Crash Energy Management Systems for Australian Rolling Stock

Transcription

1 Crash Energy Management Systems for Australian Rolling Stock

2 DOCUMENT CONTROL SHEET CRC for Rail Innovation Old Central Station, 290 Ann St, Brisbane Q 4000 GPO Box 1422 Brisbane Qld 4001 Tel: Fax: Document: Title: Crash Energy Management Systems for Australian rolling stock Project Leader: Manicka Dhanasekar Authors: Guillaume Michal, Azhar Nasir, Sun Yun, Nannan Zong, Weihua Lee, and Manicka Dhanasekar. Project No.: R3.114 Project Name: Crashworthiness of Australian rolling stock Synopsis: Australian rolling stock experiences some unique challenges typical of the Australian track network characteristics and maintenance practices. With a view to examining the provisions of the current crashworthiness standards of the Rail Industry Safety and Standards Board (RISSB) for the Australian operational environment, a Rail CRC R3.114 research project was established; this report presents the summary of the work carried out and conclusions reached. The RISSB provisions for locomotives and passenger cabins prescribe maximum longitudinal compressive and tensile forces to anti-climbers and collision posts as well as vertical forces to anti-climbers and couplers similar to the US provisions. The RISSB provisions also recommend finite element analysis to identify stress exceedance above permissible stress limits. Exceedance above permissible limits is allowed as long as no permanent deformation occurs; this is similar to the European crashworthiness provisions. Simple expressions for determining the effectiveness of the CEM derived in this report highlight the acceptability of either the increased or the reduced mass CEM designs depending on the operating speeds. It was also shown that consists possessing CEM exhibit lower derailment risk than those that do not possess CEM. Numerical models established and validated as part of the project provide adequate insight into the performance of the vehicles in terms of crash energy dissipation and survival space. REVISION/CHECKING HISTORY REVISION DATE NUMBER 0 18/11/2013 ACADEMIC REVIEW (PROGRAM LEADER) INDUSTRY REVIEW (PROJECT CHAIR) APPROVAL (RESEARCH DIRECTOR) 1 13/05/2014 Colin Cole Ivo Anic Chris Gourlay DISTRIBUTION DESTINATION REVISION Industry Participant for Review x X Copyright 2013 CRC for Rail Innovation This work is copyright. Apart from any use permitted under the Copyright Act 1968, no part may be reproduced by any process, nor may any other exclusive right be exercised, without the permission of CRC for Rail Innovation. CRC for Rail Innovation November 2013 Page ii

3 Table of Contents Executive Summary... iv List of Figures... v List of Tables... vi Abbreviations and Acronyms... vi 1. Introduction Definitions Rail crashworthiness and integration of CEM designs Overview UK and Europe designs United States of America Australia Worthiness of a CEM design Structures of CEM designs Front end Collision posts and corner posts Coupler and anti-climbers Wall filling structures Shape and tapering of columns Occupant space Finite Element Method Structure Wheelset: rolling Crash analysis Lateral collisions Crashworthiness enhancement Application of CEM designs to the prevention of derailments Multibody formulation Modelling of components Simulations Conclusions References CRC for Rail Innovation November 2013 Page iii

4 Executive Summary Australian rolling stock experiences some unique challenges typical of the Australian track network characteristics and maintenance practices. With a view to examining the provisions of the current crashworthiness standards of the Rail Industry Safety and Standards Board (RISSB) for the Australian operating environment, a Rail CRC R3.114 research project was established; this report presents the summary of the work carried out and the conclusions reached. The project was carried out by three universities (QUT, UoW, CQU) and three industry partners (RISSB, Victorian Department of Transport, QRN). Monash University and the South Australian Department of Transport also contributed to the project. Bombardier and UGL have kindly allowed the university researchers to visit their factory to gain a better understanding of the design/construction process of rolling stock. The literature review and research found that: the RISSB provisions for locomotives and passenger cars prescribe maximum longitudinal compressive and tensile forces to anti-climbers and collision posts as well as vertical forces to anti-climbers and couplers similar to the US the RISSB provisions, similar to the European crashworthiness provisions, recommend finite element analysis to identify stress concentrations and exceedance above permissible limits. Exceedance above the permissible stress is allowed as long as no permanent deformation occurs a simple expression for determining the effectiveness of the CEM derived in this report highlights the acceptability of either the increased or the reduced mass CEM designs for operational speeds between 10 m/s to 20 m/s the potential risk of lateral impact the rolling stock may experience on the Australian track network level crossings may be similar irrespective of being designed/fitted with better or poorer performing CEM designs that are determined purely from the longitudinal collision scenarios consists possessing CEM exhibit lower derailment risk compared to those consists with no CEM designs due to the improved vehicle stability during the crash event numerical models established and validated as part of the project provide adequate insight into the performance of the vehicles in terms of crash energy dissipation and survival space. CRC for Rail Innovation November 2013 Page iv

5 List of Figures Figure 1: Idealised CEM characteristics (Jacobsen 2008)... 6 Figure 2: Controlled progressive collapse (RISSB 2008) Figure 3: Energy efficiency for a type B modification. Figure 4: Energy efficiency for a type C modification Figure 6: Finite element model of a bogie assembly Figure 7: Finite element body consist Figure 8 Contribution of the traction force to the impact force Figure 9: Pitch down of the front end due to uncontrolled deformation Figure 10: failure modes of the cab car for different structural cases Figure 11: Crush deformation progress of a poorly performing structure Figure 12: Crush deformation progress of the corrected structure Figure 13: (a) Kinetic and internal energies. (b) Gross energy dissipation from Xue et al (c) Crush characteristic of the cab and (d) Force-deformation of the can end structure from Xue et al Figure 14: Deformation mode from second the software package Figure 15: Scenario of lateral collision Figure 16: From left to right, collision scenario at 5, 12 and 19m from the front edge Figure 17: Deformation modes for a lateral impact at (a) 5m, (b) 12m and (c) 19m Figure 18: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 5m of the top of the wheels Figure 19: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 12m of the top of the wheels Figure 20: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 19m of the top of the wheels Figure 21: Structural configuration of (a) intermediate vehicle and (b leading vehicle (Xue et al. 2007) Figure 22: (a) Passenger car elements and (b) details of the bogie and rail arrangements for multibody modelling Figure 23: Wheel and rail profiles Figure 24: Crush zone modelling Figure 25: Idealised crush zone s force and crush length characteristics for (a) the HE zone and (b) the LE zone Figure 26: Energy absorber modelling Figure 27: Coupler modelling (a) at slow impact and (b) at shunt impact Figure 28: (a) Obstacle offset before collision and (b) illustration of the model with 7 wagons Figure 29: Evolution of the wheel rail contact during collision for the non-cem design case Figure 30: Offset of wheelset during the collision of a non-cem design at (a) 91ms and (b) 412ms Figure 31: Evolution of the wheel rail contact during collision for the CEM design case CRC for Rail Innovation November 2013 Page v

6 List of Tables Table 1 AIS Code, HIC, and Chest Deceleration Table 2: 1990s and Current Design Standard vs. State of the Art Requirements (U.S. Department of Transport, 2006) Table 3: CEM Design Study Parameters Abbreviations and Acronyms AAR: Association of American Railroads APTA: American Public Transit Association AU: Australia CEM: Crash Energy Management EU: European Union FEM: Finite Element Method FRA: Federal Railroad Administration RISSB: Rail Industry Safety and Standards Board (Australia) SEA: Specific Energy Absorption US: United States of America CRC for Rail Innovation November 2013 Page vi

7 1. Introduction Collisions of rail vehicles can cause serious injuries and fatalities to the driver and passengers. To minimise this, rail vehicles must have adequate structural strength and energy absorption capacity to withstand the forces of the collision and ensure survival spaces for all occupants. Rail safety requirements for passenger train vehicles call for appropriate Crash Energy Management (CEM) designs comprising a high-energy (HE) crush zone at the train front and low-energy (LE) crush zones between vehicles. Crush zones are structurally designed to collapse and absorb the kinetic energy in a controlled manner. The injuries the occupants can experience are lowered through this deformation process. CEM is an essential part of the crashworthiness design of rolling stock. Frameworks and incentives for CEM implementation have been in constant evolution during the last two decades through the integration of new standards. Historical evolution of the industry, accident statistics, as well as environmental and vehicle classifications led to crashworthiness scenarios and conformity criteria based on permissible stresses for the structural material, biomechanics principles for the survivability of the passengers and minimum survival spaces. Standards play a role in transitioning the industry towards an era of structural energy management. The chain of collapsing events of CEM zones is complex and is affected by the magnitude, position and orientation of the impact; it is, therefore, expensive and time consuming to perform full scale collision tests and this limits the number of cases and designs that can be evaluated. Lab testing can assess individually novel materials and/or CEM components at a fraction of this cost while numerical models offer a cost effective approach to optimising the overall structural arrangement. Modern multibody dynamics and Finite Element Methods (FEM) can be used in the investigation of train collisions, crashworthiness and the absorption of energy in the structure. Analysis can be as simple as one-dimensional models, just evaluating interactions between the vehicles, or complex using finite element crash simulations for detailed three dimensional parametric studies (Kirkpatrick et al. 2001; (Sun et al. 2012). In Australia, the Rail Industry Safety and Standards Board (RISSB) is developing a new set of standards dubbed AS75## series. In particular, section AS7520 parts 1 to 4 (see RISSB 2008 Parts 1-4) are drafts aimed at providing specifications for both the national design and performance standards. Parts 1, 2, 3 and 4 describe the locomotives, the freight vehicles, the passenger vehicles and the maintenance vehicles respectively. The CRC rail project R3.114 aims at analysing some selected standards from other nations (the Europe and the USA), establishing statistics on Australian rail accidents and studying the applicability and limits of crashworthiness standards when applied to local conditions. As part of this exercise, a review of the CEM designs was carried out and forms part of this document. CRC for Rail Innovation November 2013 Page 1

8 After a section dedicated to general definitions, this report focuses on the evolution of crashworthiness in the standards of Europe, the US and Australia followed by a discussion on the integration of CEM designs and a presentation of the classical CEM design research and solutions. The subsequent two sections deal with modelling. First, a Finite Element Method (FEM) analysis of a longitudinal impact of a typical passenger car is presented. Results are discussed in terms of energy dissipation and details of the deformed failure modes are given. A second model, based on a semi-rigid analysis of multibody dynamics, is carried out to demonstrate the positive potential of CEM designs in limiting derailments. CRC for Rail Innovation November 2013 Page 2

9 2. Definitions Key words used in the crashworthiness literature are defined briefly in this section; a comprehensive definition is beyond the scope of this report, for which one should consult the references listed at the back pages of this report Collision Three dominant modes of train collision can be observed from the analysis of Australian rolling stock accidents (Michal et al. 2010). These are: (1) frontal; (2) rear; and (3) side/lateral collisions. Of them, the frontal collision appears to represent the greatest occurring threat to both the train drivers and the passengers. Train accidents at level crossings in which a lead locomotive is involved also challenge the frontal and the side/lateral crashworthiness. (Rear collisions were not covered as a separate scenario as for a symmetrical passenger train they can be envisaged as the same as frontal (i.e observed from the other train). Many frontal crash scenarios have been introduced in the standards to evaluate the crashworthiness of rail vehicles and that of train consists. Standards have long taken account of the seriousness of frontal collisions and emphasise the need for protection against head-on crashes (Michal et al. 2013). Lateral impacts are comparatively less considered, albeit risks are present in Australia (Michal et al. 2010). The relative lack of requirements can be related to the low level of occurrence of lateral accidents and the difficulty of providing structural strength along the length of the side panels economically within a space. Although lateral and roof strength requirements have been increasingly present in standards, they are more related to rollover situations rather than collisions per se. A detailed analysis of lateral collisions and their effects on rolling stock is provided in section 7 of this report. 2.2 Energy dissipation mechanisms Crash energy dissipation mechanisms for a passenger train must be designed to satisfy crashworthiness requirements. They must perform in accordance with a set of prescribed crash scenarios, mostly based on frontal collisions. The energy dissipation mechanisms can incorporate various structural design features and special equipment to absorb the energy without either interfering with the survival spaces of the occupants or compromising the deceleration levels. CRC for Rail Innovation November 2013 Page 3

10 Through appropriate designs, localised large deformations of the crush zones at selected locations can be minimised whilst maintaining maximum energy absorption levels. The majority of the crash energy absorption takes place at the front of the train during a frontal collision. Therefore, the crush zone at the front end is usually referred to as the high energy (HE) crush zone. Absorption of the remaining energy and the subsequent impact energy between the passenger cars happen at the coupler locations. Because of the comparatively lower energy absorption requirements between cars, these areas are known as low energy (LE) crush zones. Both the HE and the LE crush zones can take the form of a sequential crushable assembly such as push-back coupler followed by crushing tubes made of tubes or honeycombs followed by adequate corner posts. 2.3 Collision forces and structural strength In scenarios of frontal collisions, the longitudinal force is an important factor to assess the severity of the crash. Traditionally, the longitudinal forces have been specified as proof loads to ensure that a car body frame can transmit operational compressive loads and that the spaces where the drivers and the passengers occupy retain their integrity. For example, the requirement of longitudinal loading due to frontal collisions in the UK is specified by the Transport Safety Investigation (TSI) (Erskine 2003) as 1.5MN higher than the mean collapse load of the designated crush zones. As stronger survival cells for driver and passengers are highly desirable, they influence safety standards in terms of prescribing higher loads. The increase in strength will often accompany the increase in weight, and hence the collision energy is an issue the crashworthiness designs should accommodate; use of modern lightweight materials is one way to improve this factor but financial considerations often restrict their use to specific and limited zones. 2.4 Vehicle response Severe dynamic interactions occur sequentially in a frontal crash. The rail vehicle dynamics are significantly influenced by the interactions between the colliding vehicles and the performance of the crush zones. The front vehicle may vertically override its counterpart. Overriding is often associated with substantial loss of occupants space and consequently casualties. The provision of anti-climber plates at the extremities of rail vehicles is now an essential part of any crashworthiness standard. Even in a train collision with initially in-line and coincident centrelines, trains may undergo lateral buckling. For instance, when a high longitudinal load is present, the connection formed by the couplers can laterally push the ends of the vehicles due to the nature of the contact. As a result, two adjacent vehicles would laterally offset from each other leading to a sawtooth pattern of the train consist. When significant dynamics are involved, lateral buckling can lead to derailment and pile up. CRC for Rail Innovation November 2013 Page 4

11 The tendency to override and lateral buckling depend upon the nature of the collision forces, the dynamic responses of consists, as well as the initial conditions of the impact. These factors are difficult to control in full size tests. Crash modelling allows the study of these complex interaction mechanisms and provides insight into the improvements deemed necessary. 2.5 Human body injury criteria The head injury criteria (HIC), chest deceleration and neck injury criteria are often used to evaluate the mode and severity of predicted injuries (Tyrell et al. 1995). Head Injury Criteria are defined as: 2 1 t HIC t2 t1 atdt t2 t 1 t As noted in Tyrell et al. (1995), the HIC can be seen as a non-linear measure of the average acceleration the head is subject to during a given time interval that maximises the measure. The Abbreviated Injury Scale (AIS), published by the American Association for Automotive Medicine provides a basis for comparison of HIC and chest deceleration. Table 1 lists the AIS Code and the corresponding values of HIC and chest deceleration. More information can be obtained from Brell, Van Erp and Snook (1999). AIS Code HIC Table 1 AIS Code, HIC, and Chest Deceleration (Tyrell et al. 1995) Head Injury Chest Deceleration Headache or dizziness G s Single rib fracture Unconscious less than 1 hour; linear fracture Chest Injury G s 2 to 3 rib fractures; sternum fracture Unconscious 1 to 6 hours; depressed facture G s 4 or more rib fractures; 2 to 3 rib fractures with hemothorax or pneumo-thorax Unconscious 6 to 24 hours; open fracture Unconscious more than 24 hours; large hematoma G s Greater than 4 rib fractures with hemothorax or pneumo-thorax; flail chest G s Aorta laceration (partial transection) 6 >1860 Non-survivable >90 G s Non-survivable CRC for Rail Innovation November 2013 Page 5

12 3. Rail crashworthiness and integration of CEM designs The idea of crashworthiness with an engineering science focus applied to rail vehicles is relatively new. This section presents a review of the literature relevant to this concept. 3.1 Overview The history of crashworthiness research of rail vehicles is not extensive. While there are some studies dealing with crashworthy designs of new rail vehicles as well as the crashworthiness assessment of existing conventional rail vehicles, little literature focuses on analysing and improving the crashworthiness characteristics and weaknesses of existing conventional vehicles (Lewis 1994; Kirkpatrick et al. 2001; Gao & Tian 2007). This is especially true for lateral impact studies. Most of the relevant projects were launched and completed around the globe during the last few decades. It is now understood that the idea that stronger is always better is in conflict with the need to provide better occupant protection (Scholes & Lewis 1993; Smith 1996; Leutenegger 2001; Chirwa 1994; Xue et al. 2005). The new philosophy of energy management in exceptional conditions is increasingly being used in design, in addition to the strength requirements for normal operations. Controlled deformation and collapse of structural elements are augmented with dedicated devices, so that the impact energy can be dissipated safely outside the occupied areas. Crashworthy energy-absorbing zones occupy minor volumes of a train and are able to sustain large plastic deformations. This approach led to the development and use of materials and structures with relatively high volumetric energy absorption densities. CEM designs are ideally located at the extremities of the vehicles, in a way such that interference with coupling systems remains minimal during normal operations, while supplementing absorption levels during a collision. In Figure 1 a possible evolution of the concept is illustrated. Car body crush zone Underframe Stiffness Coupling Elements Figure 1: Idealised CEM characteristics (Jacobsen 2008) CRC for Rail Innovation November 2013 Page 6

13 The area defined under the plateau of the structure absorber provided by the CEM design represents an added capacity of energy absorption and is triggered after the crushing tube (coupling elements) but before the underframe deforms. It is within this zone that the deceleration can be controlled within a range commensurate to injury levels. CEM design requires a trade-off between the crush force and the crush length for a given energy absorption capacity. The higher the crush force, the smaller will be the crush length; but the deceleration levels may be lethal. The higher the crush length, the smaller will be the crush force; but the survival space may be reduced to an unacceptable volume. In order to understand the crush force/length profile, the performance of a crashworthiness design must be assessed. Even though full-scale or component crash tests are credible methods, they have serious drawbacks. They are very expensive, time consuming, limited in the definition of the test cases and require a qualified team to build, execute and measure in safe conditions a scenario and its outcomes. They can hardly be used for new vehicles in the early stages of design. Because the tests are destructive, it is impractical to consider all scenarios with a single test vehicle. Computational simulation offers some alternatives. Once a vehicle model has been established, it can be adapted to any other design, just by importing the new geometry details from, for example, CAD or Solid Works models. The key issue is, however, that the simulations need to be supported by acceptable vehicle modelling principles and proven techniques for example, representation of the interaction between the components and their characteristics. Agreeable comparisons with real crash tests are necessary. Fortunately details of crash tests carried out in the USA are widely available (Xue et al. 2004, 2005, 2007; Tyrell et al. 2005,2006; U.S Department of Transportation 2007; Mayville et al. 2002; Martinez et al. 2005; Radewagen 2002). Therefore, it makes sense to model the vehicles used in the crash tests first to evaluate whether or not the tools in the software system are adequate for modelling the crash tests. The investment in initial model establishment will, therefore, be high. As the technique can be adapted for various designs, including analysis of what-if scenarios, over all the initial investment will be paid back in the long run. As for many industrial areas, standards are instrumental in orienting the performance of a structural design and establishing minimum safety requirements. The last two decades have seen the integration of dynamic testing as part of the crashworthiness certification in the US and Europe (Michal et al. 2013). The philosophy behind these standards is generally based on three safety aspects: providing a strong survival cell (coach body) reducing crushing caused by overriding through incorporating suitable devices limiting passenger secondary impact injuries by reducing passenger acceleration impulses. CRC for Rail Innovation November 2013 Page 7

14 The European performance standard EN15227 (European Committee for Standardization 2000) also provides a detailed framework for the crashworthiness assessment of rail vehicles though simulations, speeding up development, reducing capitalisation and allowing the reuse of numerical models in the industry. In Australia, RISSB is developing (at the time of writing this report) a new set of standards aimed at providing national rolling stock specifications. These standards refer to the European and American standards for crashworthiness design with details of collision scenarios being left to be agreed between the purchaser and the manufacturer; acceptance criteria should include a limit on deceleration or a maximum collapse force resulting in the collapse being confined to the crumple zones. Publications and standards on crashworthiness of rail vehicles in the UK /Europe and the USA provide a wealth of data to support design details (Erskine 2003). Information from the UK/ Europe and the USA is presented in Sections 3.2 and 3.3 respectively. 3.2 UK and Europe designs In the UK, a number of variants of multiple units and locomotive hauled coaching stock were designed and built in the late 1940s (Erskine 2003). Many of these were based on a new standard design, Mark I (Mk I). This had a fabricated steel underframe designed to carry vertical and longitudinal loads and a relatively lightweight fabricated steel body which was intended to be non-load bearing. There were light and heavy variants of the Mk I underframe with longitudinal proof strengths varying from approximately 110 tonnes to 200 (imperial) tons. The crashworthiness of these vehicles was significantly better than earlier wooden bodied stock, but in collisions involving overriding, the longitudinally stiff vehicle underframe could cause significant damage to the body of an adjacent vehicle (Erskine 2003). In the early 1960s Mk II design was developed with a stronger focus on the longitudinal proof load capacities above floor level. The crashworthiness of these vehicles was significantly better than that of the Mk I design because of the higher resistance of the body to penetration/deformation and the lower longitudinal stiffness of the underframe. This reduced the susceptibility and the consequences of overriding in longitudinal collisions. During the 1970s a new standard design of fabricated steel coach, called the Mk III was developed, and multiple unit derivatives were designed and built. Over several years following the introduction of Mk III type vehicles, it was observed that the performance of the vehicles was very good in a variety of collisions. Large amounts of collision energy were absorbed through post yield deformation of the body structure, particularly at the vehicle ends. CRC for Rail Innovation November 2013 Page 8

15 During the 1980s British Railways decided to quantify the crashworthiness performance of so called best current practice steel bodied Mk III type vehicles, in order to produce a sound basis for new vehicles specifications. This performance was quantified by performing a longitudinal crushing test on the body ends of a Mk III coach and a Class 317 EMU to measure the force-deflection characteristic of the body ends. The principal developments of rail vehicle crashworthiness in the UK can be summarised as follow: detailed analysis of past accidents, resulting in more detailed accident reporting and the compilation and annual updating of accident statistics. The aim of this work was to correlate accident type, vehicle damage and casualty levels, and enable risk assessment, cost benefit analysis, and to monitor the accrual of the perceived safety benefits of crashworthy designs. development of validated mathematical modelling of train rake behaviour in collisions, also utilising validated vehicle collapse characteristics. In Europe, the International Union of Railways (UIC) has historically provided a series of leaflets covering design requirements for traction and rolling stock. In the late 1950s, the UIC undertook a series of instrumented load tests on the bodies of passenger coaches from a number of European countries. British Railways submitted a Mk I coach for these tests. The purpose of this work was to gain data on the strength of body shells and their ability to meet a series of longitudinal and vertical load cases, which had been developed by the UIC organisation. Subsequently, UIC leaflet 566OR specified structural design requirements for coaches for international traffic. The load cases in the leaflet have been widely used for the design of rolling stock in Europe, which included the below design requirements to enhance the performance of the vehicles during collisions: the coach body and underframe structure shall be an assembly forming a tubular beam in order to ensure the maximum protection for passengers, the connection between the side walls and the underframe shall be designed with adequate resistance to horizontal shearing forces the end walls, strengthened by anti-collision posts, shall be joined to the headstock, ends of the cant rails and roof, in such a way that the maximum amount of energy produced during a collision is absorbed first by the deformation of the end wall before the passenger compartments are deformed. Whilst the above 566OR did not specify a defined level of crashworthiness, the combination of the longitudinal design proof load cases, plus the requirements for the body ends to have collision posts and be able to absorb collision energy through deformation, ensured that vehicles achieved a reasonable level of crashworthiness. Erskine says, During the second half of the 1990s the European Committee for Standardisation produced a document entitled 'Railway applications Structural requirements of railway vehicle bodies'. This was first CRC for Rail Innovation November 2013 Page 9

16 produced in draft form with input from representatives from railway administrations throughout Europe. It was subsequently published as the European Standard EN12663 (European Committee for Standardization 2000) and British Standard BS EN 12663:2000. The standard is intended to provide a uniform basis for the structural design of railway vehicles ranging from mainline vehicles, through suburban and urban transit stock to tramways. The standard does not have any specific crashworthiness requirements, but does specify that there should be an adequate margin between the maximum design load and the load at ultimate failure. Erskine also says: with the increase in high speed international travel, European collaborative projects have been undertaken to develop a co-ordinated crashworthy strategy. The TRAINCOL ( ) and SAFETRAIN ( ) projects used train accident statistics to investigate collision scenarios and analyse train crashworthiness. The projects continued on to develop achievable designs, and proved not only the realisable safety benefits, but also the validity of theoretical modelling and prediction techniques when compared with practical testing(erskine 2003). Analyses based on the projects results were expressed in the form of limits in terms of deceleration, energy absorption and collapse distance in the cases of passenger compartment (Low Energy Ends) and driving ends (High Energy ends). analysis of the result has provided limits on vehicle vertical forces, and peak vertical offsets (misalignment). Comparison has been made of dynamic versus quasi-static validation and testing, concluding that each produced acceptable results preservation of a driver s survival space of 750 mm minimum was recommended. 3.3 United States of America Again following Erskine, the US requirement for a compressive proof strength at the coupler arose in the 1940s when it was observed that the vehicles designed to a certain load survived in collisions, whereas those of lesser strength did not. The possible introduction of high speed European trains in the US prompted a risk analysis, which considered the likelihood of an accident and its consequences. The accident likelihood was developed from a study of all relevant factors (route features, active safety measures, etc.) coupled with known accident data, to develop a number of collision scenarios. The accident consequence was derived from accident data and extensive simulation and modelling of collision behaviour. The collision behaviour included train collision dynamics, vehicle crush characteristics, occupant dynamics and injury studies. Also from Erskine, major recent research programs in the USA have followed a somewhat different approach to those in the UK and Europe. All theoretical studies have been backed up and verified by extensive testing, and have enabled comparison of collision risk between classic and crashworthy trains, including accidents involving both types. Major conclusions from this work are: CRC for Rail Innovation November 2013 Page 10

17 more favourable (lower) decelerations are experienced by passengers during collisions in trains incorporating crash energy management designs below 110 km/h both the classic and the CEM designed trains effectively preserve the passenger survival space above 110 km/h the CEM designed trains were more effective in preserving survival space than the classic designs there was a high incidence of collisions at level crossings overriding was identified as a major threat to passenger safety, both in terms of likelihood and consequence lateral buckling of a train rake was also highly correlated with collision casualties, and these phenomena prompted extensive studies for trains with speeds of over 200 km/h energy absorption ranges were defined at different locations vehicles should have sufficient rollover strength to withstand twice their own weight when on their side or roof. 3.4 Australia The proposed Australian Standards for Railway Rolling Stock, AS7520 Parts 1 to 4, describe requirements for the structural strength of a particular type of vehicle i.e. Locomotive, Freight, Passenger and Infrastructure maintenance vehicles in parts 1 to 4 respectively. The main purpose of the standards is to: 1. prescribe the minimum structural integrity level of the vehicle body to ensure safe performance under normal operating conditions 2. minimise risks to train crew and members of the general public in the event of collisions or derailments. The Australian rail vehicle design requirements are referred to as proof loads considering longitudinal, vertical and combined loadings. The critical design stress shall take the yield stress (0.2% proof limit), 80% of the ultimate stress or 80% of the critical buckling stress whichever is the smallest. If a demonstration of compliance with this standard is undertaken using Finite Element Analysis, high localised stresses may acceptably exceed the stress criteria limits set in this standard so long as one of the following conditions is fulfilled: (a) they are associated with model singularities, or (b) they would not result in significant permanent deformation being experienced by the vehicle structure when the load is removed. Methods used to demonstrate that no significant permanent deformation is experienced can include: engineering observation and judgement use of non-linear analysis to determine if there is any deformation after a load application /removal cycle relating the results of physical tests to analysis results CRC for Rail Innovation November 2013 Page 11

18 Figure 2: Controlled progressive collapse (RISSB 2008) The standard requires medium and heavy duty locomotives to have AAR-type high strength collision posts, corner posts and cab end structures whereas for light duty locomotive and passenger cab structure designs the standard is more flexible so long as the CEM recommendations are fulfilled. It is worth noting that these requirements are not applicable to freight vehicles. The collapsing of the structure shall provide a controlled deformation of the crumple zones within the unoccupied areas of the vehicles part of the consist. Figure 2, extracted from the provisional standard, shows an example of how controlled deformation and collapse can be achieved at one end of a light duty locomotive or passenger cab. However, the design standards do not provide explicit upper or lower bounds, and stipulate only that the amount of energy to be absorbed in each crumple zone shall be appropriate for the intended service and operating conditions and shall be defined based on the relevant collision scenarios. Performance is not described in terms of velocity, mass, or any other conformity criterion. The proposed crashworthiness scenarios are comparable to the European approach and are: like-for-like collision between two identical units colliding head-on collision with a freight train collision at a level crossing CRC for Rail Innovation November 2013 Page 12

19 The proposed validation of compliance with the crashworthiness requirements are based on: test of energy absorbing devices and crumple zones calibration of the numerical model of the structure numerical simulation of the design collision scenarios. The following international standards have been suggested to define the CEMS for locomotives and passenger cars with particular attention for choosing crashworthiness requirements compatible with the proof strength requirements: EN C-1 (European Committee for Standardization 2000) APTA SS-C&S (APTA Commuter Rail Executive Committee 2006) 49 CFR, Part 238, Subpart C - Tier I vehicles (Federal Railroad Administration 1997, 2010) 49 CFR, Part 238, Subpart D - Tier II vehicles (Federal Railroad Administration 1997, 2010) The standard definition of a heavy/medium/light locomotive is based on axle and trailing loadings. Axle loads greater than 22 tons and trailing loads above 5000 tons define heavy locomotives. Axle loads of less than 18 tons and trailing loads less than 1500 tons refer to light locomotives. Medium sized locomotives, generally used in interstate main and branch line operations, fall between the heavy and the light categories for both the axle and the trailing loads. In addition, Monocoque, Semi-Monocoque and Narrow-Nose locomotives have been defined. A typical Monocoque is a locomotive where the shell or skin acts as a single unit with the supporting frame to resist and transmit the loads acting on the locomotive. Another area of application of crashworthiness standards is freight rail tank cars. The crashworthiness performance of freight rail tank cars is defined as collision performance to protect against shear off of top and bottom valves and fittings in the event of a derailment. The standard requires rail tank car design to include a means of limiting the risks of tank puncture by the adjacent coupler in an override situation. Means to mitigate the risks of tank puncture include: designing a stronger tank head incorporation of a tank shield in front of the tank head moving the tank head back from the headstock of the tank car. For prevailing speeds in Australia, the RISSB standard makes use of the US loading amplitudes for medium locomotives whereas for heavy locomotives and freight wagons noticeably higher loads are specified than in either US or European standards. The higher loading considerations for heavy locomotives and freight wagons are CRC for Rail Innovation November 2013 Page 13

20 possibly due to the larger heavy haul train consist sizes in Australia. The Australian standard has no dynamic performance scenarios considering lateral impact between rolling stock and an obstacle on a level crossing, however static considerations are documented through side loading and roof loading. 4. Worthiness of a CEM design CEM designs should provide a gain of energy absorption while only adding a limited supplementary weight to the original structure to be of any interest. Because the amount of kinetic energy carried by an object with respect to speed bears a non-linear relationship but increases infinitely with the increase in speed (i.e it is unbounded), it would be very useful to examine if there exists a maximum velocity below which a CEM design is worth using. Furthermore, depending on the closing speed at impact, the total energy absorption characteristic of the CEM design may not be linear since it is expected to reach a maximum after a given critical speed. Such an analysis is complex; we here propose a semi-analytical presentation of the worthiness of a CEM device from which a notion of performance is derived. Even though the analysis is simple, it has the merit of providing a general picture of minimum performance requirements that should be expected in terms of energy absorptions for a CEM design to be of any interest. The analysis is based on the assumption that a rail vehicle of mass M, without a CEM, is envisioned for modification by integrating a CEM design. This integration is assumed to provide a change of mass dm on the overall structure which can be either positive or negative. The integration is also assumed to comply with safety requirements such that the occupied areas would remain safe during a collision. It is also assumed that extensive testing has been carried out on the new design such that the difference of total absorbed energy de a with the original design is known for various impact velocities. Such characteristics could be the result of parametric Finite Element Analysis for example. Four types of designs can be devised: CEM type A: Decrease of mass and increase of energy absorption. CEM type B: Decrease of mass and decrease of energy absorption (see Fig. 3). CEM type C: Increase of mass and increase of energy absorption (see Fig. 4). CEM type D: Increase of mass and decrease of energy absorption. It is noted that some designs may decrease the level of absorption at certain speeds while having benefits at other speeds. These will be named Hybrid A-B type or Hybrid C-D type since the increase or decrease of mass is independent of the impact velocity. It is also noted that a pure type D is not worth implementing since it will increase the overall energy to be dissipated without increasing the absorption. Type A CEMs are highly desirable since the decrease of mass ensures the reduction of vehicle energy at any speed. Types B and C are slightly more complex and are discussed in more detail in the following. CRC for Rail Innovation November 2013 Page 14

21 A type C modification is first presented. Figure 4 shows the specific energy absorption variation of the design in relation to the specific kinetic energy load it carries due to the added weight. Because the maximum energy absorption is always bounded, the difference of energy absorption with the original design as well as the specific quantity is also bounded as the impact velocity is increased. However the specific kinetic energy is not bounded and increases nonlinearly as the velocity increases. Therefore, there exists a critical point in the figure where the specific energy absorption improvement characteristic will intersect the specific kinetic energy curve. The velocity at which the intersection occurs cuts the figure into two domains. The left domain is the region where the new design is worth implementing. The right domain is the region where the design will not provide any benefits. A more complex characteristic could be obtained where the benefits are banded between two velocities thanks to a tailored CEM design being triggered above a specific velocity. For the case presented in the figure the band is from 0 to 60km/h. The left intersection is called the Low Critical Speed (LCS) of the design while the right intersection defines the High Critical Speed (HCS). LCS HCS LCS HCS Figure 3: Energy efficiency for a type B modification. Figure 4: Energy efficiency for a type C modification A type B modification is presented in Figure 3. As opposed to type C, the absorption characteristic now represents the loss of energy per loss of mass of the final design. As above, the characteristic is bounded because the loss in absorption is bounded. The specific kinetic energy curve represents the specific loss of kinetic energy due to the removal of the mass. It is not bounded as the velocity increases. In a type B modification the domain is also divided in two, LCS and HCS points are present. However the domain worth implementing is now the domain where the kinetic characteristic is above the absorption characteristic. In other words, for the case presented in the figure, the design is only worth implementing above 50km/h and does not provide benefits at lower speeds. The specific energy gain G s of the design can be quantified by using the following equation: CRC for Rail Innovation November 2013 Page 15

22 G s v 1 2 de v dm v a 2 dm 0 It represents the difference between the specific energy characteristic and the specific gain or loss of specific kinetic energy. A specific gain is obtained when the value is positive, a specific loss occurs for negative values. The performance P of the design is established by evaluating the relative gain of energy absorption in relation to the total kinetic energy of the train at a given velocity and expressed as: Pv dm Gs v dm 2dEa v M v M dmv It describes the percentage of improvement or degradation of kinetic energy absorption relative to the train kinetic energy. As shown in the equation, the heavier the train, the smaller the performance and the faster the train, the smaller the performance. Let s now assume that the variation of absorption with the original design is constant in the range of velocity from 10 to 20 m.s -1 i.e. de v de. We wish to achieve an average improvement of performance p for collisions a a between 10 and 20m.s -1. Integrating the performance over this velocity range and taking the average gives: dm de a p 1 M 100dm Therefore, the design should provide an increase of energy absorption from the original design given by (in kj) de 0.1 pm dm a The analysis for the above three cases, A, B, and C, provides the following results: 1) For a CEM system with a weight increase of 500kg installed in a 40 ton vehicle, the system must be able to provide an increase of absorbing specific energy of 250kJ, equivalent to specific energy absorption of 500 J/kg to improve the performance by 5%. Since the system increased both the absorption and the weight, it is a type C modification. Such a CEM would only provide benefits below an HCS = 114km/h. 2) A CEM with a reduced mass of 500kg would require an energy absorption increase of 150kJ, equivalent to a specific energy absorption gain of 300J per removed kg (300J/kg). Since the modification increased the absorption and decreased the mass, it is a type A modification. 3) Finally, if the design reduced the energy absorption by 100kJ then it would require a reduction of mass of 3 tons to satisfy an average increase of performance of 5%. The system would provide benefits for velocities above an LCS = 20km/h. The modification is of type B. CRC for Rail Innovation November 2013 Page 16

23 It can be seen that the performance can never be positive if the variation of mass is positive and the variation of energy absorption is negative, which indicates that type D modifications are naturally never worth implementing. From this analysis we showed that: improving crashworthiness at low speed is potentially better achieved by increasing the energy absorption of the vehicle even if this leads to an increase of mass. This relates to a type C modification. improving crashworthiness at high speed is potentially better achieved by reducing the mass of the vehicle, even if the level of absorption is reduced. This relates to a type B modification. the energy absorption characteristic and the relevant mass modification can be used to establish expected performance and allow the identification of the velocity domain suitable for its use accordingly. The energy absorption characteristic can be obtained through a parametric FEM study. CRC for Rail Innovation November 2013 Page 17

24 5. Structures of CEM designs Structural design of CEM varies depending on the intended purpose; a few designs are displayed in open literature and are reviewed in this section. It is realised that many commercial-in-confidence designs might well have been developed; the authors therefore do not claim that the information in this section is comprehensive. 5.1 Front end The modification of rolling stock front end is the most effective method of increasing crashworthiness. Xue et al increased energy absorption by 49% through simple shape adjustment of the CEM structure. Vehicle end frames and sub frames can be modified through the use of either weakened and/or strengthened beams, energy absorbers, or a change in the geometric layout of stiffening beams with reference to the draft sill. Xue et al controlled the collapse of a traditional rigid rail vehicle s end structure by increasing the cross sectional area and adding stiffening beams under the crew and passenger areas. Kirk et al weakened an original structure at judicious locations, such as the headstock and side sills, through the use of letterbox cut outs to direct deformations toward collapse areas. Gao and Tian 2007 followed a similar approach by creating holes in the corbel beam. Kirk et al showed that introducing simple letterbox cut outs alone within the end structure was not suitable for crashworthiness systems. Even if the collapse was controlled, only 1.3MJ of energy was absorbed at a collision speed of 65km per hour. To enforce the protection of the driver and passengers, Xue et al increased the cross sectional areas in occupied zones by thickening the draft sill and stiffening the plates covering the trough girders. An energy absorbing beam was integrated in the impact region. Both modifications proved to be effective in controlling the collapse zones of the traditionally rigid Mark 1 structure. Through the increase of surface area, Xue et al improved the level of localised energy absorbed by 47% to 8.65MJ, at a closing collision speed of 150km/h before reaching the driver s rigid ring. Gao et al strengthened the crew compartment by closing over the inner side of the beam, installing two inclined energy transferring beams to transfer the forces past the driver s rigid zone, and by adding two rectangular beams with cut outs in their top and bottom to act as energy absorbers and crush sequence controllers. The structural modifications, involving cut-outs combined with energy transferring beams and energy absorbing structures, improved the energy absorption to 5MJ. CRC for Rail Innovation November 2013 Page 18

25 5.2 Collision posts and corner posts Collision posts and corner posts are critical components that absorb impact force (provided in the US standards). Despite the increase in the cross sectional area of the posts from 3 1/4 (81mm) to 5 1/8 (128mm) since the 1990s, the overall weight of the structure increased by just 113kg (Table 2). This type of specification in the standards accommodates the required increase in force absorption, with a minimal overall increase in body mass. Table 2: 1990s and Current Design Standard vs. State of the Art Requirements (U.S. Department of Transport, 2006) Standard/Requirement Component 1990s Design SOA Design Collision Post 1,334kN ( lbf) at the floor without exceeding the ultimate shear strength shear strength (must be present at 1/3 points along the width of the vehicle) 1,334kN ( lbf) at 457mm(18 in.) above the floor without exceeding the material yield strength Both requirements apply for loads applied ±15 inward from the longitudinal If reinforcement is used to achieve the strength it must extend fully to 457mm (18 in.) and then taper to 762mm(30 in.) above the underframe 2,224kN ( lbf) at the floor without exceeding the ultimate 890kN ( lbf) at 762mm(30 in.) without exceeding the ultimate strength 267kN ( lbf) applied anywhere without yield All requirements apply for loads applied ±15 inward from the longitudinal Strengths must be achieved without failing connections The post must be able to deform substantially without failing the connections Corner Post (must be present at the extreme corners of the vehicle) Lateral Member (must be present between the corner and collision posts just below the cab window) 667kN( lbf) at the floor without exceeding the ultimate shear strength 134kN( lbf) at 457mm(18 in.) above the floor without exceeding the material yield strength Both requirements apply for loads applied anywhere between longitudinal to transverse inward 66.7kN( lbf) applied in the longitudinal direction anywhere between the corner and collision post without yield 1,334kN ( lbf) at the floor without exceeding the ultimate shear strength 445kN ( lbf) at 460mm (18 in.) above the floor without exceeding the yield strength 200kN ( lbf) applied anywhere along the post without yield All requirements apply for loads applied anywhere between longitudinal inward to transverse inward 66.7kN( lbf) applied in the longitudinal direction anywhere between the corner and collision post without yield Include a bulkhead in the opening below the shelf CRC for Rail Innovation November 2013 Page 19

26 5.3 Couplers and anti-climbers Couplers should sustain loads without premature activation of the crashworthiness features of a vehicle s crush zone. The FRA claims that if the pushback system and crush zone are in series, the allowable pushback load could be as high as the maximum crush load. If the two systems are independent a higher pushback load is allowable, provided it does not exceed the strength of the occupant volume (APTA Commuter Rail Executive Committee 2006). Several types of couplers incorporating crashworthiness features are commonly used such as the pushback coupler, the bolt shear mechanism, shear pin mechanism, 45-degree bolt shear mechanism, vertically orientated bolt shear and conventional crushing tube system. Anti-climbers, also known as anti-override devices, are vital in ensuring minimum vertical displacement of the rail vehicle occurs during an impact. They should also account for misalignments between cabs in a moving train, such as lateral, curving, differences in underframe height and wheel wear. Ribbed anti-climbers are the most common form of preventing vehicle override. Retrofitting of traditional vehicles to meet crashworthiness requirements can be expensive. A simple method was proposed by Kirk et al to prevent climbing with traditional train bodies, such as the Mark I structure, through the use of a cup and cone structure. This design makes use of one hole and one cone protruding from the body, each placed on the left and the right of the coupler in the headstock. In the event of a collision, the cones of each rail car penetrate the hole in the facing car, providing both vertical and lateral ad-hoc coupling and stiffening. Other non-ribbed interlocking devices, as found on the France TGV Duplex, exist. The TGV power car is fitted with an hydraulic crushing tube and energy-absorbing device integrated in the coupler, while the trailer car consists of both a hydraulic crushing tube and an energy absorbing device, with an anti-climb device that looks like a C bracket (Kirk et al 1999). 5.4 Wall filling structures Thin walled structures find particular applications in crashworthiness designs due to their energy absorbing capabilities and small volume occupancies. Traditionally, the energy absorption would occur over a narrow section of the wall, for example a 5mm thick wall, resulting in poor energy absorption performances. Thin wall filling structures increase absorption as the deformation of the narrow wall section occurs. Aluminium foam is commonly used. In addition to its energy absorbing capabilities, it helps control the mode of deformation. Zarei and Kröger 2007 showed experimentally that the crush load could be increased by approximately 20kN for quasi-static and dynamic testing of filled thin walled structures compared to their unfilled equivalents. Foam with a density of 230kg/m 3 was the most efficient absorber with a Specific Energy Absorption (SEA) of 28.53kJ/kg. Zarei and Kroger 2007 found that utilising a thin walled structure with the same initial energy CRC for Rail Innovation November 2013 Page 20

27 absorbing capabilities as that of a thin walled tube filled with foam resulted in 19% reduction in overall mass. Inversely, aluminium foam filling inducing a 12% mass increase resulted in a SEA increase of 100% overall. The use of honeycomb lattices for thin wall structures is a relatively recent development. A common application is to fill the honeycomb cells with aluminium foam. While an aluminium 5052 (230kg/m 3 ) filled honeycomb structure was found to absorb 6.7% less than the aluminium foam alone, it resulted in a more controlled deformation of the thin walled structure. Zhang, S. Liu, & Tang 2010 investigated the performance of aluminium columns filled with a kagome honeycomb lattice. A SEA of 19.14kJ/kg was reached. It was found that at equivalent mass, the kagome lattice outperformed aluminium filled sandwich columns. Thin wall filled structures are of particular interest for the management of deformations originating from lateral impacts. As opposed to front end CEM designs where volume constraints are not critical, in-wall CEM systems are restricted in thickness by occupant spaces. As such, the specific energy absorption of in-wall CEM designs should be considered in terms of specific energy absorption (SEA) per volume or thickness of materials. A lateral energy absorption requirement per square meter of panels implemented in a standard would drive passenger car and locomotive designs towards thin and crashworthy side walls in occupied areas. 5.5 Shape and tapering of columns The arrangement of tubes and columns in a sandwiched array, filled with honeycomb and foam to create an effective energy absorbing structure are good CEM elements. In order to attain the maximum SEA possible for the structure, it is important to explore the parameters which will result in a variation of the tube s SEA in addition to those of the fills. The tapering of tubes was found to influence collapse direction, while the cross-sectional area of tubes and columns influences vertical load buckling. The use of tapered tubes was explored by Liu An optimum SEA of 6.15kJ/kg was produced with a tapering angle of 5 o to the horizontal. Zarei & Kröger 2007 studied the energy absorbing profiles of square and circular tubes. Results showed that un-tapered square tubes had an optimum SEA 9.8% lower than that of a circular tube but that square tubes had a 10.4% higher bending resistance compared to circular tubes, demonstrating that simple element geometry can easily tailor crashworthiness behaviours. CRC for Rail Innovation November 2013 Page 21

28 5.6 Occupant space Since occupants do not have seat belts in railway transport, the notion of secondary impacts of the passengers is a critical aspect of crashworthiness. Although out of scope of the R3.114 project, some elements of reflexion are given. Secondary impacts are related to occupants impacting furniture in the cabin due to their difference of deceleration. Therefore, crashworthiness of a vehicle is also related to the mechanical properties of the internal furnishing such as seats and tables. An investigation into the Crashworthy Analysis of the Placentia from Parent, Tyrell, and Perlman 2004 highlighted that the rigid structure of the tables led to some fatal thorax injuries, and that compartmentalization could improve crashworthiness. Shitta-Bey et al focused on low density honeycomb sandwich lattices (<43kg/m 3 ) to dissipate occupant force during secondary impacts. CRC for Rail Innovation November 2013 Page 22

29 6. Finite Element Method Full train consists Finite Element Method (FEM) modelling and associated analysis are rare in academic publications as the work is covered in high detail in industrial designs. The science of FEM is well-suited to crash simulation with explicit solvers that can analyse the large and nonlinear deformations induced. An explicit finite element method analysis of the locomotive collision was performed in Tyrell et al and Stringfellow et al The performance of the design of crush zones for an existing passenger rail cab car was evaluated in Martinez et al and Priante et al A full-rail vehicle explicit finite element model was used to carry out train crashworthiness analysis in Gao & Tian 2007; Xue et al. 2007; Kirkpatrick & MacNeill 2002; Xue & Schmid 2005; Anghileri et al. 2008; Ujita et al. 2003).InAmbrosio 2005 and Pereira et al. 1997, the nonlinear finite elements were integrated with conventional rigid or flexible multibody descriptions in order to build better general vehicle crashworthiness models. Various FEM packages were used in the above publications. Performances and results are generally comparable between packages providing that the models remain equivalent. Typical studies of the impact behaviour of rail vehicles have been based on a collision with a rigid wall or body which is consistent with a standardised model applied in the crashworthiness part of the Technical Standards for Interoperability. As a matter of fact, in a head-on collision between two trains the contact interfaces include the initial impact between the cabs of the two trains. By comparing simulations and analysis it is found that the use of a rigid wall as an impacted obstacle in simulation modelling may mask certain structural weaknesses. However not a single case of head-on collision between two fast-moving trains has been reported in the Australian rail accident data analysis for the time span of 1997 to 2010(Michal et al. 2010). The most common collision is the one where a moving train hits a stationary (or slow moving) train, as is evident from the list of accidents reproduced below; On 23 October 1997 freight train hitting a stationary coal train (Beresfield, NSW). On 18 August 1999 passenger train hitting a stationary freight train. On 26 November 1999 freight train hitting a stationary freight train. On 05 June 2001 passenger train hitting a stationary passenger train. On 18 June 2002 passenger train hitting a slow moving passenger train. On 03 February 2003 passenger train hitting a stationary passenger train. On 16 March 2003 passenger train hitting a stationary freight train. On 25 February 2004 passenger train hitting a stationary freight train. On 19 January 2005 passenger train hitting a slow moving freight train. CRC for Rail Innovation November 2013 Page 23

30 As the above accidents all involve a moving train colliding with a stationary or near stationary train, a rigid wall impact has been considered as a conservative assumption for analysis. The numerical model studies conducted for this report have thus focussed on rigid wall impact cases. The preliminary numerical models developed and reported for this particular project have been validated by comparing them with established studies (Xue et al. 2004, 2005) before extending our studies to conduct the lateral impact analysis. 6.1 Structure The development of the crash model using Hypermesh (a well-known proprietary FEM modelling package) in this research report involved the design of the bogies and their interface with the main body and the rails. The system was modelled with limited details and the FE analysis is based on making assumptions relevant to the level of structural information that was available for this work. As the impact forces and deformation is mainly in the longitudinal direction of the vehicles, these simplified approaches are usually adequate.(xue, X, et.at, 2005) The bogie-rail interaction was taken into account through contact conditions between the rails, the wheelsets and the bogie frame. Known rail and wheelset geometries as well as observed bogie frame geometry and connections were used as a basis. The bogie consists of two wheelsets, connected to the bogie frame. Each wheelset is made of two wheels fixed to the wheelset axle. Wheelsets are mounted with bearings on the bogie. The bogie frame is connected to the main body via the centre ball or bowl located at the centre of the bogie frame. (a) rail (b) wheelset (c) bogie (d) wheel-rail contact (e) bogie-wheelset assembly Figure 5: Finite element model of a bogie assembly CRC for Rail Innovation November 2013 Page 24

31 The modelling of the train bogies was performed in stages to validate the rails, wheelset and bogie interactions. The rails were based on a AS60 profile and spaced at Cape gauge (1,067mm) (Figures 6(a), (b) and (c)). The wheelset consisted of two wheels, modelled as ANZR standard wheel profiles with a diameter of 960mm, and an axle connecting the wheels together (Figures 6(b) and (d)). The bogie frame geometry was developed from observed geometry of typical passenger trains. No suspension was included for modelling simplicity. The bogie frame consists of two box sections connecting to the outside of the wheelsets and a central section linking the two outer box sections to the frame (Figures 6(c) and (e)). The rails, wheelset and bogie frame were meshed with solid tetrahedral and quad elements (see Figure 5 (a) to (c)). The components were given elastic material properties with a density of 7,800 kg/m 3 and a Young modulus of 210 GPa, representative of standard steel. A single wheelset-rail interaction model was integrated (Figure 5 (d)) by considering surface contact constraints and a frictional coefficient of 0.5. Fixed boundary conditions of the rails and gravity forces were applied. Assuming a fully established braking state with non-spinning wheels, an initial velocity of 20 m/s was prescribed to the wheelset to generate sliding on top of the rails and observe the contact s physical correctness. Following this validation, the wheelset axle-bogie interaction was considered using similar contact conditions to simulate a bearing connection with a friction coefficient of 0.1, see Figure 5 (e). An initial velocity of 20 m/s was applied to the bogie assembly in a non-braking state to enable motion along the rails and validate the system s dynamics. (a) (b) CRC for Rail Innovation November 2013 Page 25

32 (c) (d) Figure 6: Finite element body consist The model of the train structure was developed separately from the bogie-rail model. The layout of the leading cab car of the Electric Multiple Unit (EMU) used in this study is shown in Figure 6 (a) to (d) based on the descriptions provided in (Xue et al. 2004, 2005). The driving cab is typically streamlined and has relatively large spaces in the car end areas. There are often also support facilities such as toilets, luggage rack, electric distribution cabinets and water supplies. An apparent advantage of this layout is that it provides opportunities for the design of energy dissipation zones in the vehicle end areas. For simplicity the car body was assumed to be made of steel alloy. The structure is composed of underframe, two sidewalls, roof, one end wall and one cab structure. Figure 6 (c) shows the structural model of the front end of the leading cab car. It is a simplified description of the model used in (Xue et al. 2004). The figure shows a general arrangement of the leading car front end structure at the floor level. The cab panel is a reinforced shell with strip stiffeners forming a highdensity frame. The cab-end structure is composed of the floor and the cabin (green), cross members and frames posts (shown in yellow in Figure 6 (b)), the side panels (red) and the headstock (light blue). In earlier studies (Xue et al. 2004, 2005, 2007) it was noted that the finite element impact modelling of vehicles is based on (i) half-width/full-length and (ii) half-width/half-length (quarter) structures, depending on the symmetry of the vehicle. It has also been noted that such models have been successfully used in the analysis of rail vehicle structures subjected to static loading, such as the proof load, but have not been validated for dynamic loading. In collisions, the substantial kinetic energy of an impacting rail vehicle causes the vehicle structure to suffer large plastic deformation as the energy is dissipated in structural collapse. In such conditions, the crashworthiness assessment becomes an interpretation of permanent, high strain rate, deformation behaviour - an exercise of a different nature compared to Proof load assessments where deformations are kept within the yield limit. CRC for Rail Innovation November 2013 Page 26

33 One common example of such large and permanent deformations is the downward bending of the draft sill that has been identified as a main weakness in the impact responses of conventional rail vehicles. The bottom of the draft sill near the bolster must be shallow to avoid interference with the bogie below. The front part of the draft sill must be relatively deep to contain the coupler and draft gear. The draft sills also have to sustain and transfer coupler loadings, which act on the position behind the draft gear. To satisfy these conditions, draft sills are generally formed as box shapes in their cross section with a fish-belly shape in the longitudinal direction. The front height dimension of the draft sills depend on the installation of the draft gear, with a gradual reduction to the bolster. As a result of its moment of inertia and geometry, the draft sill can sustain large lateral and longitudinal forces and is beneficial in that it can absorb impact related energies. However, it is not a strong structure in the vertical direction. The reducing height makes the draft sill liable to bending at its rear end, near the bolster. The downward bending observed in the simulation of collisions between two cab cars may go some way towards explaining the over-ride phenomenon that has manifested itself in many real life collisions. The provision of anticlimbing devices now limits the consequences related to the weak vertical performance of the draft sill. Each bogie assembly was attached to the train structure at four nodes using rigid links, and gravity was applied to the train structure. The combined model had a total of approximately 1.4 million solid elements, 70,000 shell elements and approximately 400,000 nodes. Due to the large number of elements in the model the computing time of the model was large and refinements were carried out to improve the situation. To refine the model a number of different solutions were investigated including: use of different type of mesh elements and remeshing of the rails, wheels and bogie analysis of the contact complexity: Is a wheel-rail interface model required? simplification of non-critical parts such as bogies ensembles. The surfaces of the head and web of the rails were remeshed using quad shell elements. The foot of the rail was not meshed to reduce the number of elements in the model. The mesh size varied across the head of the rail to accurately form the profile of the rail. The wheelset surfaces were remeshed using quad shell elements with a minimum mesh size of 20mm such that the curved profile of the wheel flanges could be accurately represented. The surfaces of the box frame were remeshed using quad shell elements with a minimum mesh size of 50mm. Following the refinements to the model, the total number of elements had reduced to approximately 150,000 shell elements, with the number of nodes reducing to approximately 150,000. CRC for Rail Innovation November 2013 Page 27

34 6.2 Wheelset: Rolling The modelling of the wheel/rail interaction leads to a large number of elements and carries a large computational load. This interaction can be simplified to various degrees depending on the followingscenarios: flying train: there is no explicit interaction between the train and the rails. The motion of the vehicle is simply constrained in at least one, and up to four, directions: one downward and eventually one up plus the lateral directions sled train: there is an explicit interaction of the train/rail at the wheels but the latter are not spinning. The contact at each wheel is localized to the same region. Vertical and lateral contacts can be considered normal train: there is an explicit interaction train/rail at the wheels. The wheels are rolling. The contact at the wheels is not localized. Vertical and lateral contacts can be considered. The last model is the most refined and what was initially implemented in FEM. The model is general and can handle both in-plane and off-plane curvatures of the rails, providing that the bogies include some rotational degrees of freedom. It requires a large number of FEM elements to accurately represent the interaction. This constraint is partly due to the need to accurately model the geometry of the contact and partly due to the convergence of contact algorithms. The use of a fine mesh in the contact region is only useful to describe the geometry of the contact conditions more accurately. A rigid wheelset/bogie ensemble is considered to be valid since its stiffness is considerably higher than CEM devices. Whether or not the wheels need to be modelled as rolling is a question that must be understood as: Does the rotational kinetic energy of a wheelset have a critical influence on the total energy needed to be absorbed during the impact? Wheelsets are large components. The rolling of the wheels has both forward and angular momentum that is transferred at the moment of impact. Traction and braking forces can also change the forces at the wheelrail contact and influence the overall mechanics of the impact. However, this transfer of rotational energy to forward momentum can be bounded by using an analytical model and for the purposes of the model, traction and braking torques will be ignored (the train is assumed to be in a rolling state). The transfer of energy from rotation to translation occurs at the wheel/rail interface is subject to the friction limits of the wheel-rail contact. Amonton s law of friction provides a model that relates the normal force F n to the maximum longitudinal force at the wheel tread F t. F t F n During a head-on impact phase, the dynamics of the vehicle will be affected by the vertical position of the impact point in relation to the centre of mass. An impact point above the centre of mass will increase the loading force at the rear bogie and may lift the front bogie from the rails. Conversely, an impact point below the centre of mass CRC for Rail Innovation November 2013 Page 28

35 will increase the loading at the front bogie and may lift the rear bogie. Therefore the evolution of the maximum traction force at the wheels is dependent on the speed at impact, the geometry of the vehicle and the geometry at the impact point. To make the analysis possible analytically, It is assumed that the wheels do not slip or lift off at any instant. Such idealization bounds the effect of the spinning wheels on the final impact force by maximizing the level of traction forces at the wheel/rail interface. The analysis was based on a wheelset having two wheels each being 1m in diameter and weighting 400kg. The wheelset axle was 0.2m in diameter with a mass of 350kg. The total mass of the wheelset was 1150kg. All the elements of a wheelset are assimilated as disks to express their individual moment of inertia. I is the moment of inertia, M the mass of the disk or cylinder and R the radius. The axle was modelled using the moment of inertia of a full disk. The moment of inertia of a wheel was calculated by using the equation for a thin wall disk. Although its geometry does not strictly apply to the definition of a thin wall, the use of this model bounds the level of kinetic energy that the wheel can store. The approximated total moment of inertia of the wheelset is as follows: I t = 1 2 M R 2 2 a a 2M w R w The kinetic energy of the wheelset is divided into two quantities: the translational energy E kt and the rotational energy E kr. For a rail vehicle that moves at a speed v, the equations are: E kr = M R a a R 2M 2 wv 2 1 w 2 M v 2 R Therefore, a vehicle made of n wheelsets having a total mass M T and rotational M R stores the following kinetic energies in translation and rotation: E kt = 1 2 M 2M a wv 2 E T = 1 2 M T v 2 E R = 1 2 nm R v 2 To bound the effect of the rotational kinetic energy it is assumed that a total transfer of rotational kinetic energy into translational energy occurs. Since such a transfer is carried by the instantaneous tread force at the wheel/rail interface, the variation of rotational momentum can be related to the power of the tread force. de R dt = nm R a(t).v(t) F T (t).v(t) CRC for Rail Innovation November 2013 Page 29

36 Similarly, the time variation of translational energy momentum is equal to the sum of the power of the tread forces on the wheels due to the impact force. Substituting the tread force F T leads to the equation of the impact force. The latter equation is a function of the acceleration taken at the centre of mass and a function of the translation mass and equivalent rotational mass. de T dt = M T a(t).v(t) F T (t) F I (t).v(t) F I (t) M T nm R a(t) A quantitative measure of the instantaneous effect of the rolling wheel is derived through the ratio of the tread force and the impact force. F T (t) F I (t) 1 1 M T nm R This ratio is a quantity independent of the initial velocity or time. It is an upper bound measure of the contribution of the rotational kinetic energy onto the impact force under the assumption that the total energy of the wheels is transferred into forward momentum. As shown in Figure 7, the influence of the rotation of the wheelsets can be considered linear for a ratio nm R /M T below or equal to approximately 0.1. At this level, the ratio of F T /F I is equal to 10%. It can be safely assumed that the spinning of the wheels is a negligible quantity in such conditions and that it would not affect the outcome of a simulation significantly. Theoretically the effect of rotation accounts for almost half of the actual impact force at any time of a collision phase. Figure 7 Contribution of the traction force to the impact force CRC for Rail Innovation November 2013 Page 30

37 For the presented cases, the lighter vehicle had a mass of approximately 38t, leading to a maximum possible contribution of the wheelsets on the total kinetic energy of less than 7.8%. Therefore, it appears that the modelling of the rolling wheels is not an essential requirement in the case of a head-on impact. Moreover, the relatively high stiffness of the bogie ensemble in respect to crash energy management systems points favourably toward the use of a rigid rail/wheel/bogie sub-system for modelling the crashworthiness scenario. This modelling approach, equivalent to considering the train as a sled on top of the rails, was used in the following crash analysis to reduce the complexity of the model and preserve the computational resources for the more demanding collision zones. 6.3 Crash analysis The FEM models established in Hypermesh in this research work were solved using RADIOSS. RADIOSS is an explicit FEM solver for linear and non-linear simulations. It is used in noise and vibration performance, crashworthiness, safety, and manufacturability of designs in order to bring innovative products to market faster. The model has been developed for an initial closing speed of 20 m.s -1 against a rigid wall. Since each structural component does not act in isolation, the progressive deformation of the structure as a whole depends on whether or not the collapses in any given vertical plane of the structure are synchronised. By allowing a consistent crush rate over the full height of the structure, downward and upward bending can be limited. An example of this effect is in the body bolster region as shown in Figure 8. The bolster does not crush under the impact force and transmits the impact force directly to the central underframe. The upper cab panel in the same region is not so rigid and therefore deforms under the impact force. As a result, the deformation in the underframe happens in the vehicle s bogie centre area, while the deformation in the upper panel occurs in the vehicle end area. The upper structure will produce a resisting force on the surface of the side sill of the central underframe. As a result, the central underframe will bend. This situation easily causes a shear fracture along the connecting line between the underframe and sidewall. This latter phenomenon has been observed in several collisions and is known as unzipping (Xue et al. 2004). CRC for Rail Innovation November 2013 Page 31

38 Figure 8: Pitch down of the front end due to uncontrolled deformation Table 3: CEM design study parameters Imposed Mass 30T Imposed Mass 20T Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Floor Main Body Side Beams Headstock End Bars in CEMS Middle Bars in CEMS Cross Bars in CEMS Total Mass (Kg) Body Mass (Kg) Energy (Joules) Energy /Body Mass Several structural cases were evaluated in order to improve the crush behaviour of the frontend of the train and minimise spurious bending deformations. The modifications were made by changing the thickness of the components. Table 3 presents the thickness used for each component of the assembly. The evolution of the design led to a mass range from 38 tonnes up to 52 tonnes. This range equivalently translates to a range from 7.7MJ to 10.4MJ of kinetic energy to dissipate. The failure mode of each case is shown in Figure 9. It can be observed from Table 3 and Figure 9 that CEM can be achieved by simply varying the stiffness of structural elements of the car s cabin. CRC for Rail Innovation November 2013 Page 32

39 Driver cabin Passenger area Undeformed Shape Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Figure 9: failure modes of the cab car for different structural cases Case 3 and case 7 show severe damage to the passenger areas while case 5 dissipates kinetic energy with minimal damage to the driver cabin and passenger seating space. The difference in structural system between the best performing Case 5 and the other cases is only marginal which shows the significance the structural design can make to the crashworthiness of rolling stock. The progressive collapse of a poorly designed CEM structure is CRC for Rail Innovation November 2013 Page 33

40 Crushing Force (MN) Crash Energy Management Systems for Australian Rolling Stock shown in Figure 10. The crush characteristic for such a structure flattens instead of providing restoring forces as shown in the last picture of Figure 10(b). 0ms 20ms 40ms 60ms 80ms 100ms 120ms 140ms 160ms 180ms 200ms (a) Progress of deformation of front end Time (ms) (b) Crush characteristics Figure 10: Crush deformation progress of a poorly performing structure In this case, even the provision of anti-climbers would not solve the issue as it is the aft part of the crunching zone that collapses. This example illustrates that a proof loading of anti-climbers may not guarantee inline CRC for Rail Innovation November 2013 Page 34

41 deformations. It is the complete deformation chain that should be evaluated against a measure of maximum allowable vertical bending. It has been shown (from the FEM results) that the deformation process of the car s cabin can be unstable. Both the draft sill and side sill can bend during the crash. The CEM structure should enhance the energy dissipation capability without decreasing the stability of the process. The shape of component cross-sections is critical for the crush behaviour. Changing the cross-sections of components to enhance their crush performance should minimise the reduction of the bending moments of inertia. Elements able to enhance pitch bending at a limited weighting cost are generally beneficial. The crush progress of the optimised car s cabin structure is shown in Figure 11. As is typically done (Xue et.al. 2005) the simulation is portrayed below at intervals of 20 ms, from the beginning of the crash until the whole front structure has collapsed. In this case the deformation follows a more desirable inline progressive pattern. It is closer to the recommendation of the RISSB standards: the Crash Energy Management System (CEMS) shall provide a controlled deformation and collapse of designated sections within the unoccupied areas of the vehicle to absorb collision energy (RISSB 2008). 0ms 20ms 40ms 60ms 80ms 100ms 120ms 140ms 160ms 180ms 200ms Figure 11: Crush deformation progress of the corrected structure CRC for Rail Innovation November 2013 Page 35

42 Force (MN) Energy (J) Crash Energy Management Systems for Australian Rolling Stock Figure 12 shows the total energy absorbed by the optimised cabin design over the whole period. Once the cab structure has totally collapsed at about t = 200 ms, 6 MJ of impact energy has been absorbed from the initial 8 MJ of kinetic energy. The response, shown in Figure 12(a) is comparable to the energy dissipation response studied by Xue et al and provided in Figure 12(b). Figure 12(c) shows the crush characteristic, i.e. force displacement characteristic, of the cab car structure. Excluding the impulse at the first contact point of the crash, the forces are in the region of 2 to 4 MN. This response is comparable to the studies conducted by Xue et al (Figure 12(d)) Kinetic energy Internal energy Time (ms) (a) (b) Time (ms) (c) Figure 12: (a) Kinetic and internal energies. (b) Gross energy dissipation from Xue et al (c) Crush characteristic of the cab and (d) Force-deformation of the can end structure from Xue et al The optimised structure was assessed using a second FEM commercial package (LS Dyna) as a supplementary assessment of the correctness of the deformation modes. Incompatibilities between the mesh file formats meant that the models were not strictly equivalent and variations exist between the two simulations. The kinetic energy profiles and impact forces were nonetheless similar and the general deformation mode qualitatively equivalent as shown in Figure 13. (d) CRC for Rail Innovation November 2013 Page 36

43 Lateral collisions Figure 13: Deformation mode from second the software package Studies of lateral impact of railroad vehicles is non-existent in the literature, making the development of a model and its validation difficult. Due to the thin and largely unsupported side walls and the presence of large window surfaces and the variations of designs between single and double-deck passenger cars, the pertinence of the current model is questionable. These factors limit the usefulness of the simulation. A fundamental point to understand is that the lateral impacts normally involve a significantly lower amount of kinetic energy to be dissipated compared to frontal collisions. Lateral collisions are due to road vehicles which, with the exception of heavy road trains, typically involve less than 1MJ of kinetic energy. The context is therefore very different to a longitudinal train collision scenario. Contrary to longitudinal impacts, there is unfortunately little room to allow the structure to deform without directly affecting the occupied areas. Several crashworthiness strategies could be envisioned to limit casualties: 1. The tank approach: side walls can be designed to be extremely rigid such that very little penetration is possible. Although the volume can be guaranteed to remain within acceptable bounds, this is likely to carry a significant weight penalty, unless specific materials or structures are used. This approach is only viable if the level of lateral accelerations remains within acceptable bounds. 2. The membrane approach: the large surface area available from the side walls could be used positively to spread the load over the contour frame. By arching the side panels vertically and longitudinally to an appropriate radius of curvature the stress would spread over a larger area allowing for an increase of dissipation of energy while keeping the in-wall deformation to a low level and consequently reducing the lateral acceleration. 3. Increasing occupant space: Specify a minimum distance of the seats from the side walls. The distance would need to be established through extensive simulations and testings and may reduce the occupant volume and the number of seats. CRC for Rail Innovation November 2013 Page 37

44 However, the development and analysis of such solutions are tasks that require a coordinated effort from the industry such that: a realistic structure is used throughout the development engineering and economics constraints of the design are kept within reasonable levels the lateral design requirements are agreed upon by all the industrial stakeholders for inclusion in a standard human resources and knowledge base provide an adequate foundation to build on. For instance, the development of the static and performance European standards EN12663 and EN15227 was the outcome of projects spanning over 4 to 5 years. Extensive modelling and testing were carried out and the involvement of the industry appeared significant. Developing a lateral crashworthiness system and standard is expected to carry a significant workload. As an illustration of a possible parametric study, the model was subjected to several lateral impacts. The typical scenario is shown in Figure 15. For simplicity the wheelsets were replaced with axial elements oriented vertically. Figure 14: Scenario of lateral collision Figure 15: From left to right, collision scenario at 5, 12 and 19m from the front edge CRC for Rail Innovation November 2013 Page 38

45 (a) (b) (c) Figure 16: Deformation modes for a lateral impact at (a) 5m, (b) 12m and (c) 19m The lateral impact from a moving body of 5 tons was applied at 5m, 12m and 19m from the front edge of the 26m long passenger cab at 1.5m above the cab floor level as shown in Figure 15. Figure 16 (a), (b) and (c) show the external (left) and internal (right) views of damage caused to the car body for the lateral impact at 5m, 12m and 19m from the front edge. Deformation of the occupant volume occurred in each case. The impact at 12m from the front edge case incurred the largest inward deformation. This result is driven by the fact that no significant supporting members exist in the wall or the floor in this area of the model. The two other cases are closer to the bogies and bolster, areas that are noticeably stiffer than the rest of the structure. The use of the CEMS designs presented earlier in longitudinal collisions had little effect on the outcome even for the impact close to the driver s area. No tipping of the rail vehicle occurred in the simulations and the overall lateral motion remained low. The figures that follow show the vertical (Dx), lateral (Dy) and longitudinal (Dz) displacements of the wheelsets top due to the lateral impact on the car body. For each direction it can be seen that the displacements are all below 30cm and generally bound to a couple of millimetres. CRC for Rail Innovation November 2013 Page 39

46 Figure 17(b) bottom left (5m case) and Figure 19(b) (19m case) bottom right, respectively, show the lateral displacement of the head and aft front wheels. Both figures show a lateral displacement of about 30cm (average of the head and aft front wheels) occurring in approximately 80ms, equivalent to a lateral acceleration of less than 10g. As expected, the impact incurs a high level of acceleration in those regions as the bogie and bolster assembly is particularly rigid. However, between 70ms and 100ms, the acceleration level decreases rapidly in the front and aft impact cases due to the constraints of the rails on the wheels that limit the lateral motion. Providing that the train does not tip, it can be expected that casualties will be localised around the area above the front bogie in the front impact case and aft bogie in the aft impact case. Figure 18(b) (12m case), related to the lateral mid length impact, shows a very different outcome. Because the structure is compliant in the middle of the car, a greater amount of energy is absorbed. Because the wheelsets are distant from the impact they only move by a few millimetres (~25mm). The train is less likely to derail in this situation and the level of lateral deceleration is kept to a lower value compared to the front and aft impact cases. While the overall acceleration is less of an issue for passengers, the change of volume in the region of impact can be a serious threat to the passengers in the vicinity of the wall. Measuring the level of injury experienced by the passengers would require the use of dummies in the simulation and in full scale tests. These three cases illustrated well the various mechanisms possible in a lateral collision by highlighting the advantages and drawbacks of rigidity, compliance and diminution of volume in impacted thin walls with occupants in close proximity. The simplicity of the model could not realistically provide an accurate answer nor could it give an ideal side structure but it could identify the basic mechanisms that need to be considered and balanced. CRC for Rail Innovation November 2013 Page 40

47 (a) Vertical displacement (b) Lateral displacement CRC for Rail Innovation November 2013 Page 41

48 (c) Longitudinal displacement Figure 17: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 5m of the top of the wheels CRC for Rail Innovation November 2013 Page 42

49 (a) Vertical displacement (b) Lateral displacement CRC for Rail Innovation November 2013 Page 43

50 (c) Longitudinal displacement Figure 18: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 12m of the top of the wheels (a) Vertical displacement CRC for Rail Innovation November 2013 Page 44

51 (b) Lateral displacement (c) Longitudinal displacement Figure 19: Wheels displacements are shown clockwise head back, aft back, aft front and head front for a lateral impact at 19m of the top of the wheels (a) Figure 20: Structural configuration of (a) intermediate vehicle and (b leading vehicle (Xue et al. 2007) (b) CRC for Rail Innovation November 2013 Page 45

52 7. Crashworthiness enhancement Structural modifications can enhance the energy dissipation capability without compromising the stability of the deformation process. From other studies (Xue et al. 2004, 2005, 2007), it is suggested that the structural improvements should focus on the following areas: for driver safety, a strong tube can be created around the driver region on both the underframe and cab panel to counteract the structural weaknesses, the draft sill can be enhanced at its rear end and at the position where its height changes to increase stability, the cross-section of the side sill can be changed to increase the ability of the structure to absorb energy and increase stability, energy absorbers can be added at the two sides of the draft sill. In the above study, the crashworthiness of the leading vehicle has been assessed. However, there are noticeable differences in the structural characteristics of an intermediate vehicle when compared to a leading vehicle (Xue et al. 2007). In Xue et al. 2007, the authors evaluated these differences and came to the following conclusions: difference in impact interfaces: The lead end of a train will usually impact another train or another object, while two coupled vehicles only experience the impact with each other. difference in vehicle end structures: The trailing end of the leading vehicle and both ends of intermediate vehicles feature different structural designs and therefore different opportunities for special structural components in the area near these doors. difference in impact sequences: The lead end of a train is generally the first impact interface. It is worth noting that for static requirements the end structure is designed to connect and support the headstock, the end beam and the gangway. Performance under plastic deformation is not considered in conventional designs. The impact forces stress the structure of the vehicle differently and prominently around the headstock, this means that the end structure in the gangway area becomes a key region for crashworthy design (Xue et al. 2007). Further, to allow for the door opening, the results in different end structures in intermediate vehicles. This factor means that the weakening of the structure door to the door cavities must be compensated with appropriate underframe strength. CRC for Rail Innovation November 2013 Page 46

53 Failing components and the sequence of bending are different for intermediate vehicles compared to leading vehicles (Xue et al. 2004, 2005, 2007). The bending of the leading vehicle begins with the draft-sill whereas the intermediate vehicle starts to fail due to the bending of the solebars while the draft-sill bends at a later stage. The end structures of the intermediate vehicle and of the leading vehicle differ: shapes: the intermediate vehicle has a traditional tube shape with the end-wall in a vertical plane, while the leading vehicle will have a streamlined shape. The crash responses of these two vehicles is therefore very different. the doors can be in different locations. The doors are located at the end in the intermediate vehicle and above a bolster in the leading vehicle. These details influence the crash responses in the door area. The results from (Xue et al. 2007) indicated that, the intermediate vehicle is weaker than the leading vehicle in the area near the door opening. The door opening has two consequences: the door area deforms separately as it is not part of the roof or underframe and is weak in resisting a vertical force or moment. the door opening interrupts the impact force transmission path in the sidewall. Viewing the structure as a whole, the intermediate vehicle appears stronger than the leading vehicle. Due to the size and shape of the end contact areas the collapse of the intermediate vehicle is more stable.. This deduction is based on the assumption that the leading vehicle has a tendency to bend downward. (Xue et al. 2007) In a train collision, the intermediate crash interfaces are different from the lead vehicle crash situation. i A range of obstacles can be encountered by the leading vehicle, while the intermediate impact interface are much more regular concerning two identical vehicles. (Xue et al. 2007). Multibody modelling is well suited to evaluate such complex mechanisms where several interconnected cars are subject to impact. They allow for a quick assessment of the conditions that lead to derailments. CRC for Rail Innovation November 2013 Page 47

54 8. Application of CEM designs to the prevention of derailments This section examines whether or not the CEM fitted vehicle consist can prevent derailment as the CEM dissipate the collision energy in a controlled manner. Wheel rail contact force is monitored for assessing the derailment potential. To minimise the computational time, rigid body simulation has been performed using spring and damping to model the crush zones. 9.1 Multibody formulation Multibody dynamic models can be formulated in one, two or three dimensions. One-dimensional (1D) models consider the train to be constrained longitudinally (i.e along the track). The two-dimensional (2D) models consider both longitudinal movement and the lateral buckling of the train. Three-dimensional (3D) models also consider vertical displacements (i.e over-riding of vehicles in a train) (Sun et al. 2012). Milho et al. (2003) presented a validated multibody model for the design of train crashworthiness components. A design methodology for crashworthy structures was presented in Dias & Pereira (2004). Deterministic and evolutionary algorithms were linked with simplified models based on multibody dynamics formulations, and were used in the conceptual design to obtain the best characteristics of the crashworthy train structures. A three-dimensional collision dynamics model of a multi-level passenger train was developed to study the influence of multi-level design parameters and possible train configuration variations on the reactions of a multi-level car in a collision (Priante et al. 2006; Sun et al. 2012). A collision dynamics model was used to study the kinetic and dynamic response of the individual crush zone components and the resultant car body motions prior to the tests and good agreement was obtained between the model and the test results (Jacobsen et al. 2004;Sun et al. 2012). A multi-body model was designed to carry out the crashworthiness analysis of the train-to-train collision (Parent et al. 2004; Mallon et al. 2008) and used to be comparable with a 30 mph (48.3 km/h), full-scale, train-to-train crash energy management (CEM) test (Priante & Martinez 2007). In this report, a three-dimensional model for crashworthiness analysis has been developed in GENSYS (GENSYS is a software tool for modelling vehicle-rail dynamics. Furthermore, GENSYS is a general multi-purpose software package for modelling mechanical, electrical and/or mathematical problems). The locomotive and each wagon are analysed as a fully detailed multi-body dynamics model with non-linear springs and dampers representing the secondary and primary suspensions that are used to connect the car body, bogies and wheelsets. The wheel-rail contacts are taken into account in order to observe the conditions leading to derailment (Sun et al. 2012). The non-linear springs and dampers representing the couplers are used for the connections between the locomotive and the first wagon and the inter-wagon attachments. The locomotive is modeled with and without CEM designs. The crashworthiness investigation has been carried out on the case of a longitudinal collision with a fixed obstacle. CRC for Rail Innovation November 2013 Page 48

55 9.2 Modelling of components The components of the car body and the bogie frames are each modelled as single, fully rigid mass with 6 degrees of freedom (DOF). The approach to modelling is the same as published in Sun et al. (2012) and the description is repeated below. The connections (the secondary suspensions) between the car body and the bogie frame for each bogie include: two vertical coil spring elements one spring element for the anti-roll bar, and one spring element and one damper with series flexibility for the traction rod in the direction specified by the attachment points of the coupling one lateral and two vertical bumpstops, two vertical viscous dampers, and two lateral viscous dampers and two yaw dampers with series flexibility in the direction specified by the coupling's attachment points respectively. Figure 21(a) and (b) provide detailed illustrations of these arrangements. (a) (b) Figure 21: (a) Passenger car elements and (b) details of the bogie and rail arrangements for multibody modelling CRC for Rail Innovation November 2013 Page 49

56 The wheelset is modelled as a single mass with 5 degrees of freedom (the pitch rotation is disregarded). The connections (the primary suspensions) among one bogie frame and two wheelsets include: 12 spring and damping elements in the three X, Y and Z directions two lateral and four vertical bumpstops, and four vertical viscous dampers in the direction specified by the coupling's attachment points. The rail is modeled as a massless block under a wheel. The rail is connected to the track via lateral and vertical springs and dampers. The track is modeled as a mass block under a wheelset. The track is allowed to have translations in the lateral and the vertical directions and rotation about the longitudinal direction. The connections between the track and the ground include: two vertical coil spring elements two vertical dampers with series flexibility and one lateral damper with series flexibility. The wheel and rail profiles shown in Figure 22 are chosen for the modelling of the contact characteristics. Instead of a one or two wheel-rail contact points approach, three contact points can simultaneously act at the wheel-rail interface. The calculations of tangent creep forces at the three wheel-rail contact surfaces are done in a lookup table calculated using Kalkar creep theory. Figure 22: Wheel and rail profiles CRC for Rail Innovation November 2013 Page 50

57 Crush force (kn) Crush force (kn) Crash Energy Management Systems for Australian Rolling Stock Crushing Tube Figure 23: Crush zone modelling Push back coupler Crushing tube Structure Crush length (mm) Figure 24: Idealised crush zone s force and crush length characteristics for (a) the HE zone and (b) the LE zone Push back coupler Structure Crush length (mm) The passenger areas are considered rigid and represented as single masses with 6 degrees of freedom. The components of the crush zones push back coupler, crushing tubes and structure energy absorbers are massless, see Figure 23. The obstacle is modelled as a fixed rigid body. Generally, the crush zones are designed to absorb the impact energy during the collision via the plastic compression deformation. The design of crush zones, supposed to absorb the entire impact energy without affecting the occupants, depends on factors such as the vehicle crash speed, the weight, the number of vehicles in the train, the crush zone s crush force and length characteristics (Sun et al. 2012). The initial idealised crush zone s crush force and length characteristics for the HE CRC for Rail Innovation November 2013 Page 51

58 and LE zones are chosen as shown in Figure 24. If the rigid space is affected, a mechanical stop with quite a large stiffness (e.g. 50MN/m) is assumed. The components of the HE and LE zones include one push back coupler, two crushing tubes (no crushing tubes in the LE zone) and the structure energy absorbers. Their crush lengths are taken as 300mm for the push back coupler, 400mm for the crushing tubes and 400mm for the structure energy absorbers. The push back coupler, the crushing tube and the structure energy absorbers were modelled as an element with a spring (k) in series with a friction block (F f ) in parallel with a damper (c) in the longitudinal direction. In Figure 25, the spring with stiffness coefficient is serially coupled with the friction block with a sliding force, which means that the coupling force through the element is the same for the spring part as in the friction part (Sun et al. 2012). k F f c Figure 25: Energy absorber modelling CRC for Rail Innovation November 2013 Page 52

59 Coupler for ce (M N) Coupler for ce (M N) Crash Energy Management Systems for Australian Rolling Stock Deflection (m) Deflection (m) Figure 26: Coupler modelling (a) at slow impact and (b) at shunt impact If the deformations of the ends of the element are smaller than, no sliding motion is assumed to take place in the friction block; instead all motion will take place elastically in the spring part. If the deformation of the element exceeds, the friction block will stretch or compress to ensure that the force over the spring part does not exceed the force. For example, for the modelling of push back couplers, is equal to 2000 kn and is selected to be 500 MN/m. The damper in Figure 25 represents the internal damping of the metal and a small value (0.2 kn/(m/s)) is chosen for this purpose. In addition, a two-dimensional friction block is used to model the friction between the end of the absorber and the surface of the obstacle. (Sun et al. 2012). The relationships between the coupler force and coupler deflection are shown in Figure 26. These are used for the longitudinal coupling between locomotives and wagon as well as wagon and wagon. 9.3 Simulations Several integration methods are possible to compute the motion. The two step Runge-Kutta method with step size control is selected for the simulations. The integrator has variable time steps between the selected maximum of 1 millisecond and minimum of 1 microsecond to ensure the numerical stability, and the length of the step is CRC for Rail Innovation November 2013 Page 53

60 calculated based on how fast the error increases or decreases between two consecutive time steps. (Sun et al. 2012). The level of initial symmetry of the collision was adjusted to examine the derailment potential and the effect of CEM designs on derailments. To do so, the obstacle s centre line was shifted by 0.5 m from the consist s centre line and the closing speed was set to 80km/h as shown in Figure 27. The conditions were simulated with and without CEM designs. Figure 28 shows the wheel-rail contact of the non-cem model at the leading and middle wheelsets in the leading bogie at four selected points in time 0ms, 91ms, 170ms and 412s. Three dimensional views equivalent to Figure 28(b) and (d) are given in Figure 29(a) and (b). It can be clearly seen from Figure 29 that due to the offset collision, the leading wheelset of the leading bogie derails. 0.5 m V = 80 km/h Loco Wagon Fixed Barrier (a) (b) Figure 27: (a) Obstacle offset before collision and (b) illustration of the model with 7 wagons CRC for Rail Innovation November 2013 Page 54

61 -110 kn -110 kn -110 kn -110 kn (a) 0ms (b) 91ms CRC for Rail Innovation November 2013 Page 55

62 -321 kn -316 kn (c) 170ms (d) 412ms Figure 28: Evolution of the wheel rail contact during collision for the non-cem design case CRC for Rail Innovation November 2013 Page 56

63 (a) Figure 29: Offset of wheelset during the collision of a non-cem design at (a) 91ms and (b) 412ms (b) A similar simulation is carried out for the train model with CEM design. Figure 30 shows the wheel-rail contact situations equivalent to the previous case. With the provision of the CEM components no derailment occurs and the initial wheel-rail contact forces are greatly reduced. Therefore, while it is well known that CEM designs significantly reduce the frontal impact force and longitudinal deceleration of consists, their ability to control deformation can be an asset in limiting derailments. The use of rigid and semi-rigid multibody dynamics has provided very interesting results of great interest for the matter. The computational speed is significantly higher than Finite Element Analysis and can therefore provide rapid outcomes on the overall elastic and plastic absorption properties required to arrest the train. The analysis therefore allows early estimation of acceptable crash forces and crunch volumes as well as showing if the CEM will keep the train assembly on its rails. Through such parametric studies an optimal range of values of absorbing characteristics can be estimated and used as starting point a target for the development of the structure through FEM. CRC for Rail Innovation November 2013 Page 57