2外文原文.pdf

Precision Engineering 34 2010 399 407 Contents lists available at ScienceDirect Precision Engineering journal homepage Evaluation of modelling approaches for machine tool design Daisuke Konoa Thomas Lorenzerb Sascha Weikertc Konrad Wegenerc aDept of Micro Engineering Graduate School of Engineering Kyoto University Yoshida honmachi Sakyo ku Kyoto 606 8501 Japan bInspire AG Tannenstrasse 3 8092 Zurich Switzerland cInstitute for Machine Tools and Manufacturing IWF ETH Zurich Tannenstrasse 3 8092 Zurich Switzerland a r t i c l ei n f o Article history Received 12 February 2009 Received in revised form 8 June 2009 Accepted 16 September 2009 Available online 23 September 2009 Keywords Rigid body simulation Finite element method Modal analysis Machine tool design a b s t r a c t In order to evaluate the confi guration of machine tools the IWF Axis Construction Kit ACK has been developed Thispaperdescribestheevaluationofthisapproach TheACKsupportsrigidbodysimulations and simple elastic body simulations The ACK is compared with commercial FEM software to investigate its usability and reliability Required time was compared in modelling of a machine tool The ACK needed 30 of the total required time for the FEM because of its modularity in machine modelling Then in order to investigate the reliability of the ACK static and dynamic simulations of both approaches were compared with each other and with analytical calculations on basic beam models The result showed that the ACK provided equivalent results to the FEM Static and dynamic simulations were also compared with measurements on an actual machine tool The ACK obtained almost equivalent results to the FEM Almost all lower structural mode shapes and their natural frequencies could be reproduced with the ACK when crucial parts were modelled using elastic bodies 2009 Elsevier Inc All rights reserved 1 Introduction Recently dynamic errors of machine tools such as vibration are one of the crucial problems in high precision machining Since dynamic properties of machines are greatly infl uenced by the machine confi guration the confi guration should be evaluated very early in the design phase However only few manufacturers use evaluation tools in order to check confi guration variants For this reason a lack of reliable and effortless simulation software can be stated Simulation methods for machine tools can roughly be clas sifi ed into two groups one is the fi nite element method FEM and the other is the rigid multi body simulation MBS 1 In industrial use the FEM is popular and widely used Several studieshavebeencarriedout modellingmachinecomponentswith the FEM Zaeh and Oertli have developed a model for ball screw drives by the cross coupling between axial and torsional degrees of freedom 2 Altintas and Cao have developed a FEM model of spindlescomposedofnonlinearmodelsofshaftsandbearings 3 4 Reliable results can be obtained with the FEM Furthermore the FEM is useful in the design process because many FEM software packages have useful interfaces to 3D CAD systems However the FEM needs much calculation effort to model the whole structure of a machine because of its large number of degrees of freedom nDOF In commercial software coupling set Corresponding author Tel 81 75 753 5226 fax 81 75 753 5226 E mail address Daisuke Kono t02 mbox media kyoto u ac jp D Kono tings between assembled parts are often not practical to represent guideways or ballscrews and detailed modelling in a certain level isrequiredtoavoidtheinterferencebetweencomponents Inorder toreducenDOF themachinehasbeendividedintosubcomponents or modules and these modules have been coupled with boundary conditions 5 8 Alsointhesimulationsofpositioningsystemwith fl exture hinges hinges are modelled as a simple beam element to reduce nDOF 9 Compared to the FEM nDOF can be signifi cantly reduced when using the MBS Especially in the early design stage the MBS is an appropriate tool to obtain quick and rough predictions of the machine behaviour 1 Teeuwsen et al developed a motion error model of CMM with the MBS 10 Tool motions in the time domain and modal characteristics have been simulated by the MBS also for machine tools 11 12 In recent years the MBS has been used for real time simulation of motion errors 13 The MBS and the FEM were combined to acquire more reliable results in dynamic analy ses 14 Although these studies have shown the advantage of the MBS there are only few practical software packages that focus on the construction of machine tools As can be seen in the use of the FEM a combination of 3D CAD and commercial MBS software is used to analyse the motion of a parallelkinematicmachinetool 15 Thismethodispractical How ever commercial software packages generally need more effort to model a specifi c machine since they are designed for all purpose In order to evaluate the confi guration of machine tools the Axis Construction Kit ACK has been developed at the IWF ETH Zurich Institute for Machine Tools and Manufacturing 16 Fundamental 0141 6359 see front matter 2009 Elsevier Inc All rights reserved doi 10 1016 j precisioneng 2009 09 003 400D Kono et al Precision Engineering 34 2010 399 407 Fig 1 Appearance of the machine under investigation machinepropertiescanbeobtainedinaveryshorttimebyasimple model composed of masses and springs The model is easily con structed using modules of mechanical components The ACK can handle not only the MBS but also simple elastic body simulations Static and dynamic analyses can be conducted In this paper the ACK is compared with commercial FEM soft ware to evaluate its reliability and usability First required time for modelling of a machine tool is compared Then simple beams are modelledwithbothapproaches Theobtainedresultsarecompared with each other and with analytical calculations to investigate the reliability of elastic body simulations of the ACK Finally an exist ing machine is modelled and static and dynamic behaviours are compared with measurements on the real machine The effect of elastic body simulations is also evaluated 2 Modelling of a three axis machine tool 2 1 Simulation object A high precision machine tool is modelled with the ACK and the FEM Fig 1 shows the machine under investigation The machine has been developed for shaping grooving in the XY plane by a single diamond tool It is mounted on a vibration isolation table to avoid vibration in the Z direction This machine has three linear axes Precision ball screws and servo motors are employed in each drive system The X axis uses aerostatic guideways The Y and Z axes employ linear ball guideways The major specifi cations of the machine and the vibration isolation table are shown in Table 1 Fig 2 Simulation model developed with the ACK Fig 3 Five different models used in simulations 2 2 Modelling with the Axis Construction Kit Fig 2 shows the simulation model developed with the ACK The machinepartsaresimplifi edintorigidrectanglebodies Someparts are grouped in one component and the seven components shown in Fig 2 compose the model consequently The components are coupled with discrete springs in X Y and Z directions Sliders and bases are connected with couplings which represent guideways and drives The machine base is coupled to an inertial system with weakspringsatthecornersrepresentingairspringsofthevibration isolation table The machine base the columns and the Y base are rigidly coupled with stiff springs which represent bolts or screws The ACK can handle elastic body simulations using segmented bodies coupled with each other In this paper fi ve different models are developed to investigate the effect of different elastic body rep resentations Fig 3 shows the schematic of these fi ve models The details of the models are as follows Table 1 Major specifi cations of the machine tool and the vibration isolation table Machine tool AxisXYZ Travel200mm Max feedrate10m min Drive typeBall screw servo motor Lead of ball screw10mm10mm5mm GuidewayAerostaticRolling ball Rolling ball MaterialGraniteSteelSteel Driven mass35kg47kg4kg Positioning resolution500nm500nm300nm Vibration isolation table TypeAir spring with diaphragm Natural frequency1 2 2 0Hz Mass230kg D Kono et al Precision Engineering 34 2010 399 407401 Rigid R This is the basic model All bodies are rigid The machine base the columns and the Y base are rigidly coupled The model is described in the fi rst paragraph Boltselastic BE Themachinebaseandthecolumnsareelastically coupled Columns elastic CE The columns are elastic and segmented in the Z direction Y base elastic YE The Y base is elastic and segmented in the Y direction Y base and columns elastic YCE The Y base as well as the columns is elastic The initial model parameters are identifi ed in the following procedure thedimensionsofbodiesaredeterminedfromtheirout wardforms andthedensitiesareadjustedsothatthemassesagree to the design value The stiffness of the guideways of Y and Z axes are derived from their catalogue specifi cations The stiffness of air springs drives and the X guideway are identifi ed experimentally In order to confi rm the reliability of the approach when param eters are properly selected these parameters are tuned using the result of the experimental modal analysis described in Section 4 2 The modifi ed parameters are the stiffness in the Z direction of the Y guideway the stiffness of the Y drive and the thickness of the columns and the Y base In BE the stiffness in the Z direction of the couplings between the machine base and the columns is deter minedalsofromtheresultofthemodalanalysis Becausethispaper focusesonstaticdisplacementsandnaturalfrequencieswithmode shapes damping has not to been taken into account 2 3 Modelling with commercial FEM software The machine is also modelled with a commercial FEM software package Ansys Workbench similarly as mentioned in Section 2 2 Bonded contacts are used to group the bodies The rigidity or the elasticityofbodiesisdeterminedbythebehaviourofsurfacenodes Bodies are meshed automatically under default settings With the FEM reference surfaces are required to couple components How ever thereferencesurfacesaffectthebehaviourofelasticbodies In order to suppress this infl uence reference surfaces are minimised using the extrusion with the thickness of 1 m so that only actual contacted areas become the reference 2 4 Comparison in required time for the ACK and a commercial FEM The time required for simulations with the ACK and the FEM is compared The evaluation is carried out in modelling modal anal ysis and the modifi cation of the thickness of the columns after the analysis on R The operator has all the model parameters at hand and has developed the same model several times with both software With the FEM modelling of the extrusion for the refer ence surface is omitted to investigate the time in the equivalent operations to ACK The comparison of the required time is shown in Fig 4 It is observed that the ACK needs only 30 of the total required time for the FEM Direct 3D modelling and module coupling system mainly contribute to this time reduction Comparing the time required for the modifi cation the ACK needs 25 of the time for the FEM 3 Fundamental evaluation on basic components In order to evaluate the accuracy of elastic body simulations by the ACK simulation results are compared with those by the FEM andwithanalyticalcalculationsinbasicsimulations Inelasticbody simulations the number of degrees of freedom nDOF of models has a large infl uence Therefore simulations are conducted with Fig 4 Comparison of required time with the ACK and a commercial FEM various nDOF to clarify this effect The simulation of displacements due to static forces and the modal analysis are carried out on a cantilever and a beam fi xed on both sides In this paper nDOF with the ACK and the FEM are calculated as follows nDOFACK 6 me 1 nDOFFEM 3 mn 2 where meis the number of elements segmented bodies and mnis the number of nodes For example with the FEM nDOF of a single element is 60 with a 20 node solid element composed of 8 cor ner and 12 midside nodes When elements are connected nDOF of nodes shared on boundaries decreases 3 1 Cantilever beam Fig 5 shows a cantilever used in the calculation l h and t rep resent the width the height and the thickness of the cantilever respectively In static analyses forces are applied at the load point which is on the center of the top surface The cantilever is seg mented in the Z direction with the ACK In order to compare the ACK and the FEM in the similar nDOF the cantilever is modelled without midside nodes with the FEM The physical parameters are Fig 5 Cantilever used in the calculation 402D Kono et al Precision Engineering 34 2010 399 407 Fig 6 Relationship between nDOF and deviation from the analytical displacement for a cantilever with the ACK determined according to those of steel Derived from the dimen sion of the columns of the machine the dimensions l 100mm h 500mm and t 38mm are used in the simulations 3 1 1 Displacement due to static forces The translational forces in three directions Fx Fyand Fz and moments Ma Mband Mc are applied at the load point one by one The displacements due to Ma Mband Mcare evaluated in X and Y directions at the load point The displacement due to Mcis calcu lated as the product of a distance of l 2 and an angle of twist at the top surface In this calculation the translational force of 1kN and the moment of 1Nm are loaded The deviation of the static displacement calculated by the ACK and the FEM from its analytical value is shown in Figs 6 and 7 respectively In Fig 6 it can be observed that deviations on the displacement due to Fxand Fy dx Fx and dy Fy decrease while nDOF increases from 12 to 60 With nDOF more than 60 the devia tions are almost constant This result shows that nDOF must be 30 which corresponds to 5 elements with the ACK or larger to obtain the deviation smaller than 10 The deviations are smaller than 4 when nDOF is more than 90 In Fig 7 the deviation on dx Mc is still larger than 10 even when nDOF is more than 500 This deviation is caused by the lack of midside nodes As for other displacements nDOF must be more than 40 to reduce the deviation to 10 for the FEM The devia tions are smaller than 2 with nDOF more than 200 Comparing Fig 7 Relationship between nDOF and deviation from the analytical displacement for a cantilever with the FEM Fig 8 Relationship between nDOF and deviation from the analytical natural fre quency for a cantilever with the ACK Figs 6and7 thedifferenceindeviationsissmallerthan5 between both approaches with nDOF more than 100 This result shows that the ACK has an equivalent accuracy to the FEM Furthermore the ACK typically calculates displacements larger than the FEM when nDOF is smaller which can be seen as a conservative manner of evaluating machine design 3 1 2 Modal analysis for the cantilever Thenaturalfrequenciesofthefi rstsixmodesarecalculated The fi rst six modes of this cantilever are the fi rst bending mode in the X direction the fi rst bending mode in the Y direction the second bending mode in the X direction the fi rst torsion mode the second bending mode in the Y direction and the third bending mode in the X direction The deviations of natural frequencies calculated by the ACK and the FEM from their analytical values are shown in Figs 8 and 9 respectively In both computations the third bending mode in the X direction was not observed with the smallest nDOF With the ACK nDOF must be larger than 60 which corresponds to 10 ele ments to reduce the deviation to 20 The deviations are smaller than 16 with nDOF more than 200 In Fig 9 nDOF must be larger than 100 for the FEM to obtain the deviation smaller than 20 The deviations are smaller than 15 with nDOF more than 200 Comparing Figs 8 and 9 the deviations with the ACK are smaller in the second and third modes in the X Fig 9 Relationship between nDOF and deviation from the analytical natural fre quency for a cantilever with the FEM D Kono et al Precision Engineering 34 2010 399 407403 Fig 10 Beam fi xed on both sides used in the calculation Fig 11 RelationshipbetweennDOFanddeviationfromtheanalyticaldisplacement for the fi xed beam of Fig 10 with the ACK direction with nDOF from 40 to 100 The difference of deviations is smaller than 10 between both approaches when nDOF is more than 100 3 2 Fixed beam Fig 10 shows a beam fi xed on both sides to be used for com parison In Fig 10 hFrepresents the distance in the Z direction from the center of the front surface to the load point The beam is segmented in the Y direction with the ACK In the modal anal ysis the beam is modelled without midside nodes with the FEM Fig 12 RelationshipbetweennDOFanddeviationfromtheanalyticaldisplacement for the fi xed beam of Fig 10 with the FEM Fig 13 Relationship between nDOF and deviation from the analytical natural fre quency for the fi xed beam of Fig 10 with the ACK In the static analysis with the FEM a small plate with the size of 0 5mm 10mm 10mm X Y Z isextrudedattheloadpointas thereferencesurfaceoftheload Thebeamisnotmesheduniformly because of this plate Since midside nodes could have a large infl u enceinsuchacase thebeamismodelledwithandwithoutmidside nodes in the static analysis The physical parameters are deter mined according to those of steel Derived from the dimension of the Y base of the machine the dimensions l 750mm h 250mm and t 32mm are used in the simulation However the simula tion results shall deviate from analytical Euler Bernoulli solutions because of a small ratio of l h and l t 3 2 1 Displacement due to static forces When the translational force in the X direction Fx is applied at the load point the displacements in X and Z directions at the load point dx Fx and dz Fx are calculated dx Fx is computed as the sum of translational and angular displacements Fx 1kN is loaded at hF 0mm and hF 100mm in the calculation The deviations of static displacements calculated with the ACK