EXPERIMENTAL AND NUMERICAL RESEARCH ON BULLDOZER WORKING PROCESS

CHINESE JOURNAL OF MECHANICAL ENGINEERINGVol. 20, No. 2, 2007●41●EXPERIMENTAL AND NUMERICALL YanjieRESEARCH ON BULLDOZERXU YongWORKING PROCESS*HUANG WenbinCollge of Science,China Agiutural Universily,Beijing 10083, Chinain theELFEN package. Before snmlation the soil spcimes are cxamined yhrough wrixiaitesilecompression, trixia! copression and direet shear ests to obltin mode: Cacacristits andFENG YTrelevant paramneters, then soil cuting experiments are caried out vIaamini-son binsystem wIthasonlbed of 60/120 mm in widh and 10 mm in deph cut bya.1/9 scale model blldoer blade moving withOWEN DRJthe velocity of 10 mp/s. The soil cositive model includes the ensile castite me for tensilebreakage and the compressive eatplastic rlationship with Mohr-Coulomb crterion. The cutingCivil & Computational Engineringlength in simulatiou is set as 14 of tha in the xerimet dvided into 1 869 tiangular cemems. TheCenter. ,comparison betwen the simulated results and experimental ones shows that the used model is capableUniversityof Wales Swansea,of analyzing soil dynamic behaviors qualitatively, and the predicted fracturing profles in generalns;Swansea SA2 8PP UKconform to the experiment. Hence the feasibility for analyzing soil fracturing bchaviors in tllage orother similar proceses is validated.Key words: Finite element method Discrete element methodFinite elementdistinct clement method(FE/DEM) Soil Bulldozerdisadvantages in analyzing soil behavior. The FEM is hard to0 INTRODUCTIONbreak while the DEM is hard to aggregate.To overcome the limitation ofFEM and to take the advantageSoil failure and its subsequent breakage under the action ofof DEM in dealing with the discrete system, OWEN, et alu“，soil cutting tools is common to may applications, i.e. tllagehave developed a so called combined finite/discrete elementpross with a plow in gricitural nginering and soil lveig mehod for the simolaton of muiatrini of the qusibritein civil engineering. A certain volume of soil in the process maymaterials. In addition to the classical FEM formulation, this newexperience large deformation, crack generation and furthermethod consists of two additional parts: One is the failure andfragmentation, while the resulting discrete blocks may movefracture mechanism based on the theory of fracture mechanics,wibin he suroundingn are, and Colide wih each otber. These and he other is the agoritimis for the conter ieretieon betveencomplex physical phenomena evolve transient changes fromthe discrete elements resulted from the fragmentation of thecontinuum to discontinuous states. Therefore the general theorycontinum. The finite element/distinct element method (FE/DEM)for (nuumnd related numerical methods can mnodel ooly the combined the advantages of FEM and DEM together, therefore itsoil behavior at the continuous stage, whereas cannot deal withcan be used in multi-fracturing analysis of quasibritle materialsthe consequent discontinuous state. Unlike the remarkablesuch as rock and concrete. ANTE4I described the FE/DEM in hisadvance in numerical methods for continuous problems,book in detail.comparatively less success has been achieved in the numericalAlthoughhe FE/DEM has been scessfully applied tomodeling of soil dynamics due to the diversity of the constitutive many eginering problems involving multifacturing phenom-relationships and the lack of iteraction mechanism for the ena, its use for analyzing the soil dynamics has not bee reporteddiscrete soil blocks, even though great efforts have been made inup to date. The purpose of this paper is to investigate the potentialthe field. On the other hand, the need for gaining a gooapplicability of the method to numerically analyze the fracturingunderstanding on the transient process of soil from continuous toof a dense moist clay bed disturbed by a bulldozer blade, anddiscontinuous states becomes crucial due to its important role further to validate the numeriricalresultsuts by' experimental teplaying in many engineering applications, such as dam collapse,In order to focus the application aspect of FE/DEM for theexplosion and slope stability, etc.current problem, the algorithmic detail of the FE/DE method isThere are mainly two numerical methods used in soil omitted in the paper. The detailed description of the method,dynamics: the finite element method (FEM) and the discrete however, can be found in Refs. [13-14].element method(DEM). The former is commonly used in solid1 NUMERICAL MODEL AND MATERIALmechanics and fluicdynamics, whilethe later is suitable forparticle/discontinuous systems. Many investigators studied soilPARAMETERScutting and tllage with FEM[+4. These works, however, do nottake into account crack initiation, propagation and subsequent1.1 Soil failure constitutive modelfracturing in their numerical models. Recently KARMAKAR, etThe core of the FE/DEM is to choose a proper constitutiveals, carried out an experimental study on soil failure associatedwith crack popagation for an aigcuturat tlgeg tl Compared model for the soil concemed. Sine the compresive stengh iswith the FEM, the DEM, which is based on the iteraction much greater than the tnsile strengh for qusibittle materialsbetween discrete objects and the motion law, has its advantage inthe main form of failure is mostlv due to the tensile stress. Hencedescription of discrete particle systerms or agglomerates. There the soi中国煤化工aper is a combination ofare fewer investigators studying the soil dynamic behavior with tensile:npressive elasto-plasticDEM[69. Both FEM and DEM have their own advantages and deformCNMHGipalstresatapointistensile, the stress value is checked whether it reaches the tensile* This project is supported by National Natural Science Foundation of Chinastrength according to the maximum (tensile) stress theory. If not,(No.10372113) and Royal Society-NSFC China-UK Joint Project (No.the deformnation is elastic, otherwise the point will start linearly16468). Reeved September 7. 20060 reeived in rvised form December14，softening, this sofening may continue until the consumed2006; acepted Deccember 22, 2006LI Yamnjie, et al: Experimental and numerical research on bllozer working processdeformation energy reaches a certain value limit, known as the Fig. 3. The experimental soil width is equal to the width of bladefracure energy G% then the point will break and a crack happens. (60 mm or 120 mm), and the shape of blldozer blade is 19 of theThis constitutive model of the tensile failure is shown in Fig.1.real blade size. The velocity of the blade is 10 mm/s, the length ofthe soil bin is 570 mm, and the depth of the soil disturbed byblade is 10 mm.Elastic deforming1胃plane- Linear softening易Hydrostatic.StrainFig. 1 Consitutive model of tensile failure(a) Yield surface cut off the part between π planeand apex of Mohr-CoulombFrom Fig. 1, the constitutive curve is divided into two parts:the elastic deformation and the linear softening. If the tensilestrengthf; and the element critical fracturing strain 8 are already- 一0known, the tensile constitutive curve is determined. The tensilestrengthf can be obtained by the soil uniaxial tensile test, whileRankineHydrostaictensile comerthe element critical fracturing strain E can be calculated by theaxisfracture energy theory of the Griffith brittle fracture in fracturemechanics. The curve in Fig. 1 can be described by the following(b) Yield surface added the Rankine tensile cormerequationFig. 2 Mohr-Coulomb yield surfaces(1)心一)一Strain gaugesSide bafleswhere-Tensile strength-Inelastic fracturing strain-Element critical fracturing strainNBulldozer bladeGr- --Fracture energyh0)--Elemental average dimension rfrring to the dia-Weighing sensor Four wheeksmeter of the circle whose area is equal to the elementFig. 3 Problem sketchWhen the soil is dominated by a compressive stress field, it isAccording to the geometry of the soil and blade, the problemunder elasto-plastic deformation and no fracture happens since thecan be simplified as a two-dimensional plane strain one. Thesoil model is regarded as elastic-perfect plasticity for compression.The Mobhr-Coulomb citerion is used and it can be exresedas height of the soil bed in compuationis the same. as that ofexperiment, but the length of the soil bed in computation is taken|川|=c。- o。tanφ .2) as 1/4 of that of experiment in order to reduce the computationalcost. The blade is assumed as a rigid body. In the currentwhere q, t and Co are the internal friction angle, the shear failure simulation, cracks can happen onlybetween two adjacentstress and the material cohesion respectively. It is noted that the elements at their common boundaries. It means that a singlenormal stress on acting on an inclined plane is defined here to be element cannot be split into sub elements. An unstructured meshnegative in compression. The Mohr-Coulomb failure criterion in of the domain is chosen so the discretized elements near the bladeprincipal stress space is given bytip working region are relatively smaller than the others so as tomake the computation smooth when the fracture occurs. The total(σm -n)+ (σmu +omin )sinφ =G。cos3)number of elements is 1 869. The restrictions and initial condi-tions used in the simulation are the same as the experiments. Thewhere Tmax and Omin are the maximum and minimum principalfinite element mesh and constraints are shown in Fig. 4.stresses respectively. The constitutive model for tension andcompression can be described with the Mohr-Coulomb yieldv=0.01 m/s-suface with the Rankine corner. That is to cut off the partbetween the π plane and the apex (Fig. 2a). Then the Rankinetensile cormer is added into, whose three surfaces are vertical to σI,02 and O3 principal stresses axial directions. The apex of Rankinetensile corner is the crossing point of the π plane and the中国煤化工hydrostatic axis (Fig. 2b)..2 Computation model and parametersMHCNMHGThe problem concermed in this study is the dynamic fractureof a manual compacted soil bed (570 mm in lengthx80 mm or 160mm in width x 65 mm in height) cut by a bulldozer blade movingforward horizontally at a constant velocity. The sketch is shown inFig. 42D finite element mesh and constraintsCHINESE JOURNAL OF MECHANICAL ENGINEERING●43.The parameters of soil constitutive and the discrete element:surface will move upwards as the fractured soil accumulates onthe curved surface, so the resistance can not be calculated throughcontact calculation are given in Table 1.measuring the bending moment. This difficulty is overcome by anTable 1 Parameters of soil and discrete element calculationalternative measurement. Since the acting point of the resultantforce does not change in the transverse direction, the resistanceModulus of laticit EMPa4 0*can be obtained through measuring the torsion by making a4.0*0.35*transverse offset distance, noted as a, between the screw axis andWet density Pw /(kg.m~)the central line of the blade width. Since the acting point of theSee Fig.5resultant force is normally on the central line, the torque canTensile strength fMkPa3.62simply be calculated as the product of the force multiplied by theFracture energy G(J.m)2.0ofset distance. The torque can be determined by measuring theSliding friction ceoficient f0.2shear strain on the hollow cylindrical blade shank, The sketch ofCritical time step factor in DEM Io0.3the torque measurement is shown in Fig. 8. The bulldozer bladeContact damping in DEMcan also be replaced with other soil cutting tools with the sameGravity acceleration 8/mg 3)9.81experimental system for any other purpose.Note: * is the initial value of the experiments. The dilatancy angle decreases asthe increase of the plastic strain ratio. Because of its lttle effeet on theExperiment system )simulation results, the relation between dilatancy angle and plastice strain ratiois simplified into the two phase lines (Fig.5).ControllingDriving|Cutting[ Measurement J0 0.20.40.6 0.8 T.0Plastic strain ratio即Fig. 5 Relation between dilatancy angleand plastic strain ratioFig.7 Functional modules of the experimental system2 EXPERIMENTSIn order to verify the results of the numerical simulation, anexperiment is carried out with a mini-soil bin developed. Theexperimental system is shown in Fig. 6. This system can bedivided into four parts, i.e. the controlling, the driving, the cuttingand the measurement modules. The functional module sketch is. Direction of F is normnal to the paper plane,shown in Fig. 7. The bulldozer blade is fixed on a cylindrical barand its vertical position can shiftlinked with the female screw of a screw driver set controlled by anelctromotor, so that it can move forward, backward and stopFig.8 Front view of the torque measurementautomatically when the moving part hits the limit switches. InThe outer and inner diameters of the blade pipe shank are D =rder to measure the horizontal soil resistance on the blade19 mm and d=16 mm respectively, and the shear modulus of theaccurately, two sets of data collecting schemes are usedblade is specified as G=80 GPa, the length of the torsion arm issimultaneously.(1) The soil bin is supported by four rlling bearings as .a=36 mm. According to the Wheastone bridge principle, the strainwheels on a flat tray so that the soil bin can move forward andfoil gages (R), R2, R3, R) bonded on the cylinder surface arebackward freely on the flat tray. When the blade moves forward, itorganized as full-bridge resistance strain measurement circuit, socan drive the soil bin to move forward together and the soil binthe soil resistance can be calculated from the following equationwould act on the weighing sensor.F=πG&s(D*-d^)(4)(2) On the blade shank there are four foil strain gauges which32Daare conncted to a dynamic strain-gauge indicator. Through thesetwo data Ccllelin schemes the dynamie strain and the rsistne where es is the dynamic strain in the dretien of 45.signals can be inspected and saved in the computers.The soil used in the experiments is identified as clay aftersize distribution analysis. The soil granule size distribution isdepicted in Fig. 9.号80-中国煤化工YHCNMHGFig.6 Soil bin experimental system0.01 0.001During the process of the bulldozer blade moving forward,Diameter of soil particle d/mmthe resultant force acting point of soil resistance at the bladeFig. 9 Granule size distibutionLI Yanjie, et al: Experimental and numerical research on bulldozer working processIn order to examine the characteristics and mechanicalFrom Fig. 10, it can be seen that the soil fracture profile plotsparameters of the soil, a series of soil mechanics experiments for from the simulation are very similar to those photos from thethe compacted moist soil sample are carried out, including theexperiment. Firstly, from the photos it is evident that theuniaxial tensile test, uniaxial compression test, triaxialcompacted soil is a qusi-bitle material with less plasticity.compression test and direct shear test. The physical anSecondly, the plots from the simulation shows the selected modelmechanical parameters obtained from these experiments are listedof tensile fracturing plus compressive elasto-plasticity is capablein Table 2. These parameters are not only essential for thetor the of dealing with the brittle material. It is noted that since the realnumerical modeling， but also to make the results of thesoil bed is not exactly isotropic, the soil fracturing profile in thesimulations and the experiments comparable.experiment at each time instant behaves randomly, but the regularTable2 Physical parameters of soilpatterns from several experiments can be ensured identical.ParameterValuThe soil resistances on the narrow and the wide blades in theDry density Ppsl(kg.m 51.50experiments are shown in Fig. 11, from which it can be seen thatWet density P:w/(kg ●m)1.695the soil resistance on the wide blade (120 mm in width) is aboutAverage water content w/%twice as large as that of the narow blade (60 mm in widh),Cohesion CAkPa19.3Unification of the two curves can be done to make the forceFriction angle p/°)27.5values divided by the width to get a unit width resistances, thoseDuring the soil bed preparation, a certain weight of watertwo sets and their average values are shown in Fig. 12, fromwhich it is evident the three curves are almost the same.was added into the dry soil until its water content reached 13%，and then the wet soil was kept in an airtight container for over 24140-h to make the water diffused. Before the beginning of theexperiment the soil was agitated and compressed layer by layer in台120- -Widethe soil bin to make the soil form an isotropic continuum cuboidal10bed. Before adding soil for every new layer, the surface of the80compacted layer should be roughened sO as to keep the isotropy.善603 RESULTS AND DISCUSSIONSIn this study, the initial condition and parameters used in thesimulation are basically identical to those of the experiment,01020一304050therefore, the results between simulation and experiments shouldExperimental time t/sbe comparable. The soil profiles of the experiment and the 2Dsimulation are shown in Fig. 10.ig. 11 Soil resistance on the narrowand wide blades (measured)1.0- - . Wide- Narrow0.8-0.6-(a) Experiment(t=0.0 s)(b) Simulation(t =0.0 s)b5101520253035404550Experimental time 1/sFig. 12 Soil resistance on the unit widthof wide and narrow blades (measured)(c) Experimen(t =1.0 s)(d) Simulatiom(=1.0s)The soil resistance on the blade in the 2D simulation isshown in Fig. 13. Comparing it with Fig. 12, Fig. 13 shows thatthe soil resistance fluctuates all the time as the soil fracturecontinues. When a new crack occurs, the resistance increasesinstantaneously and decreases sharply, then the soil clumps movealong the curved blade surface so that the resistance increasesslowly and smoothly to form a plateau segment. Because the soilfracture and clumps movement happen altemnatively in the whole(e) Experimen( -2.0s)(f) Simulation(t -2.05)process, the resistance curve also fluctuates in turn. Thisphenomenon observed in the experiment is also predicted byFE/DEM simulation although it is dificult to reach a goodcoincidence. The total experiment takes about 50 s while thesimula中国煤化工-1uge computational costsinvolv; ificult to make a fullHCNMHG_Ithe beginning 5 s ofresistancece values variesbetween 200 N/m and 400 N/m. Although these two curves are(g) Experiment(t -4.0s)(b) Simulation(t -4.0 s)not identical, the resuts can be regarded as reasonable in terms ofFig. 10 Soil profiles comparisons between experiment and simulationthe force magnitude.CHINESE JOURNAL OF MECHANICAL ENGINEERING●45●经400In our numerical study, the facture energy G, is found tobe one of the two very important parameterseffecting on the soilfracturing profiles (the other is the tensile strength). Fig. 15 showsthe soil fracture profles when G, = 1.0 J/m’and other parameters200remain the same as those used in the above simulation. Thematerial is much more fragile than the clay in the experiment and0hence its characteristic is totally different with the used clay.一2-言4-5古Experimental time 1/sFig. 13. Soil resistance onunitof wide and narrow blades (simulation)Fig. 14 shows the maximum and minimum principal stressvector changes for some nodes near the blade tip at different timeinstants. The maximum principal stress (that is tensile stress) isshown in the dark line and the minimum principal stress (that iscompressive stress) is in the light blue. In Fig. 14a, the very largestress field near the blade tip appears just before a crack isgenerated by the blade tip. The compressive stress is radiating(a)1=1.0sfrom the blade tip and the values were large enough to compel thesoil to yield and enter plastic state. Besides, there were tensilestresses at some nodes in the direction orthogonal to thecompressive stress so that the cracks will be generated at thesenodes. Fig. 14b ilustrates that after the soil fracturing thecompressive and tensile stress decrease sharply since the energy isreleased. In the following time the tensile stress would increase togenerate another new crack when the tensile stress reaches thetensile strength while at the part of the compressive stressdominating zone, such as the soil top surface, there is only plasticdeformation according to the constiutive relatiosbip. Fig. 14cshows the stress field at the moment when. the crack is(b)r=2.0spropagating and the soil is not acting on the blade. The stress nearFig. 15 Simuiated proflesthe blade tip is very small and the tensile stress is much less thanfor soil when Gr= 1.0 J/m2compressive stress.4 CONCLUSIONSAn FE/DEM simulative analysis and the coupled laboratoryexperiment on soil dynamics were carried out and it can beconcluded as follows.(1) Moist dense clay can be regarded as a kind ofquasi britte materials for numerical analysis. The feasibility isproved to take the advantage of the FE/DEM for predicting thesoil fracturing behaviors for tllage or other similar processes withthe constitutive model of brittle materials. The comparisonbetween the simulation and experiment indicates that the method()-0.33sis capable of analyzing soil dynamic behaviors. Nevertheless,among the many previous studies on soil dynamics with the finiteelement method, there is not any predicted picture describing thebreakage or fracturing of a continuous soil prior to our approach.(2) In order to obtain reliable numerical results it is essentialto couple the numerical modeling with experiments of soilmechanics so as to examine the specific characteristics andproperties for proper model selection and parameter specification.The reported experimental set can be used for various purposes ofsoil dynamic study with altermative cutting tools for datacollection and analysis.(b)1=2.82 s(3) Since soil is one of the most complex medium, thisapproach on soil fracturing is only the first step for justifying theapplicability of the proposed method. More investigations areneeded such as the improvement in experimental aspects and thespecification of model parameters, especially the fracture energywhich中国煤化工h; simulated results.ACKYHCNMHGThe authors was grateful for the theoretical consultant fromProf. HUANG Wenbin, China Agriculural University, China and(0)1=5.01sthe kind help from Dr RANCE J M, et al, at the RockfieldFig. 14 Maximum and minimum principal stress vectorSoftware Ltd, UK.●46●LI Yanjie, et al: Experimental and numerical research op blldozer working procesKLERCK P A, SELLERS E」, OWEN D R J. Discrete fracture in quasi-References14 KIbritle materials under compressive and tensile stress states[J]. ComputerMethods in Applied Mechanics and Engineering, 2004, 193(27-29):[1] YONG R N, HANNA A W. Finite element analysis of plane soil3 035-3 056.cutting[]J Joumal of Terramechanics, 1977, 14(3): 103-125.[2] CHI L, KUSHWAHA R L, A non-linear 3D finite element analysis ofBiographical notessoil failure with tle tos[]J Joumal of Trramechanics, 1990, 27(4;[3] MOUAZEN A M, NEMENYI M. Finite element analysis of subsoiler LI Yanjie is curently a PhD candidate in China Agricutual University China.cutting in non-homogeneous sandy loam soil[]. Soil & Tillage Research, Her research interests include discrete element method/fnite element method[4] ABU-HAMDEH H N, REEDER (1999, 51(1): 1-15..REEDER cA nonlinear 3D finite elementsimulation and mecbanical experimeats,s etc.Tel: +86-10 62736514; E mai: lianjie bjaeu@l26.comanalysis of the soil forces acting on a disk plow[小. Soil & TllgeXU Yong is currently a professor in College of Science, China AgriculturalKARMAKAR S, KUSHWAHA R L, STILLING D s D. Soil failureUniversity, China, He received his master degree from Beijing Agriculturalassociated with crack propagation for an agricultural tlge to[] SoilUniversity, China, in 1981. His research interests include computational solid& Tillage Research, 2005, 84(2): 119-126.mechanics, especially for granular materials with discrete element method and6] MOMOZU M, OIDA A, YAMAZAKI M, et al. Simulation of a soilfinite element method.loosening process by means of the modifed distinct element method[J.Tel: +86-10 62736514; E-mail: xuyong@cau.du.cnJournal of Terramechanics, 2002, 39(4): 207-220.[7] LIUSH, SUND A, WANG Y s. Numerical study of soil cllapspe HUANG Wenbin is a professor in College of Science, China Agriculturalbehavior by discrete element modeling[]. Computers and Geotechnics,University, China. His research interests include plastic mechanics and finiteclement method.}] NAKASHIMA H, OIDA A. Algorithm and implementation of soil-tie Tel: +86-10- 62736209contact analysis code based on dynamic FE-DE method[]. Journal ofFENG Y T is currently a senior lecturer inCil & Computational Engineering9] ZHANG R, LIJ Q Simulation on mechanical behavior of chesive soil Center, University of Wales Swansea, UK. His rscarc iterests include theby distinct element method[J] Journal of Trramechanics, 2006, 43(3):lgoibimss of the nonliner problem in egineeig, reenaly p303-316.[0] OWEN DRI, FENG Y T. Pallised fitdiscrete clement sioulatioTel: +44-1792295161; E-mail: y.feng@swansea. ac.ukof multifacturing solids and discrete systers([]J. Enginering Comput-OWEN D R J is currently a professor in Civil & Computational EngineeringHAN K, OWEN D R J, PERIC D. Combined fnite/discrete element andCenter, University of Wales Swansea, UK. He received his PhD degree from[1explicitiplicit simulations of peen forming process[]. Engineering Northwestern University, USA, in 1966, and received bis Doctor of ScienceComputations, 2002, 19(1): 92-118.honor from University of Wales, UK, in 1982. He is now a fellow of Royal[12] ANTE M. The combined finite discrete element method[M]. London:Academy of Enginering. His research interests include the algorithms of theJohn Wiley&Sons, Ltd, 2004.nonlinear problem in civil enginering and other pplications, the analysis of[13] PERIC D, OWEN D R J. Computational model for 3D contat prob- multfictiring wit FEDEM is bghlighed.lems with fictio based on the peraltly method[I. Itetatioial Joumal Tel: +4-17922952525 E-mil: d.j.wsases.c.kof Numerical Method in Engineering, 1992 35(6); 1289-1 309.中国煤化工MYHCNMHG