Outline – Error Budget g • Error Budget – Spreadsheets • Example – 2-axis stage • • Coordinate C di S System S Selection l i Centers – Center of stiffness – Center of static friction – Center of dynamic friction • • • Effect of Centers Stiffness Analysis Examples p – Surface grinder machine – Sensor output analysis Chapter 3 ME 551 2 Error Budget g 1 • Any project has finite resources: – Certain amount of moneyy to be allocated to each component and process. – Certain amount of time to complete the project at each phase of the project. – Certain amount of error which the machine can have (i.e. we can tolerate!). • Individual components and processes can each have only so much error. error Chapter 3 ME 551 3 Error Budget g ((Cont’d)) • M Money- and d titime aspects t off a project j t are dealt with using formal financial and managerial budgeting tools. • An error budget is used to allocate errors to components and processes. • Unfortunately, error budgeting is often done on an ad-hoc basis,, or based on the experience of the chief designer. Chapter 3 ME 551 4 Error Budget g ((Cont’d)) • A formal approach to error budgeting can be beneficial: – It can help train less experienced designers and minimize supervisory burdens on more experienced design engineers. – It can help speed up the development process and avoid needless design iterations by highlighting g g gp problem areas before p prototypes yp are built. Chapter 3 ME 551 5 Error Budget g Tools • Mathematical construction of an error budget is typically based on – Rigid body models – Homogeneous transformation matrices (HTMs) – Error gain- or sensitivity matrix • Accuracy is based on – Underlying Models – Number of error sources included – Number N b off iintermediate t di t coordinate di t fframes Chapter 3 ME 551 6 EB Tools (Cont’d) ( ) • Like financial budgets, there are many different formats. • They should be dynamic tools used during the design and manufacturing processes. • Practice is required to gain proficiency. • Practice is best obtained by playing what if scenarios. • Spreadsheet or math software (MathCad or Matlab) based error budgets are easy to use but they are not simple to construct. Chapter 3 ME 551 7 1 EB Spreadsheets p • Spreadsheets provide an easier-to-use easier to use analysis tool for initial evaluation of a design's accuracy: – Cl Closed d fform solutions l ti are iimpractical. ti l – Helps the engineer distribute error allowances amongst subsystems. subsystems – Reminds the engineer to continually evaluate each and every element in the design design. – Engineer can easily develop custom spreadsheets and merge them far easier than the specific programs written in C/C++. Chapter 3 ME 551 8 EB Spreadsheets p ((Cont’d)) • They help to minimize the cost of the machine. – Engineering time: • Quick assessment of technical feasibility • Evaluate E l t the th merit it off each h componentt and d design. d i – Rework costs: • Minimizes the chance that the machine will not meet spec. spec which would require redesign or hand finishing. • Makes error mapping easier should it be required because machines kinematics would already be modeled. modeled • A basic spreadsheet can be constructed and then customized as needed. • Dynamic and thermal error modules can be constructed. Chapter 3 ME 551 9 Example p – Stacked Axes1,2 −ε z1 ⎡ 1 ⎢ε 1 R ⎢ z1 1T= ⎢ −ε y1 ε x1 ⎢ 0 ⎣ 0 −ε z2 ⎡ 1 ⎢ε 1 1 ⎢ z2 2T= ⎢ −ε y2 ε x2 ⎢ 0 ⎣ 0 R Ptooll (actual ) = R1T ⋅ 12T ⋅ 2 Ptooll = R2T ⋅ [t x 0 ε y2 −ε x2 1 0 ⎥ ⎦ a2 + δ x2 ⎤ b2 + δ y2 ⎥⎥ c2 + δ z2 ⎥ ⎥ 1 ⎦ 1 ⎡ a1 + a2 + t x ⎤ ⎢b + b + t ⎥ R Ptool (ideal ) = ⎢ 1 2 y ⎥ ⎢ c1 + c2 + t z ⎥ ⎢ ⎥ 1 ⎣ ⎦ t z 1]T ty ε y1 a1 + δ x1 ⎤ −ε x1 b1 + δ y1 ⎥⎥ c1 + δ z1 ⎥ 1 ⎡ −t y (ε z1 + ε z 2 ) + t z (ε y1 + ε y 2 ) − b2ε z1 + c2ε y1 + δ x1 + δ x 2 ⎤ ⎢ t (ε + ε ) − t (ε + ε ) + a ε − c ε + δ + δ ⎥ z x1 x2 2 z1 2 x1 y1 y2 ⎥ R R Ptool (ideal ) − Ptool (actual ) = ⎢ x z1 z 2 ⎢ −t x (ε y1 + ε y 2 ) + t y (ε x1 + ε x 2 ) − a2ε y1 + b2ε x1 + δ z1 + δ z 2 ⎥ ⎢ ⎥ 0 ⎣ ⎦ Chapter 3 ME 551 10 Example p ((Cont’d)) ⎡δ xt ⎤ ⎡ −t y (ε z1 + ε z 2 ) + t z (ε y1 + ε y 2 ) − b2ε z1 + c2ε y1 + δ x1 + δ x 2 ⎤ ⎥ ⎢ ⎥ ⎢ ⎢δ yyt ⎥ = ⎢ t x (ε z1 + ε z 2 ) − t z (ε x1 + ε x 2 ) + a2ε z1 − c2ε x1 + δ y1 + δ y 2 ⎥ ⎢⎣δ zt ⎥⎦ ⎢⎣ −t x (ε y1 + ε y 2 ) + t y (ε x1 + ε x 2 ) − a2ε y1 + b2ε x1 + δ z1 + δ z 2 ⎥⎦ Axis 1 2 Feature Roll Yaw Pit h Pitch P-X S-Y S-Z Pitch Yaw Roll S-X SY S-Y P-Z Allow. Error Allow “ μm εx1 εy1 εz1 δx1 δy1 δz1 εx2 εy2 εz2 δx2 δy2 δz2 Offset (mm) X Y Z |tz+c2| |ty+b2| |tx+a2| |tz+c2| |ty+b2| |tx+a + 2| | z| |t |tz| |ty| |tx| Measurement Maximum possible error per axis Chapter 3 ||ty| |tx| Error (μm) X Y Z εx1|tz+c2| εx1|ty+b2| εy1|tx+a2| εy1|tz+c2| εz1|ty+b2| εz1|tx+a + 2| δx1 δy1 δz1 |tz|εx2 |ty|εx2 |tx|εy2 |tz|εy2 |ty|εz2 |tx|εz2 δx2 δy2 δz2 * * ME 551 * * * * 11 Numerical Example p Axis Feature Allow. Error “ μm Offset (mm) X 5 Roll 4 300 Yaw 5 500 1 Pitch PX P-X 10 00 10.00 S-Y 12.50 S-Z 20.00 4 Pitch Roll 4 400 2 Yaw 5 600 S-X 10.00 S-Y 12.50 PZ P-Z 20 00 20.00 Measurement Maximum possible error per axis Y Z 300 500 500 500 X 6.00 10.00 10 00 10.00 Y Z 7.50 10.00 10.00 10.00 12.50 400 600 600 600 8.00 8 00 8.00 15.00 10.00 20.00 12.00 12 00 12.00 15.00 12.50 12.00 71.00 δ max = 712 + 65.52 + 962 = 136 [ μm] Chapter 3 Error (μm) 12.00 65.50 20 00 20.00 12.00 96.00 In this case, δmax < 50 μm is desired, so keep on trying! ME 551 12 Notes on Error Budgets g 1,3 • When dealing g with error budgets, g , the worst case scenario is considered. – The coordinates ((or parameters) p ) are changed g to reflect various extreme positions of various elements. • The most challenging g g p part is to estimate (“guesstimate”) the deviations (or motion errors): – As mentioned before, they depend on many different factors. • For traditional designs, angular errors are taken as ±5 arcsec/100 mm. • For precision designs, they are less than ±3 arcsec/100 mm. Chapter 3 ME 551 13 Estimation of Angular g Errors1 • • • A linear axis consists of the bed, bearing rails, rails bearing blocks, blocks and carriage. We presume that each block has a running i parallelism ll li error (δ) and d that the bearing and its mounting each has a similar level of precision. Errors in the system are then conservatively modeled assuming all act at once in multiple directions about the center of stiffness: – – – – Vertical straightness = δ Horizontal straightness = δ Roll angle = 2δ/W Pitch and yaw angles = 2δ/L Pitch- Chapter 3 ME 551 14 Coordinate System y Selection1 • During the design phase, a coordinate system is typically located at each bearing interface. • The origin of this coordinate system is located on the plane of interface. – It needs to be located such that an axis is perpendicular to the plane of bearings bearings. – It should also pass through the bearing's center of stiffness: • It is the pivot point at which no rotation occurs when a force is directly applied th through h that th t point. i t Chapter 3 ME 551 15 Coord. System y ((Cont’d)) • F For design, d i it is i customary t t place to l th coordinate the di t system at the center of stiffness: – This choice decreases error accounting oversights and increases the potential for achieving an accurate model. – Enables you to design the machine in the stick-figure mode. – When the required stiffness of an axis is specified, the designer can select the bearings and their spacing. • Each coordinate system has 6 stiffness values: – 3 translational stiffness values: kx, ky, and kz [N/m]. – 3 torsional stiffness values: kθx, kθyy, and kθz [N/rad]. Chapter 3 ME 551 16 Coord. System y ((Cont’d)) • Once the error budget based on HTMs at the centers of stiffness has been made, one can do a sensor error budget: – An extra coordinate frame can be put at the exterior of the axis where measurements will be made after the machine is built. – The sensor error budget accounts for the fact that the sensor may be registering some of the Abbe offset error. – The sensor error budget only considers what motions the sensors measure, and what they are sensitive to. – The sensors are generally located between the center of stiffness and the toolpoint. toolpoint Chapter 3 ME 551 17 Coord. System y ((Cont’d)) • F For metrology t l on an existing i ti machine, hi th coordinate the di t systems are often lumped at the tool tip: – Errors are measured as if the axes are moved individually. individually – Straightness errors are measured along with Abbe errors. – One can g go from the design g HTMs at the centers of stiffness to the measurement HTMs by a simple HTM. • This allows the designers to see if there estimates of the errors were correct. • Angular errors take place about the axes whose origin is located at the center of stiffness. Chapter 3 ME 551 18 Coord. System y ((Cont’d)) • As the bearings move along the axis, the bearing surface it lf has itself h straightness t i ht and d angular l errors associated i t d with ith its surface. – Once the machine is built, built one would like to measure the errors and see if they correspond to the predicted ones. • A straightness g error being g measured is a combination of a straightness error and an angular error: – It is often not practical to place a straightness measuring device i id a machine inside hi componentt (i.e. (i att the th origin i i off the th coordinate di t system located at the pivot point!). • HTMs are utilized to determine expected straightness and angular error at the point on the structure at which the measurements will be made. Chapter 3 ME 551 19 Center of Stiffness1 • It is the point at which no angular motion of the structure occurs when h a force f i applied is li d to t a locked-in-place l k di l axis. i • It is also the point about which angular motion occurs when h forces f are applied li d elsewhere l h on the th body. b d • It can be found using a center-of-mass type of calculation ( is substituted for (K f M). ) – To find the location of the center of stiffness with respect to an arbitrary coordinate system: N x= ∑K x n n n =1 N ∑K n =1 Chapter 3 n ME 551 20 Center of Static Friction1 • After the center of stiffness is found, the bearing blocks‘ coordinates should be defined from this point. point • One can extend the concept of the center of stiffness to define the centers of friction. – The center of static friction is the point at which when a force is applied, no net moment is generated by the resulting static friction forces. forces – To find the location of the center of static friction with respect to an arbitrary coordinate system: N x fs = ∑F n =1 N ∑F n =1 Chapter 3 x s ,n n ME 551 s ,n 21 Center of Dynamic y Friction1 • Can be used to determine the best place for an actuator on a system with a V and flat arrangement of sliding contact bearings. • The center of dynamic friction is the point at which no net moment is generated by the dynamic friction forces pp when a force is applied. • To find the location of the center of dynamic friction with respect to an arbitrary coordinate system: N x fd = ∑F n =1 N ∑F n =1 Chapter 3 x d ,n n d ,n ME 551 22 Centers and Their Effects1 • For a system to be insensitive to actuation forces: – If the actuation force acts through the center of mass: • There is no inertial moment on the body. • There are no reaction forces on the bearings. • The center of mass and the center of stiffness do not necessarily have to be located at the same point. – However, for stacked multi-axis structures: • The centers of mass of the axes move. • It is i wise i to t locate l t the th point i t off actuation t ti att the th nominal i l center t of mass. Chapter 3 ME 551 23 Centers (Cont’d) ( ) • If the force is located between the center of friction and the center of stiffness: – There will be no moment acting on the system. – The center of friction and the center of stiffness do not necessarily have to be located at the same point. Chapter 3 ME 551 24 Stiffness Analysis y 1 • To analyze the stiffness of the machine, the kinematic chains of the machine must be taken into consideration. consideration • Closed kinematic chains are analyzed to determine motion errors of an equivalent open-chain ((HTM)) model: – Carriages supported by bearings – Closed structural loop • Bridge-type • Parallel mechanisms (Hexapod) Chapter 3 ME 551 25 Stiffness of a Linear Axis Consider the spatial transform of an (linear/translational) axis. The corresponding HTM can be written as −ε zi ε yi ai + δ xi ⎤ ⎡ 1 ⎢ε 1 −ε xi bi + δ yi ⎥⎥ zi i-1 ⎢ iT= ⎢ −ε yi ε xi 1 ci + δ zi ⎥ ⎢ ⎥ 0 0 1 ⎦ ⎣ 0 where ai, bi, ci are the offsets between the coordinate frames. The stiffness of this axis can be defined as F = Kδ ⇒ δ = K −1 F = CF where K ∈ ℜ6×6 is the stiffness matrix; C = K-1 is the compliance matrix matrix. Note that F ⎡⎣ Fxi Fyi Fzi M xi M yi M zi ⎤⎦ δ ⎡⎣δ xi δ yi δ zi ε xi ε yi ε zi ⎤⎦ T T Here, F is the generalized force vector that constitutes the external forces and moments translated to the origin g of the coordinate frame {{i}. } Chapter 3 ME 551 26 Stiffness1 (Cont’d) ( ) • For a symmetrical linear motion axis with all bearing blocks having equal stiffness, one can often assume that the Center of Stiffness is located at the geometric center of the bearings. Chapter 3 ME 551 27 Simplified p Carriage g Model kT k As a starting point, the carriage can p as a rigid g body y be represented supported on four (linear) bearing blocks. F is presumed to act on the center of stiffness. Hence, the compliance matrix of the carriage simply boils down to Transmission System k Y k L k Z Travel Direction Center of Stiffness X Rigid Carriage k ⎛ 1 1 1 1 1 1 ⎞ C = diag ⎜ , , , 2 , 2 , 2 ⎟ ⎝ kT 4k 4k W k L k L k ⎠ k W k k Bearing Block Chapter 3 As the first step, each structural member transmitting force / moment is regarded g as rigid. g Then, the deviations for each axis is computed via the given model. ME 551 28 Roller Guide Stiffness4 4 Ball rows 4 Contact C t t points i t 4 Roller rows Li contact Line t t Tensile D Deformation n [ m] Compressive 4 Ball rows 2 Contact points Load [kN] 2 Ball rows 4 Contact points Chapter 3 ME 551 6 Ball rows 2 Contact points 29 Elastic Deformations of Elements As the next step, all bearings are assumed rigid and the elastic deformation of each element ((or structural member)) is taken into consideration. For the sake of simplicity, p y, one can idealize a stacked-axis as a fixed-end beam. Recall that the deflection (δ) and the rotation (θ) of its free-end under the action of singular force (F) and the moment (M) can be given as FL3 ML2 δ= + 3EI 2EI Chapter 3 FL2 ML θ= + 2EI EI ME 551 30 Typical yp Deflections in Machining g Processes3 Machining Operation Average force in feed direction [N] Acceptable deformation [μm] Roughing w/ conventional tools 1500 100 Roughing w/ 100–125 mm. diam. tools 3000 100 Semi-finishing 1000 50 Finishing 200 <10 Chapter 3 ME 551 31 Deformation Analysis of a Typical Milling Machine4 Chapter 1 ME 551 32 Example p – Surface Grinder1 Column ZC Grinding i di Wheel le in d Sp e iag arr e iag arr YC Table Bed • • Consider the design of a CNC surface grinder grinder. Let us analyze the motion errors associated with this machine. – Load induced errors are significant in metal working machinery. Chapter 3 ME 551 33 Example p – Coordinate Frames Chapter 3 ME 551 34 Example p – Stiffness Analysis y t yt yt r t zt xt Ft Fr Fr ⎡δ xt ⎤ ⎡ k xt−1 ⎢δ ⎥ ⎢ ⎢ yt ⎥ ⎢ 0 ⎢δ zt ⎥ ⎢ 0 ⎢ ⎥=⎢ ⎢ε xt ⎥ ⎢ 0 ⎢ε yt ⎥ ⎢ 0 ⎢ ⎥ ⎢ ⎢⎣ ε zt ⎥⎦ ⎢⎣ 0 ⎡ 1 k −1 ⎡δ x 5 ⎤ ⎢ 4 5 ⎢δ ⎥ ⎢ 0 ⎢ y5 ⎥ ⎢ ⎢δ z 5 ⎥ ⎢ 0 ⎢ ⎥=⎢ 0 ⎢ε x5 ⎥ ⎢ ⎢ε y 5 ⎥ ⎢ 0 ⎢ ⎥ ⎢ ⎣⎢ ε z 5 ⎦⎥ ⎢ 0 ⎣ Chapter 3 ME 551 0 k yt−1 0 0 0 0 0 0 0 0 k zt−1 0 0 kθ−xt1 0 0 0 0 0 kθ−yt1 0 0 0 0 0 1 4 0 0 0 k 0 0 kT−51 0 0 0 0 0 0 −1 5 1 L25 k5−1 0 0 0 0 0 0 0 1 L25 k 0 −1 5 0 ⎤ ⎡ Ft ⎤ ⎥⎢ ⎥ 0 ⎥ ⎢ Fr ⎥ 0 ⎥⎢ 0 ⎥ ⎥⎢ ⎥ 0 ⎥ ⎢ − Fr 2t ⎥ 0 ⎥ ⎢ Ft 2t ⎥ ⎥⎢ ⎥ kθ−zt1 ⎥⎦ ⎢⎣ Ft r ⎥⎦ 0 ⎤⎡ ⎥⎢ 0 ⎥⎢ 0 ⎥ ⎢⎢ ⎥ 0 ⎥ ⎢ − Fr ⎥⎢ 0 ⎥ ⎢ Ft ⎥⎢ 1 k −1 ⎥ ⎢ F W52 5 ⎦ ⎣ t ⎤ ⎥ ⎥ ⎥ 0 ⎥ t − L5 2 +L ⎥ ⎥ t − L5 ⎥ 2 +L ⎥ r − H25 ⎥ ⎦ Ft Fr ( ( ( ) ) 35 ) Stiffness Analysis y (Cont’d) ( ) 1 −1 ⎡δ x 4 ⎤ ⎡⎢ 4 k4 ⎢ ⎥ 0 ⎢δ y 4 ⎥ ⎢⎢ ⎢δ z 4 ⎥ ⎢ 0 ⎢ ⎥=⎢ 0 ⎢ε x 4 ⎥ ⎢ε y 4 ⎥ ⎢⎢ 0 ⎢ ⎥ ⎢ ⎣⎢ ε z 4 ⎦⎥ ⎣⎢ 0 ⎡ k −1 ⎡δ x 2 ⎤ ⎢ T 2 ⎢δ ⎥ ⎢ 0 ⎢ y2 ⎥ ⎢ ⎢δ z 2 ⎥ ⎢ 0 ⎢ ⎥=⎢ 0 ⎢ε x 2 ⎥ ⎢ ⎢ε y 2 ⎥ ⎢ 0 ⎢ ⎥ ⎢⎣ ε z 2 ⎥⎦ ⎢ 0 ⎣⎢ Chapter 3 0 0 0 0 kT−14 0 0 0 1 −1 k 4 4 0 0 0 1 k −1 L2 4 0 0 0 0 0 0 0 0 0 0 k2−1 0 0 0 0 1 −1 k 4 2 0 0 0 0 0 0 0 4 1 4 1 L2 k4−1 0 0 0 0 1 k −1 W2 2 0 0 0 0 2 1 L22 0 ⎤⎡ Ft ⎥⎢ F r ⎥⎢ ⎥⎢ 0 ⎥⎢ − Fr ( t +2L + z ) ⎥⎢ ⎥⎢ Ft ( t +2L + z ) 0 ⎥ ⎢ H H 4 + L4 4 1 k −1 ⎥⎥ ⎢⎣ Fr 2 − Ft 2 − r L24 4 ⎦ ) − Ft ⎤⎡ ⎥⎢ − Fr ⎥⎢ 0 ⎥⎢ ⎥⎢ t −W ⎥ ⎢ Fr 2 2 + z − a ⎥⎢ 0 ⎥ ⎢ − F t −W2 + z − a t 2 −1 ⎥ ⎢ 1 2 k2 + F x Ft ( y − r ) ⎢ L2 ⎦⎥ ⎣ r ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥⎦ 0 0 0 k2−1 0 ( 0 0 0 0 ME 551 ( ( ) ) 36 Example p – Sensor Analysis y • Consider a single-stage positioning system with a linear scale5. • Let us determine the errors registered by the sensor reading (output) in this configuration. Chapter 3 ME 551 37 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ Example p – Coordinate Frames Chapter 3 ME 551 38 Example p ((Cont’d)) The HTMs of the system can be written as L ⎤ ⎡1 0 0 −ε z11 ε y11 x + δ x11 ⎤ ⎡ 1 ⎢ 0 1 0 −( a + b) ⎥ ⎢ε ⎥ 1 −ε x1 − a + δ y1 ⎥ R R ⎢ ⎥ ⎢ z1 ST= 1T= ⎢0 0 1 ⎢ −ε y1 ε x1 c ⎥ 1 δ z1 ⎥ ⎢ ⎥ ⎢ ⎥ 1 ⎦ 0 0 1 ⎦ ⎣0 0 0 ⎣ 0 The global coordinates of the measurement (reference) point become R Pm = R1T ⋅ 1Pm = RST ⋅ S Pm where 1Pm = [0 –b c 1]T . Hence, S Pm = ( T) R S −1 T ⋅ 1Pm R 1 Note that the first element of SPm is essentially the sensor reading (xm): xm = x − L + bε z1 + cε y1 + δ x1 As can be seen, the sensor output constitutes not only the servo-error but also some of the Abbe offset errors. errors Chapter 3 ME 551 39 References 1. A. H. Slocum, Precision Machine Design, SME Press, 1992. 1992 • A. H. Slocum, ME 2.075 Course Notes, MIT, 2001. 2 Cranfield Unit for Precision Engineering (CUPE), 2. (CUPE) Precision Engineering Course Notes, Cranfield Institute of Technology (UK), (UK) 1998. 1998 3. L.N. L. deLacalle, A. Lamikiz, Machine Tools for High g, Springer-Verlag, p g g, 2009. Performance Machining, 4. M. Weck, Werkzeugmachinen (Band 2), SpringerVerlag, 2005. 5. Heidenhain Corp., Linear Encoders Improve the Machining Accuracy, www.heidenhain.de. Chapter 3 ME 551 40