Overhead guard physical tests vs LS-DYNA FE simulations.
The lifting and excavating industry are not as advanced as automotive in the use of modern CAE tools in the early stages of design and development of heavy machinery. There is still a lack of confidence in the integrity of the results from FE simulations and optimisation and this becomes a barrier to the adoption of virtual prototyping for vehicle verification. R&D of Tata Steel has performed tests on two forklift truck overhead guards supplied by a major manufacturer. Based on the international standard for Falling Object Protective Structures (FOPS) as an initial input to the method of testing, the main aim of this study was to generate as much test data as possible to correlate the Finite Element (FE) simulations of two tests - a static and a dynamic test. The static test was developed to deform the overhead guard plastically in a slow controlled manner, so it would be easier to correlate the measured data to FE simulation. A dynamic test is always very difficult to correlate due to the fact that the whole test is over in less than a second and the dynamic response of the structure makes the data extremely noisy. However, this paper demonstrates that paying close attention to all of the input data for an FE model results in a close correlation between the test and the FE simulation even in highly dynamic conditions.
CITATION: Cafolla, J., Smart, D., and Warner, B., "Overhead Guard Physical Tests vs LS-DYNA FE Simulations," SAE Int. J. Commer. Veh. 9(2):2016, doi:10.4271/2016-01-9017.
ROPS, FOPS and TOPS are well known acronyms in agricultural, material handling, mining and construction industry. Roll Over; Falling Object and Tip Over Protective Structures have to be used as appropriate in all tractors, combine harvesters, forklift trucks, bulldozers, excavators, compactors, loaders, etc. ROPS, FOPS and TOPS are defined in various standards which define the requirements that the structure has to pass. Prototype testing is the only way to pass these standards and often the statement 'If it looks strong and bulky, it must do the job' governs the design.
Finite element (FE) methods are rare in aiding the physical tests. Historically a lot of research has been done on ROPS in the 1970s as a reaction to the number of fatalities in over-turned tractors. Mathematical models developed by Chisholm  formed the basis for standards development such as SAE J2194  and OECD Standard Code 4 . Late 1990s and early 2000s saw the use of FE in ROPS validations in many research papers such as [4,5,6] just to list a few. Despite the fact that the FOPS standards have been developed similarly to ROPS standards, there is much less evidence of FOPS FE simulations to validate the standards or develop any mathematical or numerical approach to FOPS. Kaneda and Tamagawa  have analyzed FOPS using the commercial software PAM-CRASH. Their findings have shown good correlations between the measured and calculated values in terms of time-displacement measurements.
There is an increasing demand for light-weighting, reductions in C[O.sub.2] emissions and cost saving in the off-highway vehicle sector. This encourages the sector to look for alternatives and the success of aerospace and automotive industries in using the FE technique is becoming an attractive way of improving a design. The trust in Computer Aided Engineering (CAE) in heavy machinery has still a long way to go though. This study is aiming to reinforce this trust.
Two overhead guards (OHG) used on a forklift truck (Figure 1) have been kindly donated by Hyster-Yale Group to R&D of Tata Steel. The FOPS standard  for a forklift truck has been used as a basis to develop the two tests - static and dynamic. The static test was developed to ensure that enough data is collected to be able to correlate with FE simulation. The dynamic test was designed to be as close to the real test as possible (given the fact that the standard test is done on the full vehicle) to see how close the FE simulation can get to the test data. A dynamic test is usually difficult to correlate due to the noisiness of the data and short duration of the test. Material testing is essential for this type of study and this was planned in the testing schedule.
The paper describes the non-standard physical tests, the FE simulations and a comparison of the results. The study is split into four sections. The first section briefly explains the FOPS standard and the motivation behind the development of Tata Steel's own test procedures to do this study. Material testing is explained briefly in the second section, together with its importance for FE analysis and the material data source. The third section touches on the FE method, software versions and few less common LS-DYNA features used during the FE simulations. The fourth and main section then describes in much more detail the tests and FE set-up, results obtained and their comparison with a conclusion in the Summary section.
OHG TEST STANDARD
The chosen 3.5tonne forklift truck undergoes testing given by the standard . The truck is tested in its full assembly (Figure 1) and according to the standard it is subjected to a sequence of the following tests: a non-destructive static test, 10 times 45kg cube drop test and an impact lumber drop test. If applicable, there could also be an angle impact drop test, penetration test and/or the 45kg cube dropped onto the cowl or other structures intended to protect the operator's legs and fleet.
The static load test consists of applying a uniformly distributed load on the top of the OHG for at least one minute. The test load depends on the rated capacity of the truck. The test is passed if no part of the OHG or its attachment exceeds 10mm of permanent vertical deformation.
The drop test with a 45kg cube consists of a sequence of 10 drops of the cube strategically located 1525mm above the OHG (locations and their tolerance are given by the standard). After the 10th drop, minor cracks are permissible, but any completely broken parts will fail the test. Permanent deformation after the last drop may not exceed 19mm.
The impact drop test consists of dropping a large lumber load onto the OHG. The weight and height of the load depends on the truck rated capacity. Same as with the previous test, any broken parts would fail the test; however, minor cracks are permissible.
The R&D department of Tata Steel has not followed these standard procedures, but has developed their own. There was couple of reasons for doing so. Firstly, only the frame of the OHG has been available for testing and not the whole forklift truck as is required in the standard test. Secondly, there was a desire to collect as much data from the test as possible, especially from the static test. The standard static test does not deform the OHG significantly; therefore it would have been difficult to pinpoint the exact locations for the strain gauges to measure some data of reasonable values. The frame needed to deform such that plastic deformations were achieved in pre-defined locations.
The material used for the OHG was specified as S355J2 (EN10025-2:2004). For a good correlation of the test with FE simulation it was essential to do at least a basic material tensile test. No extra material samples were available so after the two tests the OHGs were cut up to obtain the tensile test samples. The samples were taken from the areas of the OHG where there was no plastic deformation during the two tests, such as the rear of the roof bars and legs (Figure 2).
The legs of the OHG have been subjected to forming strains during manufacture, so it was expected to get higher strength material from the samples taken from the legs. This is exactly what has happened. Figure 2 shows the quasi-static true stress-true plastic strain curves as calculated from measured engineering stress-strain curves obtained from tensile test compared to the curves from the Tata Steel's AURORA material database. It is clear that the tensile test on the roof bars samples matches the S355 database curve to a high degree. Samples taken from the legs have resulted in higher strength curve at the beginning matching the S420 stress-strain data from the AURORA database, but then gradually joining the stress-strain curve of S355 at about 0.2 true plastic strain. This is exactly the behaviour of material hardened by forming.
Usually the quasi-static stress-strain curves are sufficient input for the static tests. However, the dynamic testing requires the strain rate stress-strain curves. Due to the lack of sufficient amount of material from the remaining OHGs and lack of resource, it was not possible to obtain the strain rate stress-strain data by testing. The Tata Steel's AURORA database has the full range of stress-strain data for both S355 and S420. Using these and the quasi-static stress-strain curves from the tensile testing the two full sets of stress-strain data have been adopted and appropriately adjusted to be used in the present FE simulations in the form as shown in Figure 3.
FE SIMULATIONS SET-UP
The static simulation has used the implicit solution of LS-DYNA [10,11]. The force displacement curve from the test has been adopted for the FE simulation. The test displacement-time curve has been smoothed and applied in FE as displacement boundary conditions imposed on the rigid loading block. In the test, the OHG has been pre-loaded up to 10kN to check that all the instrumentation was working. The load has been released and all instrumentation reset to 0 before the start of the static test. This pre-loading of the OHG has also been modeled in the FE simulation.
In the test, the max load has been held for about 10 minutes, while all FaroArm measurements have been taken. The FE simulation has omitted this time; the loading curve was reversed straight after the max load had been reached and slowly downloaded to complete the release of the load on the OHG.
Element formulation of type 2 - Belytschko-Tsay shell elements with 5 through thickness integration points modeled the frame. Approximate size of the shells was 8-12mm to create a reasonably sized model which would not take too long to run and therefore could be used as a guide in development stage of a design. It was felt that the mesh sensitivity study was not needed since the work preceding this correlation study on the same frame has proven to give satisfactory results.
The '*MAT_PIECEWISE_LINEAR_PLASTICITY' card with its option for tabulated data has been used for both static and dynamic simulations.
Modeling of welds has not been considered - the joints were created by careful alignment of the mesh. This simplified the requirements for contact modeling by omitting the self contact of the structure and any possible weld/joint contacts often used in automotive crash event modeling. The only contact needed was therefore the automatic nodes to surface contact between the loading block bottom surface and the frame nodes with friction coefficient 0.1.
The dynamic analysis has used a combination of explicit and implicit solutions. The explicit solution had modeled the impact event while the implicit solution has been used to 'calm down' the vibrations of the structure after the impact. The last state of the deformed shape, including the stresses and strains has been saved in a 'dynain' file after each drop using the '*INTERFACE_SPRINGBACK_LSDYNA' card. This file was then used for the subsequent next drop simulation as an input file. This way the deformations and strains from all ten drops were accounted for when looking at the final deformation of the OHG.
Before the release of LS-DYNA Version 971 R7.1.1, the 10 drops had to be submitted separately meaning that each individual simulation of a drop had to finish creating the dynain file as the input before the next drop could be manually submitted. After the release of LS-DYNA Version 971 R7.1.1, the '*CASE_option' card had simplified the process of submitting all 10 drops in only one input file.
THE PHYSICAL TESTS SET-UP AND ITS REPRESENTATION IN FE MODEL
Both OHGs (one for the static test and one for the dynamic test) were bolted and welded to the strong floor. Most of the instrumentation attached to the OHGs has been used in both tests. The following gives the detail description of all the instrumentation and how it was related to the FE modeling.
Wire Transducers Measurements
For the static test only, a frame has been fixed around the OHG. Seven wire transducers were then attached to this frame and the OHG to measure the continuous displacements of the OHG's roof. Figure 4 shows the locations of the wire transducers in the test and the nodal IDs of the corresponding positions in the FE model. The measurements in both test and FE simulations provided the force displacement curves for comparison.
Force-displacement curves are usually the first indications of a good correlation result. Figure 5 shows the comparison of the force-displacement curves as measured in test by the wire transducers and FE curves obtained from the displacement of the corresponding nodes. For example WT03 and FE node ID 11040 (Figure 4) had the same location on the OHG in test and FE model, respectively.
The solid lines in Figure 5 are the force-displacement curves from FE and the dashed lines are from the test. The vertical displacements measured in the Y-direction reflect the directionality of the displacement. Some of the points have moved upwards. For example the curves for point WT08 in test and node ID 4842 in FE simulation are at the negative side of vertical displacements in Figure 5. This point represents one of the free ends of the centre part at the split in the side rail for the battery crane hoist. The second free end - WT15 in test and node ID 5194 - has the largest discrepancy between test and the FE. At these two points the ends are by their nature free to move. The OHG is a flexible structure allowing these two free ends to have larger discrepancy between test and FE than any other points.
The rest of the measurements shown in Figure 5 obtained from wire transducers in the test (dashed lines) are very close to those obtained from FE simulation (solid lines).
The FaroArm is a portable coordinate measuring arm fixed to the floor next to the OHG (Figure 6). The end of the moving arm can measure the coordinates of any point in space relative to the reference coordinate system defined by four fixed points on the floor.
A further 26 points were positioned on the OHG as shown in Figure 7. They were marked by numbers 5 to 30 to identify the locations of the measured points in the tests.
Unlike the continuous curves in force-displacement results, FaroArm measurements gave discrete displacement measurements. Three measurements were taken; at the beginning of the test, at the maximum load and finally at the end of the test. The marked locations of the points for the FaroArm measurements were matched with the nodes in the same locations on the FE model. The displacements of these nodes at the maximum load and at the end of the simulation have been recorded and compared to the test measurements.
Similarly to the wire transducers, the comparison of the test to the FE simulation results is good, as shown in Figure 8. The points at the end of the split in the side rail for the battery crane hoist (points 8, 9,10,16,17 and 18 in Figure 7; results in Figure 8) show larger discrepancies between the test and FE simulation results due to the increased flexibility of the whole roof structure in this area. The rest of the points are within 20% for the measurements at max load and up to about 30% for the measurements at the end of the test. Given the flexibility of the structure and that there is only one available test result, the correlation is very good.
Nine 3D Rosettes and two 1D strain gauges were strategically placed around the OHG to measure the strains. These were then compared with the surface strains of the shell elements from the FE model picked from the same OHG locations as in the test.
The 3D Rosettes measure strain in three directions - 0[degrees], 90[degrees] and 45[degrees]. To be able to compare the strains to FE, the principal strains need to be calculated from these three dimensional strains. The well known formula  to do so is:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where [[epsilon].sub.1] and [[epsilon].sub.2] are Major and Minor strains respectively; [[epsilon].sub.0], [[epsilon].sub.45] and [[epsilon].sub.90] are directional surface strains measured by 3D Rosettes as shown in Figure 9.
The majority of software for post-processing FE simulations already has the calculation of the principal strains coded in. The Major and Minor strains from FE simulations were therefore directly read as output results of the simulations (Figure 9).
Strain output from FE simulation was averaged over the element area and given at the mid-point of the element plane as indicated in Figure 9. Based on the location of the 3D Rosette the four elements surrounding this location have been chosen and their strains compared to test strains. There were 9 different locations on the structure, all chosen strategically where the highest strains could be expected during the tests. A large amount of data has been generated and analyzed.
The test results in all locations correlated very well with the FE simulations. As an example, Figure 10 shows the results of the strain gauge location 2 (shown in Figure 9). There are two distinct graphs showing the results. The first two plots show major and minor strain plotted on the y-axis, respectively, against the force on the x-axis.
The third plot is the major vs. minor strain graph. In all three plots the orange line represents the strain measured in the test. The red, purple, blue and green lines represent the strain obtained from FE simulation for the elements of the same color in the top right picture of Figure 10. The strain levels are very small in absolute values. They change rapidly around the structure, so just a small change in the location could mean a significant change in the strain level. This is shown also in the two principal strains contour plots at the bottom of the Figure 10. The major and minor contour strains are plotted at the max load. The strains are very localized in the frame and picking the correct position of the strain gauge was critical.
The orange line in the graphs representing the test is about in the middle of the other lines representing the strain results of individual elements in FE simulations. The test results are therefore approximately an average of the four element results for both major and minor strains in this area. The third graph shows the major strain plotted against minor strain with the dotted line being the calculated average of the elemental results and this result is close to the test result. In general, the strain measurements in test and FE simulation are of the right levels and showing the correct trend.
Very similar behavior can be observed on the other eight 3D Rosettes measurements when comparing them to the FE simulation results.
The two one-directional strain gauges could not be used to calculate the principal strains. At these two locations there was a regular element mesh where one side of each element was parallel to the one-directional gauge; therefore this strain direction has been used for the comparison of the test and FE strains. Similarly as with the 3D Rosettes the agreement between the values and trends was good.
High-Speed Camera Measurements
The high-speed camera was used only in the dynamic tests to capture the impact of the loading block on the OHG. It was positioned to the side of the OHG (Figure 6). The recording captured the movement of the markers placed on the OHG just next to points 8, 15 and 16 (Figure 7) and on the loading block itself.
The movement of the marker on the loading block enabled the impact velocity of the block to be calculated. The impact velocity in all drops but one was slightly lower than the predicted 5.47m/s calculated from:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where [v.sub.0] = 0m/s is the initial velocity of the drop from height h = 1525mm and g is the gravitational constant. The measured impact velocity values (Figure 11) were used in the drop test simulations. For drop 8 the calculated value was used, due to an error in recording during the test.
Figure 11. Variable velocity measured by high-speed camera Drop 1 5.64 Drop 2 5.15 Drop 3 5.18 Drop 4 4.72 Drop 5 5.1 Drop 6 5.41 Drop 7 5.23 Drop 9 4.91 Drop 10 5.17 Note: Table made from bar graph.
The markers positioned next to the points 8, 15 and 16 enabled the first maximum initial displacements of these three points below and above the starting position to be measured.
Figure 12 shows these displacements for two of the points - 8 and 15. Point 16 had a similar result to point 8. The blue bars represent the initial movement (usually downward) followed by the upward movement (red bars) as the frame bounces back passing the initial starting position. The movement is indicated by either "B" - max distance below the starting position and "A" - max distance above the starting position. The green and purple bars represent the movements of nodes N5770 and N9408 from FE simulations in similar manner as the points 8 and 15 in test, respectively.
Points 8 and 16 are at the free end of the U-shaped cut-out of the side rail and their dynamic behavior is more chaotic than point 15 being placed approximately in the middle of the roof. As a result of this it is not surprising that point 8 has a couple of drops in which the test shows the first crest above the starting position while the FE simulation shows it below the starting point or vice versa (drops 2, 6 and 7 for point 8 in Figure 12).
Overall the trend is similar in both; test and FE simulation. The differences between test and FE simulation for the "negative" measurements of point 15 are, for all but one, better than 20%. It is more for the points 8 and 16, but giving the nature of the test and FE simulation - being dynamic - this is a very good correlation result.
The FaroArm has been used in the dynamic test in the same way as it was in the static test. The measurements were taken before and after each drop. Figure 13 shows the results of final displacements of individual points as shown in Figure 7 after the last drop. The agreement between the test and FE simulation is very good. Comparing the FE results and test, approximately half of the measured points differ only by about 5% while the difference of the rest of the measured points is within 20%. Given that it is difficult to correlate a dynamic test with FE, these results show that with careful consideration of all the conditions it is possible to get a good agreement between the test and FE simulation.
The strain measurements in the dynamic test were measured exactly the same way as in the static test. Therefore the calculation of the principal strains is the same too. There was, however, one major difference. While static test measured the strains from the beginning of the test to its end, the dynamic drops happened very quickly. The strain measurements in each drop were measured over the period of about 4 seconds starting well before the impact. The impact represented only about 0.12 seconds of the recorded data. The rest of the data was discarded. Also each strain gauge has been fully reset after each drop; this was taken into consideration when post-processing the data. There was no force measurement in the dynamic tests therefore the final
comparison graphs were represented by the line plots where time was on the x-axis and principal strain on the yaxis. As in the static test the comparison between the results of test and FE simulation show that FE results are similar to the test results considering the elements surrounding the gauge location. The test curve does not consistently match an element output or the average of the four elements output for different drops, but all five curves on each graph are closely clustered within a given range. As an example Figure 14 shows the results of gauge 2 and drops 1 and 10.
The majority of the measurements taken during the test are in good agreement with the FE simulation output for both static and dynamic tests. This study was aiming to increase our customer's confidence in FE simulations in an area of FOPS and ROPS structural design and integrity. The intention was to demonstrate a good level of correlation between FE simulation and practical testing in order to validate virtual prototyping models developed on other projects with Hyster-Yale. The results comparison between the test and FE showed the same trend. Majority of individual measurements have shown the difference less than 10%. Where the results were represented by curves, the shape of test and FE result curves was very close. The customer has been very impressed with both the testing at the laboratories of Tata Steel in Rotherham and the test / FE simulation correlation. The integrity of the correlation is based on one set of tests only and further testing is required to completely validate the analysis tests for all situations. Finite element simulations will never give exact answers due to the many assumptions that have to be considered as well as other factors, such as, variability in material, processes, test circumstances, etc. However, this study demonstrates that finite element simulation is an excellent tool for overall prediction of structure behavior and the use of virtual prototyping would be a benefit in cost and weight reduction of off-highway vehicle structures.
[1.] Chisholm C. J.: A Mathematical Model of Tractor Overturning and Impact Behaviour, Journal of Agricultural Engineering Research, Volume 24, Issue 4, December 1979
[2.] SAE International Surface Vehicle Recommended Practive, "Roll-Over Protective Structures (ROPS) for Wheeled Agricultural Tractors," SAE Standard J1637, Reaf. April 2009.
[3.] OECD: Standard Codes for the Official Testing of Protective Structures Mounted on Agricultural and Forestry Tractors, OECD, Paris, 2005
[4.] Alfaro J. R., Arana I., Arazuri S., Jaren C.: Assessing the Safety Provided by SAE J 2194 Standard and Code 4 Standard code for testing ROPS, using Finite Element Analysis, Journal of Biosystems Engineering, Volume 105, Issue 2, February 2010
[5.] Fabbri A., Ward S.: Validation of a Finite Element Program for the Design of Roll-over Protective Framed Structures (ROPS) for Agricultural Tractors, Journal of Biosystems Engineering, Volume 81, Issue 3, March 2002
[6.] Harris J. R., Mucino V. H., Etherton J. R., Snyder K. A., Means K. H.: Finite Element Modeling of ROPS in Static Testing and Rear Overturns, Journal of Agricultural Safety and Health, Volume 6 (3), August 2000
[7.] Kaneda S., Tamagawa T.: Introduction of Simulation of Falling Object Protective Structures, KOMATSU Technical Report, Volume 49, No. 151, 2003
[8.] ISO/TC 110/SC 2 Safety of powered industrial trucks, "Industrial trucks - Overhead guard - Specification and testing", ISO 6055:2004, June 2004
[9.] Benham P. P., Crawford R. J.: Mechanics of Engineering Materials, Longman Scientific and Technical, 1987
[10.] LS-DYNA Keyword user's manual, Volume I, LS-DYNA R7.1 (revision: 5471), May 2014
[11.] LS-DYNA Keyword user's manual, Volume II - Material Models, LS-DYNA R7.1 (revision: 5442), May 2014
Dr Janka Cafolla
CEng MIMechE, R&D
Tata Steel, UK
Project Manager - L&E Sector, R&D
Tata Steel, UK
The authors would like to express their thanks to Graham Gould from Hyster-Yale, who provided invaluable advice on tests set-up; and to all in Tata Steel UK Limited, R&D, who helped with the preparation and execution of the tests, namely: Mick Large, Frank Coombes, Ian Lister, Stephen Danks, Neil Bennett, George Ibbeson, David Norman and Tony Mudrak
FOPS - Falling Object Protective Structure
ROPS - Roll-Over Protective Structure
TOPS - Tip-Over Protective Structure
OHG - Over-Head Guard
FE(A) - Finite Element (Analysis)
CAE - Computer Aided Engineering
R&D - Research and Development
AURORA - Tata Steel Material Database
1D - 1 Dimensional
3D - 3 Dimensional
Janka Cafolla and Derick Smart Tata Steel UK Limited
Barry Warner Hyster-Yale UK Limited
|Printer friendly Cite/link Email Feedback|
|Author:||Cafolla, Janka; Smart, Derick; Warner, Barry|
|Publication:||SAE International Journal of Commercial Vehicles|
|Date:||Oct 1, 2016|
|Previous Article:||Track-based aerodynamic testing of a heavy-duty vehicle: Coast-down measurements.|
|Next Article:||Effect of cab suspension configuration and location on tractor semi-trailer driver comfort.|