Design of An FDM Positioning System and Application of An Error-Cost Multiobjective Optimization Approach
Design of An FDM Positioning System and Application of An Error-Cost Multiobjective Optimization Approach
Design of An FDM Positioning System and Application of An Error-Cost Multiobjective Optimization Approach
The intense growth of additive manufacturing technologies and their applications in the last few years have given rise to an increased demand for high accuracy parts. However, the development of these technologies faces several challenges, such as the needed increase in quality and strength of the nal products (Gibson et al., 2010; Cunico, 2011). The main goal of this work is to evaluate a systematic approach for the design of additive manufacturing equipment in which both cost and nal accuracy of the machine are considered. As the basis of this study, a layout of a positioning system such as the stereolithography or selective laser sintering equipment (Gibson et al., 2010) was selected. In these systems, the
The current issue and full text archive of this journal is available at www.emeraldinsight.com/1355-2546.htm
displacement of the tool head is a direct result of the action of mechanical elements and does not to involve mirrors or galvanometers. Figure 1 shows a schematic of the selected concept, which is equivalent to a rst generation fused deposition modelling (FDM) machine, such as a 3D Modelerw or FDM series (Wang, 2010). It is also interesting to note that variations of this concept were adapted for simultaneous deposition and polymerisation (SDP), laminated object modeling (LOM) and 3D print technologies. Consequently, this work can also be used for the improvement of other similar technologies (Gibson et al., 2010; Cunico, 2011). For this research, a numerical model that allows the identication of the cost and nal error of the equipment
The authors would like to acknowledge the nancial support of CAPES, in addition to technical support of the Department of Post Graduation in o Paulo (campus Sa o Mechanical Engineering of the University of Sa Carlos), for providing access to infrastructure and laboratory. Received: 11 November 2011 Revised: 26 December 2011, 13 March 2012 Accepted: 16 March 2012
Rapid Prototyping Journal 19/5 (2013) 344 352 q Emerald Group Publishing Limited [ISSN 1355-2546] [DOI 10.1108/RPJ-11-2011-0117]
344
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
After estimating the cost-tolerance functions of the main components of the positioning system, we formulated a general cost-function equation that takes into account the manufacturing tolerances, machine element accuracy and the number of machine elements, as presented in equation (2). Through the use of this equation, it is also possible to evaluate other equipment layouts in addition to identifying the contribution of each component for the nal cost of the equipment: C total C g x N ax C g y N ay C g z N az w1 C t x N bx C t y N by C t z N bz w2 C m w3
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
Figure 2 Relation of main positioning machine elements, accuracy grade, general error and relative cost
Machine Element Accuracy Grade General Error Relative Cost
normal (c )
38 m 29 m 17 m 12 m 6 m 3 m 2 m
Linear Guide
Ultraprecision (UP)
IT10
500 m
100 m 210 m 52 m 23 m 12 m
Ball Screw
C03
BallScrew-C7
1.2
1.6
1.8
BallScrew-C5 2
where: Ctotal Cgx, Cgy and Cgz Ctx, Cty and Ctz Cm w1 is the sum of the machine element and manufacturing operation costs. are the costs of linear guides of the x-, y-, and z-axes, respectively. are the costs of the transmission elements of the x-, y-, and z-axes, respectively. is the relative cost of the manufacturing operation. is the weight of the relative cost of the linear guides. 346
is the weight of the relative cost of the transmission elements. w3 is the weight of the relative cost of the manufacturing operation. Nax, Nay and Naz are the quantities of linear guides used in the model. Nbx, Nby and Nbz are the quantities of transmission elements used in the model. w2 In accordance with the proposed equipment layout, the quantity of linear guides for the x- and y-axes is equal to 2, using no linear guide and four transmission elements
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
Figure 4 Regression curve of linear guide elements, accuracy and relative cost
45 m Linear Bearing - C/H 35 m 30 m Radial error 25 m 20 m 15 m 10 m 5 m 0 m 0.5 1 1.5 2 2.5 3 3.5 4 Relative cost
y = 120.31e1.132x R2 = 0.9667
Another point of interest is the effect of process parameters on the nal accuracy of the part. These parameters include deposition and extrusion velocities, extrusion and chamber temperatures, height of the deposition layer and the trajectory strategy. Due to the quantity of variables involved in the process, the reduction of positioning system errors is essential to avoid either distortions of the process or the interference of computational compensation of errors (Venkatesh and Izman, 2008; Tan et al., 2008; Tong et al., 2004; Dornfeld and Lee, 2007; Slocum, 1992). Numerical model The errors obtained from the theoretical model of a rst generation FDM machine layout (Figure 6) were compared with the announced accuracy of FDM 1600 and FDM 8000 for the evaluation of the numerical model of the equipment nal error (Crump et al., 2009; Wang, 2010; Stratasys, 1994, 1998). In this gure, the main axes used in the composition of the error budgeting model are shown in addition to the main dimensions of the positioning system. Therefore, it is possible to determine the axial translation coordinates for the numerical model. For the identication of the nal error of the system, the homogeneous transformation matrix (HTM) of each coordinate system (axis) was multiplied taking into consideration the translational and rotational errors, as observed in equation (3) (Venkatesh and Izman, 2008; Slocum, 1992):
R
for the z-axis. Similarly, the weight of the cost of the transmission elements is found to be 3.4, making the general cost of transmission higher than the linear guide in that proportion. We also dened the value of the manufacturing cost weight as 1 to consider the minimal cost necessary to fabricate the equipment. It is also important to note that this study focused on geometric errors, making it primarily useful for the conceptual design process. In other applications, additional sources of error should also be studied, such as kinematics, dynamics and thermal errors. Because these errors are caused by external forces, such as thermal expansion, material instability, friction, vibration and random sources, they should be considered in a second step (detailed design), where detailed specications can be developed for materials, dimensions and forces, among other parameters (Venkatesh and Izman, 2008; Tan et al., 2008; Dornfeld and Lee, 2007; Slocum, 1992). Therefore, we emphasise that the intention of this proposal is to help the designer select a suitable machine layout for the development of a novel positioning system in the preliminary stages of a project. 347
To introduce the errors from the machine elements into the HTM, the translation and error matrices were determined, as illustrated in equation (4). In that equation, a T berror is the HTM from axis a to axis b considering errors, while a T b is the translation matrix from axis a to axis b without errors. The error matrix (a E b ) considers only geometric error from the interface between axis a and axis b, where X, Y and Z are the position elements in the x-, y- and z-directions, respectively, while (1x, 1y, 1z) and (dx, dy, dz) are the rotational and linear
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
Figure 6 FDM machine schematic (schematic basic dimensions) and axes (X, Y, Z, T)
1 0 0 X
3 2
2 1z 1 1x
1y
dx
T berror
6 60 1 6 a T b a E b 6 6 60 0 4 0 0 2 1 6 6 1z 6 6 6 6 21y 4 0
0 3 X dx 7 Y dy 7 7 7 7 Z dz 7 5 1
7 21x dy 7 7 7 7 1 dz 7 5 0 1
7 6 7 6 1z1 1 21x1 25 dy1 7 6 6 7 7 6 21y1 1x1 1 50 d z 1 5 4 0 0 0 1 3 2 1 21z1 1y1 250 2 X dx1 7 6 7 6 1z1 1 21x1 25 dy1 7 6 6 7 7 6 21y1 1x1 1 50 d z 1 5 4 0 0 0 1 5 2 1 21z2 1y2
21z1
1y1
150 2 X dx1
dx2
After applying the main equipment dimensions shown in Figure 6, the HTM of each axis of the system was found, as presented in equations (5)-(7). It is important to note that the errors of those equations were dependent on the machine element errors and manufacturing tolerances: 348
7 6 6 1z2 1 21x2 125 2 Y dy2 7 7 6 6 7 7 6 21y2 1x2 1 2 25 d z 2 5 4 0 0 0 1 3 2 1 21z2 1y2 dx2 7 6 6 1z 1 21x2 225 2 Y dy2 7 7 6 2 7 6 7 6 2 1 1 1 2 25 d z 7 6 y2 x2 2 5 4 0 0 0 1 6
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
7 6 6 1z 1 21x3 2150 dy3 7 7 6 3 7 6 7 6 1 225 Z dz3 7 6 21y3 1x3 5 4 0 0 0 1 3 2 1 21z3 1y3 2250 dx3 7 6 6 1z 1 21x3 2250 dy3 7 7 6 3 7 6 7 6 1 225 Z dz3 7 6 21y3 1x3 5 4 0 0 0 1 7
21z3
1y3
2150 dx3
Minimize ! f cos t ; error Ctotal w Error 1 2 w s:t : 1 # Cg # 5 1 # Ct # 3 1 # C m # 15 This equation also shows the constraint functions, which bound the relative cost of the linear guide between 1 and 5. Additionally, the cost of the transmission elements was limited between 1 and 3, while the relative cost of the manufacturing tolerances could be as great as 15. Initially, we dened all the variables in the optimisation study as 1 to identify the local optimum that provided the lowest values for the relative cost. This denition forced the optimisation problem to indicate the lowest solutions despite the existence of other local optimum values. 9
Finally, the maximum equipment error can be found from equation (8), which describes the multiplication of the HTM, which takes errors into consideration, and the inverse matrix, which takes into account the expected position of the machine (Venkatesh and Izman, 2008; Slocum, 1992). In this case, because we have considered the centre of the table as the origin of the system, the values of (X, Y, and Z) are null:
2
21
dx
0 0
321
7 Y7 7 7 7 Z7 5 1
Optimisation of system For the optimisation study, we applied the goal attainment method, by dening a problem with multiple objectives. In addition, a Newton modied algorithm was performed to solve the cost-error system (Rao, 2009). Due to the dependency between the tolerance and the cost of machine elements, we dened seven variables to perform the optimisation study: Cg x ; C g y ; C g z ; C t x ; C t y ; C tz ; and Cm . As these variables refer to the cost of components and manufacturing processes, their respective machine element tolerance values had to be updated in each iteration of the optimisation study. For this study, only the main dimensions of the FDM 1600 layout were considered, whereas equipment with different built areas or layouts implies different optimisation results. Equation (9) shows the optimisation problem, which consists of the minimisation of the total cost and nal error. In this equation, the weight of objective function (w) represents the importance of each function to the optimisation study:
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
Error direction
which provided the cheapest solution (relative cost 1), provided an average error of 733.69 mm, while the minimum error was found in the y-direction (approximately 718.7 mm). In Figure 7, it can also be observed that the minimal average error (37.19 mm) implies an increase of 2.83 times the equipment cost, which was found for the importance weight equal to 0. In that case, the deviation of the errors was 0.51 mm, the highest error value found in the z-direction. It can be noted that this analysis may be relevant for the identication of the protability of equipment layout during the preliminary phases of a project, expressing graphically the advantages and disadvantages that inuence the changes of accuracy and cost. Additionally, the Pareto optimum also indicates the limit between feasible and unfeasible solutions, emphasising the minimal cost that can be achieved for the same error. Table II indicates the equivalent machine elements for the solutions found in the Pareto optimum. Through this table, it is possible to observe the values of the total relative costs, importance weight and errors of the respective optimal solutions. In addition, the design parameters that promote these results are indicated, such as the tolerances of the manufacturing processes, linear guides and transmission elements. Another point that was made apparent in this study is that currently available equipment are already made with
specications close to the optimum solutions, which employ importance weights equal to 0.5 (balanced solution). This emphasises that the current equipment layout concepts are already well dimensioned. Despite this nding, if the cost of the equipment remains constant, the implementation of the optimum solution would decrease the error by 20 mm. Additionally, this study also indicates that, for the same error, it is still possible to decrease the cost of the equipment by 10 per cent. Moreover, the increase in accuracy regardless of the reduction of or remaining nal error might imply a change in the layout concept, instead of a simple adjustment of the accuracy of the machine elements. Therefore, it would be possible to improve the accuracy of the equipment by almost 120 mm, if the relative cost of equipment increased 1.26 times. It is also important to note that although this study considered only geometric errors, by ignoring other potential mechanical, electronic and controlling errors, such as a dynamics error or motor resolution, this study can be a useful tool for determining positioning system concepts and improving the accuracy of additive manufacturing equipment, such as FDM machines.
Conclusion
This work evaluated a general model for the estimation of the nal error of the layout of a rst generation FDM positioning system.
Figure 7 Pareto optimum of optimisation study, relating optimal values of cost function and average of nal error
3
2 1.9 Feasible Solutions FDM 1,600 balanced solutions Optimal Pareto
Cost
2.5
2 Cost
1.5 100 125 150 Final Error (m) 175 200
1.5 Feasible Solutions 1 Unfeasible Solutions 0.5 0 100 200 300 400 Final Error (m) 500 600 700 800
350
Weight X error (mm) Y error (mm) Z error (mm) Average error (mm) Relative cost Manufacturing tolerance Linear guide xy
351
2 37.38 2 43.09 2 52.85 2 63.24 2 78.07 2 105.6 2 140.1 2 176.6 2 240.4 2 399.3 2 730.8
2 36.62 2 41.94 2 51.13 2 60.99 2 75.22 2 102.1 2 135.8 2 171.2 2 233.1 2 388.3 2 718.7
2 37.58 2 44.45 2 71.66 2 109.5 2 156 2 216.6 2 294.5 2 383 2 541.8 2 740.9 2 751.6
37.19 43.16 58.55 77.92 103.09 141.46 190.10 243.61 338.48 509.47 733.69
2.83 2.51 2.17 1.94 1.75 1.57 1.41 1.30 1.18 1.06 1.00
IT1 IT2 IT2 IT3 IT4 IT4 IT5 IT5 IT6 IT6 IT7
Linear guide (UP) Linear guide (SP) Linear guide (P) Linear guide (C) Linear bearing (H) Linear bearing (H) Linear bearing (C) Linear bearing (C) Linear bearing (C) Linear bearing (C) Linear bearing (C)
Ball screw C5 Ball screw C5 Ball screw C5 Ball screw C5 Ball screw C7 Ball screw C7 Timing belt pulley IT7 Timing belt pulley IT7 Timing belt pulley IT7 Ball screw C10 Lead screw
Ball screw C5 Ball screw C5 Ball screw C7 Timing belt pulley IT7 Timing belt pulley IT7 Ball screw C10 Ball screw CIO Lead screw Lead screw Lead screw Lead screw
Design of an FDM positioning system Marlon Wesley Machado Cunico and Jonas de Carvalho
In addition, a systematic approach to select the main machine elements for similar FDM positioning systems, such as LOM, SDP and 3DP equipment, was presented. Despite the acceptance of the high prices of advanced manufacturing, reduction of manufacturing costs implies an increase in prot of the equipment, emphasising the importance of this work during the conceptual design. A deviation of 2 mm was found by comparing the proposed numerical model, which considered only geometric errors, to two commercially available equipment units. This deviation was found to be the average of the nal error between the higher FDM 8000 equivalent model and the lower FDM 1600 equivalent model. This could possibly be the result of the increase in tolerance as a function of larger dimensions, despite the use of the same IT grade. In addition to the identication of cost-accuracy functions, the minimisation of the optimisation study resulted in a nal error of 37 mm for maximum accuracy. In order for the cost function to equal the error function, the optimal result was found to equal 141.46 mm, which is lower than the equivalent value of the commercial equipment that we used as the basis of our study (168 mm). The design layout of the solution that obtained the highest accuracy employs ultra-precision grade linear guides and C5 grade ball screws, in addition to a manufacturing tolerance equal to an IT7 grade. However, this concept requires an increase of 2.83 times the total cost of cheapest equipment concept studied. Additionally, it was found that the current layout of commercially available equipment is close to the optimum solution of the optimisation problem if the weight of the functions were considered equal at a value of 0.5. This may indicate that the improvement of FDM accuracy should occur in parallel with the change of layout design to prevent a signicant increase in the total cost of equipment.
References
Crump, S.S., Batchelder, S.J., Sampson, T., Zinniel, R.L. and Barnett, J. (2009), System and method for building threedimensional objects with metal-based alloys, US Patent, Stratasys, Inc., Edina, MN. Cunico, M.M.W.M. (2011), Development of new rapid prototyping process, Rapid Prototyping Journal, Vol. 17 No. 2, p. 6. Dong, Z., Hu, W. and Xue, D. (1994), New production costtolerance models for tolerance synthesis, Journal of Engineering for Industry Transaction of ASME, Vol. 116, pp. 199-206. Dornfeld, D.A. and Lee, D.E. (2007), Precision Manufacturing, Springer, New York, NY, p. 775.
Gates (2006), Timing Belt Theory, Gates Mectrol, p. 18, available at: www.gates.com/ (accessed June 2011). Gates (2011), Urethane Timing Belts and Pulley, Gates Mectrol, p. 64, available at: www.gates.com/ (accessed June 2011). Gibson, I., Rosen, D.W. and Stucker, B. (2010), Photopolymerization processes, Additive Manufacturing Technologies, Springer, New York, NY, pp. 61-102. Hiwin (2006), Linear Bearing Technical Information, HIWIN Linear Motion Products & Technology, p. 252, available at: www.hiwin.com/ (accessed June 2011). Hiwin (2008), Ball Screw Technical Information, HIWIN Linear Motion Products & Technology, p. 252, available at: www.hiwin.com/ (accessed June 2011). Rao, S.S. (2009), Engineering Optimization: Theory and Practice, Wiley, New York, NY. Shigley, J.E. and Mischke, C.R. (1996), Standard Handbook of Machine Design, McGraw-Hill, New York, NY, p. 1723. Slocum, A.H. (1992), Precision Machine Design, SME, Dearborn, MI, p. 750. Stratasys (1994), FDM System Documentation , Reference Manual, Stratasys, Edina, MN. Stratasys (1998), FDM System Documentation , Reference Manual, Stratasys, Edina, MN. Tan, K.K., Lee, T.H. and Huang, S. (2008), Precision Motion Control: Design and Implementation, Springer, New York, NY, p. 272. THK (2007), Power Linear Bush, THK, p. 12, available at: www.thk.com (accessed June 2011). THK (2011), Precision Ball Screws: DIN Standard Compliant Ball Screw, THK, p. 36, available at: www.thk.com (accessed June 2011). Thomson (2009a), Lead Screws, Ball Screws and Ball Splines, Thomson Linear, p. 252, available at: www.thomsonlinear. com/ (accessed June 2011). Thomson (2009b), Thomson Roundrail Linear Guides and Components, Thomson Linear, p. 268, available at: www.thomsonlinear.com/ (accessed June 2011). Tong, K., Lehtihet, E.A. and Joshi, S. (2004), Software error compensation for rapid prototyping, Precision Engineering, Vol. 28 No. 3, pp. 280-292. Venkatesh, V.C. and Izman, S. (2008), Precision Engineering, McGraw-Hill, New York, NY, p. 418. Wang, B. (2010), Integrated Product, Process and Enterprise Design, Vol. 2, Springer, New York, NY, p. 492.
Corresponding author
Marlon Wesley Machado Cunico can be contacted at: marloncunico@yahoo.com.br
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints
352