d Note: MwRSF = Midwest Roadside Safety Facility; TTI = Texas A&M Transportation Institute. Table 5.2. Full-scale crash tests used for validations.

67 Test Model Set Up Angular Rotations Change in Velocity N J C on cr et e B ar ri er w / K ia R io 1 10 0C N J C on cr et e B ar ri er w / Si lv er ad o 22 70 P G 4( 1S ) w / C 25 00 2 00 0P G 4( 1S ) w / D od ge R am 2 27 0P M G S w / G eo -M et ro 8 20 C M G S w / K ia R io 1 10 0C Table 5.3. Summary validation results—change in vehicle velocities and rotations. (continued on next page)

68 M G S w / D od ge R am 2 27 0P Table 5.3. (Continued). in Figure 5.13. The RSVVP procedure consists of apply- ing statistical tests to determine how well the simulation curves compare with data collected from the test. The figure shows sample results of RSVVP for single-channel (graphs a through f) and multichannel (graph g) comparisons. Various means of comparing the data are shown in each comparison, including the following (moving from the upper left to the lower right): • Time history plot. The red line indicates the simulated data and the blue line indicates the test data for the crash event. Each data point is a measure of the acceleration recorded. • Plot of integrated time histories. Integrating the change of acceleration data allows the changes in velocity to be plotted. A general decrease in velocity is noted, as expected, although there is some deviation between the test and the simulation after the impact. • MPC metrics. This statistical metric provides a measure of “goodness of fit” between the two curves. Three param- eters are used for the evaluation: the magnitude (M), phase (P), and comprehensive (C, combined magnitude and phase). A value of less than 40 for M, P, and C is considered passing the criteria. • ANOVA metrics. Analysis of variance (ANOVA) is also used to compare the test and simulation curves goodness of fit. Two parameters are used for the comparison: the average residual between the curves and the standard devia- tion of the residuals. Values of less than 5% for the average residual and 35% for the standard deviation are considered passing the criteria. In this example, the metric meets the criteria and hence the boxes are labeled “pass.” • Residuals plots (time history, histogram, and cumulative). These plots show the residual (i.e., difference between the two curves in different forms). In the first plot, time history, the residual is shown versus time. In the second, the residual is shown in a histogram format where the percentage of the residual is plotted against the percentage of its occurrence. In the third plot, the cumulative sum of residuals is plotted. The program allows various types of single-channel data to be analyzed. The common crash test and simulation metrics compared are as follows: • X-acceleration: change in acceleration in the original direction of travel of the vehicle • Y-acceleration: change in acceleration in the lateral direction of travel of the vehicle • Z-acceleration: change in acceleration in the vertical direction of travel of the vehicle • Yaw rate: rate of change in original direction of travel of the vehicle • Roll rate: rate of change in lateral direction of travel of the vehicle • Pitch rate: rate of change in vertical direction of travel of the vehicle Because not all measurements have the same impor- tance in the tests, (e.g., in some tests little roll, pitch, or x-acceleration observed), these low magnitude channels could fail the evaluation metrics even if the simulation is valid. To overcome this problem, a multichannel compari- son is incorporated in the validation process, where each channel is given a weighting factor based on magnitude. A sample multichannel is shown in Figure 5.13. In this case, the figure indicates that the simulation passes on the multi- channel comparison metrics. In addition to graphs shown for the single-channel comparisons, this graph includes relative weights that were computed for each of the chan- nels used on the evaluation. These are used to weight the importance to the overall comparison of the two sets of data (test and simulations). In addition to RSVVP evaluations comparing the time history from the transducers mounted on the vehicle, NCHRP Web-Only Document 179 procedure establishes PIRTs aimed at comparing other aspects of the impact such as occupant risk numbers, barrier maximum deflections, and rotations. Table 5.4 shows a sample PIRT comparison. PIRTs for each of the seven cases selected are provided in Appendix C.

69 (a) X-acceleration (b) Y-acceleration (c) Z-acceleration (d) Roll (e) Pitch (f) Yaw Single-channel RSVVP comparisons. Multichannel RSVVP Comparisons Figure 5.13. Sample RSVVP single- and multichannel evaluations.

70 Evaluation Criteria Known Result Analysis Result Relative Diff. (%) Agree? St ru ct ur al A de qu ac y A A1 Test article should contain and redirect the vehicle; the vehicle should not penetrate, underride, or override the installation, although controlled lateral deflection of the test article is acceptable. Yes Yes YES A2 The relative difference in the maximum dynamic deflection is less than 20%. 1 m 0.960 m 4.0 % YES A3 The relative difference in the time of vehicle-barrier contact is less than 20%. 0.7 s 0.65 s 7.1 % YES A4 The relative difference in the number of broken or significantly bent posts is less than 20%. 4 4 YES A5 Barrier did not fail (Answer Yes or No). Yes Yes YES A6 There were no failures of connector elements (Answer Yes or No). Yes Yes YES A7 There was no significant snagging between the vehicle wheels and barrier elements (Answer Yes or No). No No YES A8 There was no significant snagging between vehicle body components and barrier elements (Answer Yes or No). Yes Yes YES O cc up an t R is k D Detached elements, fragments, or other debris from the test article should not penetrate or show potential for penetrating the occupant compartment, or present an undue hazard to other traffic, pedestrians, or personnel in a work zone (Answer Yes or No). Yes Yes YES F F1 The vehicle should remain upright during and after the collision. The maximum pitch and roll angles are not to exceed 75°. Yes Yes YES F2 Maximum vehicle roll: relative difference is less than 20% or absolute difference is less than 5°. 10 (0.45 s) 9 (0.35 s) 10% 1° YES F3 Maximum vehicle pitch: relative difference is less than 20% or absolute difference is less than 5°. 7 (0.67 s) 12 (0.67 s) 71% 5° YES F4 Maximum vehicle yaw: relative difference is less than 20% or absolute difference is less than 5°. 38 (0.8 s) 36 (0.72 s) 5.2% 2° YES L L1 The occupant impact velocity in the longitudinal direction should not exceed 12 m/sec and the occupant ridedown acceleration in the longitudinal direction should not exceed 20 g. Yes Yes YES L2 Longitudinal OIV (m/s): Relative difference is less than 20% or absolute difference is less than 2 m/s. 7.1 6.4 9.8% 0.7 m/s YES L3 Lateral OIV (m/s): Relative difference is less than 20% or absolute difference is less than 2 m/s. 4.4 5.4 22.7% 1.0 m/s YES L4 Longitudinal ORA (g): Relative difference is less than 20% or absolute difference is less than 4 g. 7.9 11.5 45.6% 3.6 g YES L5 Lateral ORA (g): Relative difference is less than 20% or absolute difference is less than 4 g. 8.4 10.1 20.2% 1.7 g YES V eh ic le T ra je ct or y M M1 The exit angle from the preferable test article should be less than 60% of test impact angle, measured at the time of vehicle loss of contact with test device. No No YES M2 Exit angle at loss of contact: relative difference is less than 20% or absolute difference is less than 5°. 16 18 11% 2° YES Table 5.4. Sample PIRT results from the validations.

71 5.5 Crash Simulation Parameters The research was initiated to answer a variety of questions over a range of conditions for barriers on CSRS. Over the course of the research, the questions were refined and the focus on critical conditions or situations sharpened. The follow- ing sections describe the simulation approach used to estab- lish useful insights and details to get answers to the research questions. 5.5.1 Analysis Conditions The findings of the literature review, state DOT survey, and crash data analysis indicated that there was a large set of parameters that potentially affect the safety performance of longitudinal barriers when placed on CSRS. These were discussed with NCHRP Project 22-29A panel in a review of the VDA. The VDA showed that there are differences in interface effectiveness that can be attributed to the degree of curvature and superelevation, shoulder width and slope, as well as barrier type and placement. The following sections describe the critical factors associ- ated with roadway conditions, barrier types and placement, and impact conditions that were identified. These parameters and associated ranges were the focus of the simulation efforts. Other parameters were considered after the basic influences were determined. 5.5.1.1 Roadway Design Conditions Deliberations with NCHRP Project 22-29A panel led to defining the primary road design conditions to be analyzed, including curvature and superelevation, shoulder width and slope, and side slope. Various degrees of roadway curvatures reflect the range of superelevation applications commonly found on the highway. These range from tight curves used on ramps to gentle sweeping curves. A total of six road- way curve conditions with different curvatures and super- elevations were used in the VDA. These conditions were selected based on the Green Book design superelevation tables. The analyses incorporated three superelevations (6%, 8%, and 12%). For each superelevation, two curvatures were selected representing the minimum radii at the 50-mph (80-km/h) and 80-mph (130-km/h) design speeds. The curvatures/superelevation combinations were as follows: • 614 ft (187 m)/12% • 2,130 ft (649 m)/12% • 758 ft (231 m)/8% • 2,670 ft (814 m)/8% • 833 ft (254 m)/6% • 3,050 ft (930 m)/6% To investigate the effects of roadside shoulder, different shoulder angles and widths were analyzed. Three shoul- der widths were considered in the analyses: 4 ft (1.22 m), 8 ft (2.44 m), and 12 ft (3.66 m). Four shoulder angles were included in the analyses: 0%, 3%, 6%, and 8%. An impor- tant note here is that the shoulder angle is different than the conventional shoulder slope defined relative to true horizontal plane (see Section 4.4.4). A negative roadside/median slope of 12H:1V relative to shoulder was to be used for all simulations. 5.5.1.2 Barrier Types and Placement The following three types of barriers were investigated (a concrete safety shape and two variations of the strong-post W-beam guardrail based on the state DOT survey responses): • Concrete barrier: NJ concrete barrier with a height of 32 in. (813 mm) • Strong-post W-beam guardrail: G4(1S) with heights of 27¾ in. and 29 in. (705 mm and 737 mm) • Strong-post W-beam guardrail: MGS with a height of 31 in. (787 mm) Selections of these barriers were made considering the following: • NJ Concrete Barrier. This classic, widely used concrete safety-shape barrier was a starting point, because vehicle vaulting and rollovers have been attributed to the slop- ing sides of the barrier profile. The VDA indicated that there would be no underride or override interface issues for the 1100C or 2270P vehicles. However, the NCHRP Project 22-29A panel expressed the concern that the two- stage slopes of the barrier had been seen to cause small vehicle rollovers, so these were simulated. The simulations included both the small car and the pickup truck. • G41S W-Beam Barrier. This widely used barrier was first accepted after the adoption of NCHRP Report 350, and it has been widely deployed. Its original design has a height of 27¾ in. The VDA indicated that there could be interface issues for the larger vehicle. Tests with the Silverado and the G4(1S) resulted in vaulting of the barrier. An FHWA techni- cal memorandum in 2010 recommended a nominal height of 29 in. for new installations (Nicol 2010). The simulations were run at both the original height and the 29-in. height recommended by FHWA. • MGS Barrier. This newer W-beam barrier was designed to accommodate vehicles with higher centers of gravity with a rail height of 31 in. This barrier was accepted by the FHWA in 2005. The VDA indicated that there could be underride interface issues for the small car, so the focus of the simula- tions was on the 1100C vehicle. There were no simulations for the pickup.

72 The barriers in all FE simulations were placed at the edge of the “operational” shoulder. Placement further off the shoulder was found to be an uncommon practice. Three barrier vertical orientations were analyzed includ- ing true-vertical orientation, perpendicular to the shoulder surface, and perpendicular to the road surface. 5.5.1.3 Impact Conditions The following impact conditions were based on MASH for the vehicle types, speed, and angle: • Vehicle Type – 2270P pickup truck: 2007 Chevrolet Silverado model – 1100C small car: 2010 Toyota Yaris model • Impact Conditions – Impact angle 25° – Impact speed 62 mph (100 km/h) To limit the number of simulation runs required, the strategy was to use VDA simulation to bracket the poten- tial problem conditions. Based on the VDA results, cases in which there was (1) very poor interface between the vehicle and the barrier and (2) the vehicle was likely to override the barrier were not simulated. Cases that showed marginal or good performance were evaluated using the FE simulations. Other factors that were used to minimize the number of simulations included the following: • Curvature. There is evidence that the sharper the curvature (smaller radius), the more likely serious crash problems will result. Therefore, the plan was to simulate the mid- range curvatures and superelevations [i.e., 758 ft (231 m)/8% and 2,670 ft (814 m)/8%] and, based on the outcome from these simulations, the other curvatures/superelevations would be simulated. • Impact Angle. Only the usual 25° impact angle was used in the simulations. Higher impact angles were initially investigated, but it was determined that these barriers are not designed for these impact angles and would not likely meet the MASH criteria even for flat surfaces and straight barriers. • Barrier Offset. Barrier lateral placement was limited to the break point between the shoulder and roadside/median. Even though other placements were initially consid- ered, after consultation with the NCHRP Project 22-29A panel, it was decided that other lateral placements are uncommon. • Barrier Height. Simulations were limited to the standard barrier heights. Taken together, these factors resulted in more than 150 simulation runs. After summarizing the results and discussing them with the NCHRP Project 22-29A panel, a second round of simulations was undertaken to address concerns. This involved another 100 simulations to inves- tigate other shape concrete barriers and to add depth to the insights developed on the safety performance of barriers on CSRS. Each simulation took between 20 h and 40 h of CPU time to provide detailed analyses of crash events involving typical barriers on CSRS with varying features. Table 5.5 shows the distribution of these simulations. Each simulation has been assigned a case number to facilitate the evaluation process and outline useful analyses and comparisons. The number of runs in each cell is unequal for various reasons. A main reason was that when the VDA results indicated that there was a “poor interface,” the need for FE simulations was less critical. Because it was infeasible to simulate every CSRS condition under consideration given the amount of computing time that would be required, a selection was made based on realistic operational limits. 5.5.1.4 Analysis Assumptions A number of different conditions for road departures were considered. The following assumptions were made prior to the analyses: • The roadside had a firm surface. Ploughing into the surface by tires was negligible. • Vehicles were “tracking” as they entered the roadside (i.e., vehicle initial speed is in the same direction as its longitudinal axis). Vehicle Concrete Barrier G41S W-Beam Barrier MGS Barrier NJ F Shape 27¾-in. Height 29-in. Height 31-in. Height Small Car (1100C) 59 Runs (Cases 101−175) 0 Runs 0 Runs 0 Runs 22 Runs (Cases 601−652) Pickup (2270P) 50 Runs (Cases 201−274) 45 Runs (Cases 901−963) 38 Runs (Cases 801−876) 50 Runs (Cases 301−358) 0 Runs Table 5.5. Summary of simulation runs.

73 • There were no driver inputs (e.g., steering, braking) that affect the vehicle. • The road friction was made identical in all runs using a friction coefficient of 0.9. • There was a smooth transition between the pavement and shoulder and the shoulder and side slope. Where these assumptions do not hold, other effects will occur that will alter the stability of the vehicle. Other condi- tions related to these assumptions could be modeled, but they were not at this stage. 5.5.2 Evaluation Criteria The performance of the longitudinal barriers in the crash simulations was evaluated in accordance with the criteria presented in MASH. The simulations replicated MASH Test 3-10 for the small car (1,100-kg test vehicle) and MASH Test 3-11 for the pickup truck (2,270-kg test vehicle). The fundamental criteria for crashworthiness evaluation are shown in Table 5.6. These ensure that there is adequate structural integrity of the barrier, all occupant risk metrics are met, and vehicle trajectories are acceptable. The primary difference in this effort was that the barrier was deployed on one of the CSRS configurations previously defined. For this research, the MASH tests provide a useful imme- diate and long-term value in the assessment of the perfor- mance of barriers used on CSRS. The implications of the MASH (and its predecessors) on performance are well understood because highway engineers have considered a common group of metrics in a structured approach for many years. The simulation approach also provides data that allows other metrics to be analyzed including deformations (barrier and post deflection), barrier component forces and stresses, and vehicle lift. In addition, simulation technology allows unique views of the vehicle-to-barrier contacts to be gener- ated to better understand unusual results. 5.6 Crash Simulation Results 5.6.1 Simulation Analysis Summaries Figure 5.14 depicts a typical summary of the simulation results. It provides a pictorial view of the impact with a barrier on a CSRS and the resulting behavior of the vehicle. It documents the CSRS conditions including radius, super- elevation rate, shoulder width and slope, barrier orienta- tion, and speed and angle of impact. The diagram depicts the time sequence of vehicle and barrier interaction in the crash event. Below the diagram is the MASH evaluation sum- mary, which shows the key metrics generated by the simu- lation and whether the results passed or failed the MASH criteria. Similar summaries were generated for each of the FE simulations run. The full set of summaries is included in Appendix D. 5.6.2 General Comparative Analyses The simulation results permitted considerations of the influences of curvature, road profile, barrier vertical ori- entation, impact angle, impact speed, and combinations of these factors on safety performance. The findings pro- vided the basis for formulating barrier design, selection, and installation guidelines. Multiple metrics were generated in the analysis, but all are not documented here. The following sections describe the findings from types of analyses that were undertaken. Structural Adequacy A: Test article should contain and redirect the vehicle; the vehicle should not penetrate, underride, or override the installation, although controlled lateral deflection of the test article is acceptable. Occupant Risk D: Detached elements, fragments, or other debris from the test article should not penetrate or show potential for penetrating the occupant compartment, or present an undue hazard to other traffic, pedestrians, or personnel in a work zone. F: The vehicle should remain upright during and after the collision, although moderate roll, pitching, and yawing are acceptable. H: The OIV in the longitudinal direction should not exceed 40 ft/s and the ORA in the longitudinal direction should not exceed 20 g. I: Longitudinal and lateral ORA should fall below the preferred value of 15.0 g, or at least below the maximum allowed value of 20.49 g. Vehicle Trajectory For redirective devices, the vehicle shall exit within the prescribed box. Table 5.6. MASH crashworthiness evaluation criteria for simulation analyses.

74 5.6.2.1 Influence of Barrier Orientation Figure 5.15 depicts typical results for situations with the same radius, superelevation, and shoulder configurations for the NJ concrete barrier but different barrier orientations. The barriers were installed with normal and vertical orienta- tions. The first and second panels provide the MASH results and time sequence diagram for a 2270P vehicle impact- ing the barrier on a CSRS at 25° and 100 km/h for a tight curve with similar shoulder width conditions. It can be seen that the barrier impact event results in failure due to vehicle roll for the true-vertical orientation case, while the normal orien- tation case shows a pass. The MASH evaluation results reflect similar unacceptable degrees of maximum vehicle roll (91.77°, exceeding the MASH maximum 75° roll criterion) for the true-vertical orientation and acceptable (28.54°) roll angle for the normal orientation case. The values for OIV and ORA in the longitudinal and lateral directions are similar for both vertical and normal orientation cases and are below the MASH maxi- mum values. Similarly, panels 3 and 4 compare the impact event for the 2270P vehicle for the normal and vertical orientation with a slight shoulder angle change. The results reflect similar patterns, but failures for the true-vertical orientation for both cases. These results suggest that the normal orientation shows better performance than the true-vertical one for these CSRS conditions using the NJ concrete barrier. The true-vertical orientation also has a higher propensity for vehicle instability. 5.6.2.2 Influence of Curvature Figure 5.16 depicts simulation results from two different radius conditions with similar shoulder configurations for the NJ concrete barrier installed with a normal orientation. The first and third panel provide the MASH results and time sequence diagram for an 1100C vehicle impacting the barrier on a CSRS at 25° and 100 km/h. While there are some scal- ing differences in the diagrams, it can be seen that the barrier impact event results in similar performance for the 758-ft and 2,670-ft curvatures. The MASH evaluation results also reflect similar degrees of maximum vehicle roll and pitch (e.g., roll 26.59° and 25.85°, and pitch 21.99° and 21.72°). Similarly, the values for OIV and ORA in the longitudinal and lateral directions are the same order of magnitude. Panels 2 and 4 compare the results for the 2270P vehicle. The results reflect comparable degrees of maximum vehicle roll and pitch (e.g., 2270P - NJ Concrete Barrier (102) Radius Super Shoulder Width Shoulder Angle Barrier Orient. Speed Angle 614 ft 12% 4 ft 6% Normal to Road 100 [km/h] 25° Evaluation Criteria A Test article should contain and redirect the vehicle; the vehicle should not penetrate, underride, or override the installation although controlled lateral deflection of the test article is acceptable. Pass D Detached elements, fragments, or other debris from the test article should not penetrate or show potential for penetrating the occupant compartment, or present an undue hazard to other traffic, pedestrians, or personnel in a work zone. Pass F The vehicle should remain upright during and after the collision. The maximum pitch and roll angles are not to exceed 75°. Max Roll (°) 28.54 Pass Max Pitch (°) 23.38 H Longitudinal and lateral OIV should fall below the preferred value of 30 ft/s (9.1 m/s), or at least below the maximum allowed value of 40 ft/s (12.2 m/s). Vx (m/s) –5.29 Pass Vy (m/s) 8.15 I Longitudinal and lateral ORA should fall below the preferred value of 15.0 g, or at least below the maximum allowed value of 20.49 g. Ax (g) 9.92 Pass Ay (g) 17.65 Figure 5.14. Sample simulation analysis summary report.

75 Parameters and Results Case Time Sequence View CSRS: Radius 614 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 91.77 (Fail) * Max Pitch – 61.09 (Fail) * H – OIV – Vx – –4.67 (Pass) Vy – 7.61 (Pass) I – ORA – Ax –12.34 (Pass) Ay – 18.65 (Pass) 101 V CSRS: Radius 614 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 28.54 (Pass) Max Pitch – 23.38 (Pass) H – OIV – Vx – –5.29 (Pass) Vy – 8.15 (Pass) I – ORA – Ax – 9.92 (Pass) Ay – 17.65 (Pass) 102 N CSRS: Radius: 614 ft; Superelevation: 6%; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: Normal; Impact Speed/Angle: 100kmh/25o CSRS: Radius 614 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 81.65 (Fail)* Max Pitch – 55.67 (Fail)* H – OIV – Vx – –4.73 (Pass) Vy – 7.65 (Pass) I – ORA – Ax – –15.12 (Pass) Ay – –16.55 (Pass) 103 V CSRS: Radius 614 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 29.15 (Pass) Max Pitch – 26.25 (Pass) H – OIV – Vx – –5.47 (Pass) Vy – 8.19 (Pass) I – ORA – Ax – –10.16 (Pass) Ay – 17.69 (Pass) 104 N *The combination of high simulated roll and pitch reflect considerable vehicle instability that can result in a rollover in the time after the simulation is terminated. CSRS: Radius: 614 ft; Superelevation: 12%; Shoulder Width: 4 ft; Shoulder Angle: 6% Barrier: NJ Concrete; Orientation: Normal (N) or True Vertical (V); Impact Speed/Angle: 100kmh/25º Figure 5.15. Comparison of barrier orientation effects. roll 33.66° and 35.84°, and pitch 35.84° and 31.6°). The values for OIV and ORA for the longitudinal and lateral directions are also of a similar order of magnitude. These results suggest that for these CSRS conditions, the NJ concrete barrier perfor- mance was similar for these two curvatures. Figure 5.17 depicts the results for two different radius condi- tions with similar shoulder configurations for the NJ concrete barrier installed with a true-vertical orientation. The first and third panel provide the MASH results and time sequence dia- gram for an 1100C vehicle impacting the barrier on a CSRS at 25° and 100 km/h. It can be observed that the barrier impact event resulted in different performances for the 758-ft and 2,670-ft curvatures. The MASH evaluation results reflect high vehicle roll and pitch for both curvatures (e.g., roll 53.93° and 70.93°, and pitch 58.68° and 81.87°), but only the impacts for the small car on the large radii curve exceeded allowable levels. The values for OIV and ORA for the longitudinal and lateral directions are the same order of magnitude. Panels 2 and 4 compare the impact event for the 2270P vehicle. The results reflect similar degrees of maximum

76 vehicle roll and pitch (e.g., roll 57.53° and 59.73°, and pitch 42.15° and 47.05°) for the pickup. The values for OIV and ORA for the longitudinal and lateral directions are the same order of magnitude. These results show a more pronounced roll effect due to the barrier orientation, but not a significant effect due to the difference in curvature. 5.6.2.3 Influence of Barrier Type Figure 5.18, 5.19, and 5.20 depict the different barriers analyzed for increasing curve radii and otherwise similar conditions. Because not all combinations of CSRS con- ditions were simulated, it is not possible to make impact comparisons across the three barrier types. Variation in the crash behaviors is apparent, demonstrating that the varying of CSRS parameters influences the vehicle-to-barrier inter- face and the performance of the barrier. These figures show that the response for the small car and pickup truck varied. For all cases, the OIV and ORA values for the longitudinal and lateral directions are of a similar order of magnitude and direction for the similar barrier type. This may suggest that current barriers can function effectively across a variety of CSRS conditions. These comparisons show that it is possible to analyze the performance differences across a range of CSRS conditions for typical barriers. It may also be possible to analytically define the influence patterns for safety performance, but that was not possible without simulation results for all combina- Parameters and Results Case Time Sequence View CSRS: Radius 758 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 26.59 (Pass) Max Pitch – 21.99 (Pass) H – OIV – Vx – –5.47 (Pass) Vy – 9.76 (Pass) I – ORA – Ax – –3.19 (Pass) Ay – 14.23 (Pass) 224 CSRS: Radius 758 ft, 8% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 33.66 (Pass) Max Pitch – 29.86 (Pass) H – OIV – Vx – –5.39 (Pass) Vy – 8.24 (Pass) I – ORA – Ax – 13.38 (Pass) Ay – 18.43 (Pass) 124 CSRS: Radius 2,670 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 25.85 (Pass) Max Pitch – 21.72 (Pass) H – OIV – Vx – –5.49 (Pass) Vy – 9.72 (Pass) I – ORA – Ax – –3.27 (Pass) Ay – –13.54 (Pass) 264 CSRS: Radius 2,670 ft, 8% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 35.84 (Pass) Max Pitch – 31.6 (Pass) H – OIV – Vx – –5.57 (Pass) Vy – 8.13 (Pass) I – ORA – Ax – –12.09 (Pass) Ay – 18.28 (Pass) 164 CSRS: Radius: 758 and 2,670 ft; Superelevation: 8%; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: Normal; Impact Speed/Angle: 100kmh/25º CSRS: Radius: 2,670 ft; Superelevation: 8%; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: Normal; Impact Speed/Angle: 100kmh/25º Figure 5.16. Comparison of radius effects for NJ concrete barrier with normal orientation.

77 tions of factors. The simulation analysis generated summary tables to reflect the pass/fail patterns across the various CSRS conditions analyzed. 5.6.3 Barrier-Specific Results Performance envelopes were also generated to provide a quantitative assessment of the influence of roadway and barrier design elements on the outcome of the crashes. These included the influences of curvature, road profile, barrier lateral position, barrier vertical orientation, and combina- tions thereof. The following sections provide a summary of the observations, comparisons, and conclusions drawn from the simulation analyses for each type of barrier. A global set of conclusions is generated to provide a basis for decisions about testing in the next phase of the project and for the development of guidance. 5.6.3.1 Concrete Barrier Results Table 5.7 contains a summary of the simulation runs that were made for the NJ concrete barrier under the designated Parameters and Results Case Time Sequence View CSRS: Radius 758 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 53.93 (Pass) Max Pitch – 58.68 (Pass) H – OIV – Vx – –5.33 (Pass) Vy – 9.59 (Pass) I – ORA – Ax – –4.42 (Pass) Ay – 11.14 (Pass) 223 CSRS: Radius 758 ft, 8% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 57.53 (Fail)* Max Pitch – 42.15 (Fail)* H – OIV – Vx – –4.75 (Pass) Vy – 7.89 (Pass) I – ORA – Ax – 10.47 (Pass) Ay – 17.64 (Pass) 123 CSRS: Radius 2,670 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 70.93 (Fail)* Max Pitch – 81.87 (Fail)* H – OIV – Vx – –5.14 (Pass) Vy – 9.43 (Pass) I – ORA – Ax – –4.44 (Pass) Ay – 10.41 (Pass) 263 CSRS: Radius 2,670 ft, 8% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 59.73 (Fail)* Max Pitch – 47.05 (Fail)* H – OIV – Vx – –4.83(Pass) Vy – 7.95 (Pass) I – ORA – Ax – 9.71 (Pass) Ay – 18.12 (Pass) 163 *The combination of high simulated roll and pitch reflect considerable vehicle instability that can result in a rollover in the time after the simulation is terminated. CSRS: Radius: 758 and 2,670 ft; Superelevation: 8%; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: True Vertical; Impact Speed/Angle: 100kmh/25º CSRS: Radius: 2,670 ft; Superelevation: 8%; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: True Vertical; Impact Speed/Angle: 100kmh/25º Figure 5.17. Comparison of radius effects for NJ concrete barrier with true-vertical orientation.

78 Parameters and Results Case Time Sequence View CSRS: Radius 614 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 54.42 (Pass) Max Pitch – 26.43 (Pass) H – OIV – Vx – –5.02 (Pass) Vy – 7.86 (Pass) I – ORA – Ax – –12.32 (Pass) Ay – 17.26 (Pass) 107 CSRS: Radius 758 ft, 8% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 62.37 (Fail) Max Pitch – 25.18 (Fail)* H – OIV – Vx – –5.05 (Pass)* Vy – 7.97 (Pass) I – ORA – Ax – –14.75 (Pass) Ay – 17.22 (Pass) 125 CSRS: Radius 833 ft, 6% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 49.51 (Pass) Max Pitch – 20.67 (Pass) H – OIV – Vx – –4.89 (Pass) Vy – 8.02 (Pass) I – ORA – Ax – –9.34 (Pass) Ay – –18.30 (Pass) 133 CSRS: Radius 2,130 ft, 12% super Vehicle: 2270P A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 34.13 (Pass) Max Pitch – 31.01 (Pass) H – OIV – Vx – –5.14 (Pass) Vy – 8.04 (Pass) I – ORA – Ax – –10.13 (Pass) Ay – 17.39 (Pass) 151 *The combination of high simulated roll and pitch reflect considerable vehicle in stability that can result in a rollover in the time after the simulation is terminated. CSRS: Radius: 614 to 2,130 ft; Superelevation: Variable; Shoulder Width: 4 ft; Shoulder Angle: 8% Barrier: NJ Concrete; Orientation: Vertical; Impact Speed/Angle: 100kmh/25º Figure 5.18. Barrier type effects: NJ concrete barrier. CSRS parameters and provides the evaluation results derived for the 2270P and 1100C vehicles. The runs covered six different curvatures and superelevation conditions. Four shoulder angles (i.e., slope) and three shoulder width conditions were analyzed. Three barrier orientation con- ditions were also considered: true vertical, perpendicular to the shoulder surface, and perpendicular to the roadway surface. The cells of the matrix are shaded based on the MASH evaluations for the Test 3-10 and Test 3-11 impact conditions. The cases highlighted in red failed the MASH criteria and those in dark green met the criteria. All cases that are shaded in light green (with “*”) were assumed to meet the MASH requirements based on having less severe conditions. The clustering of the dark green-shaded cells around the red cell indicate attempts to assess the degree of effects leading to failures. The following observations are made from the data for cases depicted in this table: • All failures to meet the MASH requirements for the 1100C and 2270P vehicles resulted from exceeding the maximum allowable roll angle.

79 CSRS: Radius: 614 to 3,050 ft; Superelevation: Variable; Shoulder Width: 12 ft; Shoulder Angle: 8% Barrier: MGS; Orientation: Normal; Impact Speed/Angle: 100kmh/25o Parameters and Results Case Time Sequence View CSRS: Radius 614 ft, 12% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 5.23 (Pass) Max Pitch – 3.85 (Pass) H – OIV – Vx – –6.74 (Pass) Vy – 5.83 (Pass) I – ORA – Ax – –15.50 (Pass) Ay – 12.17 (Pass) 603 1100C CSRS: Radius 758 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 5.30 (Pass) Max Pitch – 4.52 (Pass) H – OIV – Vx – –6.78 (Pass) Vy – 5.88 (Pass) I – ORA – Ax – –17.26 (Pass) Ay – 10.98 (Pass) 611 1100C CSRS: Radius 853 ft, 6% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 5.23 (Pass) Max Pitch – 5.45 (Pass) H – OIV – Vx – –8.86 (Pass) Vy – 5.68 (Pass) I – ORA – Ax – –12.42 (Pass) Ay – 10.45 (Pass) 627 1100C CSRS: Radius 2,670 ft, 8% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 5.21 (Pass) Max Pitch – 3.23 (Pass) H – OIV – Vx – –6.88 (Pass) Vy – 5.71(Pass) I – ORA – Ax – –18.56(Pass) Ay – 11.86 (Pass) 641 1100C CSRS: Radius 3,050 ft, 6% super Vehicle: 1100C A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 5.30 (Pass) Max Pitch – 4.70 (Pass) H – OIV – Vx – –9.98 (Pass) Vy – 5.52 (Pass) I – ORA – Ax – –11.77 (Pass) Ay – 10.54 (Pass) 651 1100C Figure 5.19. Barrier type effects: MGS barrier.

80 CSRS: Radius: 614 to 2,670 ft; Superelevation: Variable; Shoulder Width: 8 ft; Shoulder Angle: 6% Barrier: G41S (@ 29 in.); Orientation: Normal; Impact Speed/Angle: 100kmh/25o Parameters and Results Case Time Sequence View CSRS: Radius 614 ft, 12% A – Containment (Pass) D – Detached Elements (Fail) F – Max Roll – 17.61 (Pass) Max Pitch – 12.04 (Pass) 305 2270P CSRS: Radius 758 ft, 8% super MASH Evaluations: super MASH Evaluations: A – Containment (Fail) D – Detached Elements (Pass) F – Max Roll – 20.05 (Pass) Max Pitch – 17.96 (Pass) 175 315 2270P CSRS: Radius 833 ft, 6% super MASH Evaluations: A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 11.59 (Pass) Max Pitch – –7.51 (Pass) 329 2270P CSRS: Radius 2,130 ft, 12% super MASH Evaluations: A – Containment (Fail) D – Detached Elements (Pass) F – Max Roll – 33.84 (Pass) Max Pitch – 14.74 (Pass) 345 2270P CSRS: Radius 2,670 ft, 8% super MASH Evaluations: A – Containment (Pass) D – Detached Elements (Pass) F – Max Roll – 11.82 (Pass) Max Pitch – –6.35 (Pass) 355 2270P Figure 5.20. Barrier type effects: G41S barrier. • Most failures occurred when barriers were in true-vertical orientation. There was only one failure when the barrier was oriented normal to shoulder (for the 1100C vehicle) and one when normal to road for the large vehicle. • More failures were noted with higher superelevation: 12% superelevations had larger number of failures than the 8%, and the 6% superelevation had the lowest number of failures. • Narrower shoulder widths had more failures than the wider shoulder widths; more failures were noted with the 4-ft than the 8-ft shoulder widths, and the 12-ft shoulder width had the lowest number of failed cases. • The larger the shoulder angle (i.e., the larger the difference in angle between the road and shoulder), the higher the number of failures. The 8% shoulder angle had more cases that failed than the 6%, which had more failed cases than the 3%, and the 0% had the lowest number of failures. These observations do not reflect the degree of failure. Determining that would be possible by looking at the results

81 Notes: * = Barrier performance extrapolated based on other simulation results. Simulation case numbers are shown in parentheses. Table 5.7. Performance table for 32-in. NJ concrete barrier.

82 for each of the failed cases. Some efforts to discern the effects of CSRS conditions using analytical means were not success- ful, given the dispersion effects and interrelationship between factors investigated. These would not reflect other differences in speeds, impact angle, vehicle loading, and driver reactions that are associated with real crashes. Simulations were also performed for the F-shape con- crete barrier. Since the F-shape is known to introduce less vehicle instability than the NJ concrete barrier, only cases where the NJ concrete barrier did not meet MASH require- ments were simulated. The results are shown in Table 5.8. When comparing Table 5.7 with Table 5.8, it can be seen that the F-shape shows improved performance over the NJ concrete barrier, although not in a consistent way across all orientations. 5.6.3.2 W-Beam Barrier Results Table 5.9 contains a summary of the simulation runs made for the W-beam barriers (i.e., G41S and MGS) for various parameters and their results. The runs covered the curvatures and superelevation conditions selected. The same set of shoul- der slope and width conditions as the concrete barriers was analyzed. Because W-beam barriers are traditionally installed in a true-vertical orientation, only this case was considered in the performance table. These results are based on analysis of the most critical vehicle. Because there were concerns about underride and pocketing of the small vehicle for the MGS, the simulations were undertaken with the 1100C vehicle. The larger and consequently higher vehicle (2270P) was used for the G4(1S) simulations, because this barrier is more suscep- tible to overrides. The G4(1S) barrier was initially analyzed based on a barrier height of 27¾ in. All simulation runs made for the barrier at that height led to unacceptable results. The FHWA Technical Memorandum dated May 17, 2010, indi- cated FHWA’s preference for using 29-in.-high barriers over 27¾-in.-high ones and recommended that agencies consider adopting 31-in. barrier designs (Nicol 2010). Consequently, for barrier heights less than 31 in., a height of 29 in. was used in the simulations. The following observations are made from the data in this table: • Narrower shoulder widths had more failed cases than the wider shoulders. More failures were noted with the 4-ft shoulder width than the 8-ft shoulder width, and the 12-ft shoulder width had the lowest number of failed cases. • Larger shoulder angles (i.e., higher difference in angle between the road and shoulder) led to more failures. The 8% shoulder angle had more cases that failed than the 6%, which had more failed cases than the 3%, and the 0% had the fewest failed cases. • G4(1S) W-beam barriers at 27¾ in. high did not meet MASH requirements for the range of conditions simulated. It was concluded that their performance would not be acceptable for CSRS conditions. • The G4(1S) W-beam barrier 29 in. high met MASH require- ments for most cases where there were wider shoulders. There were 11 failures and 2 marginal results for 4-ft wide shoulders compared with 2 failures for 8-ft shoulders. These all occurred across all radii and superelevations analyzed. • The higher MGS met the MASH requirements for all cases. The 11 simulations undertaken covered a valid cross section of CSRS conditions without a failure. These observations do not reflect the degree of failure. That would be possible by comparing the detailed simulation results for each of the cases. Marginal passes are indicated by orange shading. Efforts to discern the effects of CSRS conditions using analytical means were not successful given the dispersion effects and interrelationship between factors. These would not reflect other differences in speeds, impact angle, vehicle loading, and driver reactions that are associated with real crashes. 5.7 Conclusions This effort successfully applied the FE models and crash simulation to analyze the safety performance of longitudi- nal barriers typically used on CSRS under varying impact conditions. The effort led to the following observations and conclusions: • For the most part, the VDA results were similar to the selected FE simulations. There was evidence in some cases that the vehicle-to-barrier interface data did not fully reflect safety performance. This can be attributed to the effects of barrier design and placement (e.g., face shape and orienta- tion) that were not explicitly considered in the VDA, as well as the inherent barrier “strength” that was able to redirect a vehicle even if the interface was not ideal. • These simulations focused on MASH conditions for the general norm of TL-3 for impacts of longitudinal barri- ers placed on the outside edge of CSRS curves. Thus, the results cover only impacts by the MASH defined small car (1100C) and large pickup truck (2270P). • This analysis focused on longitudinal barriers deployed on the outside of level curves on typical CSRS across a range of parameters. It focused on meeting MASH TL-3 evaluation criteria that are useful for barrier selection decisions. Deeper analyses of the crash impact dynamics would be needed to assess critical severities and deter- mine means to improve barrier design and placement guidelines.

83 Notes: * = Barrier performance extrapolated based on other simulation results. Simulation case numbers are shown in parentheses. Table 5.8. Performance table for 32-in. F-shape concrete barrier.

84 Notes: * = Barrier performance extrapolated based on other simulation results. Simulation case numbers are shown in parentheses. Mar. = marginal pass. Table 5.9. Performance table for W-beam guardrails.

85 • The G41S barrier with a height of 27¾ in. had demonstrated vaulting failures in tests and simulations undertaken in other efforts conducted to demonstrate the efficacy of the MASH requirements. The FHWA also issued a technical memorandum recommending that the height of the bar- rier be raised to 29 in. to address such problems. These problems were noted in the early simulations for deploy- ments on CSRS, so no further runs were performed. There was no evidence of underride or snagging issues with the small vehicle. The MGS analysis was conducted last and began with a concern of underride issues. The early FE simulation runs showed no evidence of this, so the number of runs was limited. • Analysis of barrier orientation was an objective of this research. It was noted in the performance envelopes that concrete barriers oriented in true vertical were most likely to fail. It can be noted that there appears to be less vehicle instability for the larger radius curves for either vehicle type, and the impact metrics are similar. • The number of factors identified at the outset created a very large analysis matrix making it hard to isolate individual effects. For example, the combined effects of shoulder width and slope for the six CSRS curvature and superelevation conditions for the three barriers for large and small vehicles would have required close to 600 simulations. Some addi- tional benchmarking runs might make it possible to get further insights on relative effects. • The NJ concrete barrier provided sufficient containment and redirection of the large and small impacting vehicles, but there were many cases where rollovers occurred for the small vehicle, probably due to the influence of the barrier’s orientation. These rollovers were lower for orien- tations normal to road surface. There was no indication of vaulting problems for the situations analyzed. • Secondary analysis of the 32-in. F-shape concrete barrier indicated that their variations in face slope resulted in fewer failures. The analysis did not attempt to determine the specific influence of face slope differences on safety performance. • The efficacy of the G4(1S) at the 27¾-in. height was assessed in the simulation runs, and a similar pattern of failures for 4-ft shoulders suggests a problem. Given this has been a widely used barrier, additional investigations may be warranted. • The MGS barrier showed no indication of override issues in the VDA, so there was limited effort to simulate impacts with the large vehicle. The simulations for the small car suggested that underride and snagging are not issues. These results suggest that there may be isolated safety per- formance issues associated with the deployment of standard longitudinal barriers on CSRS. In general, the results indicate that current practices provide for reasonable safety expecta- tions for the vehicles considered.