Sunteți pe pagina 1din 10

Journal of Manufacturing Systems Vol. 19/No.

4 2000

Stamping Strip Layout for Optimal Raw Material Utilization


T.J. Nye, Mechanical Engineering Dept., McMaster University, Hamilton, Ontario, Canada. E-mail: nyet@mcmaster.ca

Abstract
Stamping dies are used to produce very large numbers of identical parts from sheet metal. Due to the high volumes of parts produced, even small inefficiencies in material utilization per part can lead to very large amounts of wasted material over a dies life. This paper develops an exact algorithm for orienting the part on the strip to maximize material utilization. The algorithm optimally nests convex or nonconvex blanks on a strip and predicts both the orientation and strip width that minimize material usage. Technological constraints, such as blank orientation constraints due to planar anisotropy, are also incorporated into the algorithm. The algorithms use is demonstrated with examples that show how sensitive material utilization can be to small changes in blank orientation in the die. Keywords: Stamping, Die Design, Optimization, Material Utilization, Minkowski Sum, Design Tools

Introduction
During the design process for stamping dies, decisions must be made about the orientation of the stamped part on the strip. The orientation determines how efficiently raw material is utilized, and in an operation such as stamping where large amounts of material are processed, small inefficiencies per piece can accumulate into huge wastes of material in the long term. Maximizing material utilization in stamping is of paramount importance. Raw materials typically represent 75% or more of total costs in stamping facilities,1 so a poorly designed die can significantly increase a companys operating costs over its life. For example, in a blanking press running at 200 strokes per minute, a die will save one ton of material in each eight-hour shift if its design is adjusted to save just 10 grams per part. Once tooling is built, the amount of material waste per part is fixed for the (usually long) life of the tools. Thus, it is very important to design tooling from the start to minimize waste.

Substantial life cycle savings result by selecting optimal blank orientation when the tooling is designed. This task, however, is quite challenging to perform manually because, as the blank orientation changes, both the width and the pitch (that is, the distance between adjacent blanks) of the blank on the strip change. The measure of interest is the utilization of the strip material, which is a function of the area of strip material used per blank. Selecting the orientation that minimizes pitch does not necessarily maximize material utilization, and manually evaluating area when both the pitch and width change is difficult. The computational needs of this problem are well suited to the capabilities of CAD systems, provided an algorithm such as the one described here is available. Applications of this algorithm are in computer-aided design of stamping dies and in expert systems for automated tooling generation, such as the systems demonstrated in recent years by Cheok and Nee,2 Cheok, Foong, and Nee,3 Choi et al.,4 Huang, Ismail, and Hon,5 Ismail, Chen, and Hon,6 Lin and Hsu,7 Lu, Weidong, and Lihua,8 Prasad and Somasundaram,9 and Singh and Sekhon.10 This paper is concerned with the problem of orienting a single, arbitrary blank on a strip so that raw material is optimally utilized. The blank is represented as a simple polygon, that is, as an irregular polygon that may or may not be convex, but has no holes. (Holes in a stamping are ignored for blank layout.) Blanks with curved edges are approximated as polygons, with the approximation improving as the number of vertices on the polygon increases. It is assumed that the width of the strip is determined during the layout optimization rather than being prespecified before the layout work starts. The cost of the raw material is assumed to be a fixed value per unit weight (or unit area) regardless of the width chosen for the strip; that is, there are no penalties for selecting widths that are not stock or preferred

239

Journal of Manufacturing Systems Vol. 19/No. 4 2000

widths. In the remaining discussion, bridge width, wB, the extra amount of material around the blank that forms the skeleton, is accommodated by growing the blank by one half the bridge width before optimization starts. Optionally, an additional edge width, wE, can be added to the skeleton, as shown in Figure 1. Naturally, blanks are not allowed to overlap, and strip width, wS, is set equal to the minimum width necessary for the blank at a given orientation. Given these assumptions, this paper describes a new, exact algorithm for finding optimal blank layout on a strip. Previous work on this problem is surveyed in the next section, followed by an introduction to the Minkowski sum, a concept from the computational geometry field that is especially useful in determining whether adjacent polygons overlap. The Minkowski sum is then incorporated into an algorithm that optimizes material utilization, and an example is used to demonstrate the use of the algorithm. Finally, a discussion of conclusions is given.

wS

wB

wE

Figure 1 Adding Bridge Width to Blanks

Previous Work
The earliest approach to this problem seems to be due to Adamowitz and Albano.11 They proposed packing a blank into a rectangular box of minimum area that is then repeated along a strip. The orientation of the blank in the minimum area box is used as the optimal orientation when designing the blanking die. Nee12 and Martin and Stephenson13 have also used this approach. In general, however, this approach does not lead to optimal layouts. For example, consider a parallelogram-shaped blank. A parallelogram will nest along a strip with zero waste, but an enclosing rectangle will add waste material at each end of the blank. Thus, using the enclosing rectangle can preclude an optimal layout from being achieved. Enclosing the blank within other packing shapes has also been examined. Chow14 packed blanks into known interlocking shapes that were then repeated along the strip. Qu and Sanders15 packed blanks into composites of nonoverlapping rectangles. Dori and Ben Bassat16 and Karoupi and Loftus17 packed blanks into convex polygons. Just as with packing blanks into rectangles, these approaches build waste material into the layout that may prohibit finding the optimal layout for a specific shape. Meta-heuristics have also been applied to the blank layout problem. Simulated annealing has been used by Jain, Feynes, and Richter18 and Theodoracates and

Grimsley.19 Ismail and Hon20 have successfully used genetic algorithms for such a layout problem. These methods perform well relative to other methods when a number of different parts need to be nested together, as the complexity of such problems increases combinatorially. Unfortunately, when determining the layout for a single blank on a strip, they suffer from the disadvantages of not being guaranteed to find the optimum orientation, and they may require significant computational times. Incremental rotation algorithms have been proposed by Chow,14 Nee,21 Prasad, Somasundaram, and Rao,22 and Lin and Hsu.7 In these, the blank is rotated by a fixed increment, then the strip width and pitch are determined at that orientation. After rotating the blank through many increments, the orientation that provided the best material utilization is chosen. These algorithms, as proposed, are not exact in that they only examine utilization at fixed increments such as one (such as Prasad, Somasundaram, and Rao) or two (such as Nee12) degrees. The optimum orientation will usually fall between increments, and because even small inefficiencies per blank can accumulate into significant wastes of raw material over the life of a die, there is value in improving on this approach. The automated die design systems mentioned previously,2-10 when described, all utilize incremental rotation algorithms in their blank layout modules. An exact blank layout algorithm has been described by Joshi and Sudit.23 They started by providing a proof that the optimal layout of blanks placed on a strip of finite length will converge to the optimal layout on blanks of a strip of infinite length as the finite strip becomes relatively long, as is common for

240

Journal of Manufacturing Systems Vol. 19/No. 4 2000

the raw material in metal stamping. To ensure a nonoverlapping constraint between adjacent blanks was satisfied for any blank orientation, the concept of the obstacle space was employed. The obstacle space, which is closely related to no-fit polygons and the Minkowski sum, is a geometric construction in which one polygon is grown by the shape of another polygon. If the second polygon is translated such that its reference point falls on the perimeter of this obstacle space, it will touch but not overlap the first polygon. By using the obstacle space to define the feasible set of relative translations, Joshi and Sudit showed that when the width of the strip is greater than that required by the blank in any orientation, the optimal orientation is that which minimizes the pitch between blanks, that is, the orientation that gives the smallest distance from the center of the obstacle space to a point on its perimeter. Joshi and Sudit then adapted this method to provide optimal blank orientations for the case where the width of the strip is predetermined and may be less than the blank requires for particular orientations. In stamping, predetermined strip width is usually not the case, however. Die design normally proceeds without restriction on strip width, and once the design is finalized, the minimum necessary strip width is determined and specified for material purchase. The following discussion introduces a new exact algorithm, one that is capable of determining jointly optimal orientations and strip widths through the use of the Minkowski sum, a concept similar to the obstacle space used by Joshi and Sudit.23
y

The Minkowski Sum


The Minkowski sum is a concept found in the computational geometry field24 that finds application in areas such as robotic motion planning (such as Canny25) and optimization of cutting patterns in garment production (for example, Li and Milenkovic26). The power of the Minkowski sum is that its use simplifies a polygon-polygon overlap query into a pointin-polygon query, significantly reducing computational effort. When applied to the blank layout on a strip problem, it also greatly simplifies pitch and width calculations, as will be shown. The Minkowski sum for two polygons, A and B, is defined as follows: A B = {a + b | a A, b B} (1)

This is shown graphically in Figure 2. For detecting overlaps between polygons, say A and B, the Minkowski sum A ( B) is used, where ( B) is polygon B rotated 180 around the origin. If A ( B) contains the origin, A and B overlap,27 as shown in Figure 3. The Minkowski sums shape is invariant for a given A and B but translates as A or B translate. For blank layout, the same blank is repeated on the strip, so the relevant Minkowski sum is A ( A). For example, the Minkowski sum for a parallelogram-shaped blank is shown in Figure 4. Regardless of the coordinate system origin used for defining the vertices of A, A ( A) is centered at the origin. If two blanks are considered, A1 and A2, they will overlap if A1 ( A2) contains the origin. To eliminate
y

A B

A B

A A ( B) x
a

B
b

( B)

Figure 2 Minkowski Sum of Two Polygons

Figure 3 Overlap Detection with Minkowski Sums

241

Journal of Manufacturing Systems Vol. 19/No. 4 2000

Width

A Sweepline

A ( A) Pitch A ( A) A

Figure 4 Minkowski Sum for a Parallelogram-Shaped Blank


A

Sweepline

overlaps, the blanks need to be translated relative to each other. This relative translation causes a similar translation of the Minkowski sum, and when the Minkowski sum no longer contains the origin, the blank overlap is eliminated. This property leads to the observation that the distance from the origin to the perimeter of A ( A) in any given direction is the spacing, or pitch, between blanks on a strip oriented in that direction. Thus, for a given blank, finding blank pitch for any strip orientation is simply a matter of finding the distance to the edge of the Minkowski sum. This is shown in Figure 5, where the strip orientation is described by a sweepline vector, that is, a vector anchored at the origin that sweeps through a range of orientations. Similarly, the strip width for any blank orientation is just the maximum perpendicular distance between the strip longitudinal axis (through the origin) and the perimeter of A ( A). This is shown in Figure 5 by the dashed lines parallel to the sweepline vector. With these two observations, material utilization can be easily calculated as a function of blank orientation once the Minkowski sum is generated.

A Width

( A)

Pitch

Sweepline A

Pitch

( A)

Width

Figure 5 Use of Minkowski Sum in Finding Strip Pitch and Width

AB =

1 n ( xi yi +1 yi xi +1 ) 2 i =1

(3)

Utilization Optimization
The raw material utilization, , is defined as follows:
= AB AB = AS p wS

(2)

where AB is the blank area, AS is the strip area, is the strip pitch, and wS is the strip width. For a polygon, blank area is given by the following:24

where (xi, yi) are the coordinates of vertex i on the polygonal blank. (The n vertices are assumed numbered consecutively in CCW order around the blank, with a modulus-n numbering system.) To calculate strip pitch and strip width, an angular sweepline is used within the Minkowski sum A ( A). The sweepline is anchored at the origin and traverses through a total rotation of 180 (a rotation of 360 is not necessary due to symmetry). In physical terms, the sweepline can be thought of as the longitudinal axis of the strip. Rotating the sweepline and strip in the CCW direction through

242

Journal of Manufacturing Systems Vol. 19/No. 4 2000

an angle of is equivalent to rotating the blank in the CW direction through an angle of on a stationary strip. The pitch (distance from the origin to the Minkowski sum edge) varies smoothly as the sweepline traverses an edge of the Minkowski sum, but changes abruptly as the sweepline crosses a vertex to the next edge. These points where the sweepline passes vertices are termed event points. A given edge traversed by the sweepline is defined by the start vertex, vS, and the finish vertex, vF, with vS being located CW of vF. As the sweepline crosses the event point (that is, when the sweepline reaches vF and an angle of tan 1(yF/xF)), these vertices each increment to the next vertex on A ( A). This process is shown in Figure 6. The sweepline starts at vertex 2 (top figure), giving vS = 2 and vF = 3. As the sweepline rotates CCW, it traverses edge 2-3. Once vertex 3 is reached (bottom figure), the sweepline passes on to edge 3-4, incrementing the start and finish vertices to vS = 3 and vF = 4, respectively. Between the event points, strip pitch, , is calculated as the distance from the origin to the intersection point of the sweepline and Minkowski sum (see the Appendix for the derivation), that is,
p=

vF, vT

A 1

A) 2 vS

vT

vF

vS

4 vF, vT

3 vS

( x F yS x S y F ) ( yS yF )cos() + ( x F xS )sin()

(4)
1 2

Strip width is the maximum perpendicular distance between the sweepline and the points on the Minkowski sum. This distance will always coincide with a vertex on the Minkowski sum, termed the top vertex, vT. For example, in Figure 6 the maximum perpendicular distance from the sweepline is shown by the dotted lines. The top vertex starts at vertex 3, then moves to vertex 4 as the sweepline becomes parallel to edge 3-4. As the sweepline rotates, the top vertex progresses around the Minkowski sum, and the complete set of top vertices forms the convex hull of the Minkowski sum. (This can be seen in Figure 8.) This shifting of top vertices from one vertex to the next also triggers event points. These event points occur when the sweepline becomes parallel to the edge of the convex hull vTvT+1, that is, at the angle tan 1((yT+1 yT)/(xT+1 xT)). For a given top vertex, the strip width is then found as follows:

Figure 6 Shifting of Event Vertices as Sweepline Rotates

wS = xT sin( ) + yT cos( ) + 2wE

(5)

The quantity 2wE is included to add edge width to the strip (Figure 1). Strip area becomes:
AS =

( x F yS xS yF )( xT sin() + yT cos() + 2wE ) ( yS yF )cos() + ( x F xS )sin()

(6)

Material utilization can be expressed as a function of sweepline orientation through the use of Eqs. (2) to (5):
= AB ( x F yS xS yF )( xT sin() + yT cos() + 2wE ) ( yS yF )cos() + ( x F xS )sin()

(7)

243

Journal of Manufacturing Systems Vol. 19/No. 4 2000

Taking the first derivative and simplifying gives the following:


d = d AB

(8)

vertices on the Minkowski sum so that a range of 180 is spanned. These angles are the event points for vS and vF. Step 3: If wE 0, for each pair vS and vF, calculate *. If * falls between vS and vF, include an event point for *. Step 4: Calculate the convex hull of the vertices on the Minkowski sum. Step 5: Find the angles of the edges of the convex hull. These angles are the event points for vT. Step 6: Add event points at the limits of technologically feasible orientations. Delete any event points that are technologically infeasible. Step 7: Calculate strip utilization at each event point. Select the orientation(s) with the highest utilization as the optimal orientation(s). Algorithms for the calculation of the Minkowski sum (step 1) and convex hull (step 4) can be found in computational geometry texts, such as ORourke24 or de Berg et al.27 Technological constraints on blank orientation, such as due to planar anisotropy, are dealt with simply in this algorithm. New event points are added to the problem at the limits of technologically acceptable rotations. Any existing event points at nontechnologically feasible rotations are then deleted. All remaining event points are technologically feasible, so only a feasible orientation will be chosen as optimal.

( x F xS ) yT ( yS yF ) xT + 2wE (( x F xS )cos() ( yS yF )sin()) ( x F yS xS yF )( xT sin() + yT cos() + 2wE )2

The behavior of can be considered for two cases. First, if no additional edge width is added to the strip, wE = 0 and the third term in the numerator of Eq. (8) vanishes. The rest of the numerator is invariant to changes in , as is the first term in the denominator. The second term in the denominator is a function of , but because it is squared, its sign will always be positive. Thus, the sign of Eq. (8) is invariant with changes in , so utilization will either monotonically increase, monotonically decrease, or remain constant as the sweepline passes along the edge vSvF. This implies that optima must occur at the rotation constraints, that is, the event points. The global optimum is then found by simply evaluating the utilization at each event point where the sweepline reaches a vertex on the Minkowski sum and selecting the event point orientation that provides the best utilization. In the second case, where wE is not zero, setting Eq. (8) to zero and solving for * gives the following:
2 ( x F x S ) f + ( x F x S ) yS yT + xT (2 yS yF yS ) + x S yT yF * = tan 1 2 ( yS yF ) f yT ( x F x S ) + xT ( x F x S )( yS yF )

(9)

where:
2 f = 4 wE ( x F x S ) + ( yS yF ) (( x F x S ) yT + ( yS yF ) xT ) 2 2

Examples
Thus, it may be possible for a local maximal point to occur at a sweepline angle between vS and vF. In this case where wE is not zero, * must be evaluated for each edge of the Minkowski sum. On a particular edge, if * falls between vS and vF, a new event point is created at * and its utilization evaluated. The optimization algorithm is then constructed as follows: Step 1: Calculate the Minkowski sum A the polygonal blank A. ( A) of Figure 7 shows a T-shaped blank surrounded by its associated Minkowski sum. The convex hull of the Minkowski sum is also shown with dotted lines. The sweepline is assumed to start at v4 and travel CCW until it reaches v10, for a total range of rotation of 180. This process is shown for each event point in Figure 8, read right to left, top to bottom. Each drawing in Figure 8 shows the sweepline at an event point, either for shifts in vS and vF vertices or for shifts in vT vertices. The angle of each event point was calculated, along with the relevant pitch [Eq. (4)], strip width [Eq. (5)], and utilization [Eq. (7)], and summarized in Table 1. (No technological con-

Step 2: Find the angles between the origin and the

244

Journal of Manufacturing Systems Vol. 19/No. 4 2000

( 6, 8) v8 ( 4, 8) (0, 6) ( 4, 6) ( 10, 2) v10 ( 6, 2) v9 (0, 0) (2, 0) (2, 6)

(6, 8) v7

(6, 6) (10, 2) v6 (6, 2) v5

v11 ( 10, 2)

( 6,

2) v12

v3 (6,

2) v4 (10, 2)

v1 ( 6, 8)

v2 (6, 8)

Figure 7 Example Blank (Heavy Line), Its Minkowski Sum (Solid Line), and the Convex Hull (Dotted Line)

Table 1 Event Points for T-Shaped Blank Example

Event Point 1 2 3 4 5 6 7 8 9 10 11

Angle () -11.31 0.00 11.31 18.43 53.13 56.31 90.00 123.69 126.87 161.56 168.69

Event

Pitch

Width Utilization 9.022 8 9.022 9.486 9.600 9.430 10 9.430 9.600 9.486 9.022 34.8% 40.0% 34.8% 53.3% 33.3% 35.3% 40.0% 35.3% 33.3% 53.3% 34.8%

vS 4, vF 5, 10.198 vT 7 vT 8 10 vS 5, vF 6 10.198 vS 6, vF 7 6.324 vS 7, vF 8 10 vT 10 9.614 vT 11 8 vT 1 9.614 vS 8, vF 9 10 vS 9, vF 10 6.324 vS 10, vF 11 10.198

straints have been included in this example and wE = 0.) Inspection shows the optimal blank orientation occurs at = 18.43 and = 161.56, with a corresponding material utilization of 53.3%. The value of optimal blank layout can also be demonstrated using this example. Suppose the dimensions of the part are in centimeters and the part is to be made from 3 mm thick steel. Suppose also that the part is produced on a 200 stroke-perminute blanking press an average of 10 hours per week. If a conventional incremental rotation algorithm was used to design the strip layout, with a oneor two-degree increment angle, the blank would be

Figure 8 Sweepline Event Points. The Dotted Line Shows the Maximum Perpendicular Distance from the Sweepline (Heavy Arrow) to the Edge of the Minkowski Sum.

oriented at 18. The optimal algorithm developed here would select an orientation of 18.43. The difference, less than one half degree, if ignored would lead to more than 18 tons of additional strip material wasted per year. Over its life, the value of this extra waste could well exceed the cost of the blank-

245

Journal of Manufacturing Systems Vol. 19/No. 4 2000

55.0% 50.0% 45.0% 40.0% 35.0% 30.0% 20 30 80 Sweepline angle (degrees) 130 180

Material utilization (percent)

Figure 9 Material Utilization as a Function of Sweepline Angle

Figure 10 A More Complex Part in Its Optimal Orientation

ing die itself. Thus, the effort necessary to optimize blank orientation is amply rewarded. Utilization as a continuous function of sweepline angles has also been calculated (using an incremental rotation algorithm with incremental rotations interpolated between the event points) and is presented in Figure 9. The large slopes of portions of the utilization function demonstrate that the efficiency of a blanking die can be very sensitive to blank orientation. A second example, shown in Figure 10, is based on a more complexly shaped part, described in Choi et al.4 This part contains 22 vertices, and it leads to 34 event points being investigated. The utilization function at various orientations is given in Figure 11. As can be seen by the single peak, optimal utilization is sensitive to blank orientation in this part as well. Execution time for the algorithm with this part was 0.10 seconds on a 550 MHz Pentium III PC. In contrast, the execution time using this algorithm for a proprietary automotive blank with 442 vertices, leading to 460 event points, was 11.9 seconds. Clearly, the costs involved with employing this algorithm during the design of a die are negligible compared to potential lifetime material savings.

0.6 0.5 Material utilization 0.4 0.3 0.2 0 50 100 Blank orientation (degrees) 150

Figure 11 Utilization as a Function of Orientation for the Part from Figure 10

Conclusions
A new, exact algorithm for optimally orienting blanks on strips has been described in this paper. This algorithm overcomes the limitations of previous approaches to this problem to quickly calculate blank orientations and strip widths that maximize material utilization. Implementation in CAD systems and tooling-design expert systems is quite straightforward. The relative ease with which mater-

ial utilization is calculated is a benefit of using the Minkowski sum, which no doubt will also prove very useful in other types of nesting problems. As stamping is almost by definition a high-volume manufacturing operation, even tiny per-part savings can accumulate into large potential material savings over a stamping tools life. The algorithm developed here will add negligible time or cost to the tooling design process, so its use is justified in any tooling design effort. A number of research opportunities arise from this work. For instance, it is generally known that blanking more than one part from the same strip increases material utilization. An exact algorithm to optimize the layout of two or more blanks on a strip would be very useful for these problems. A special subclass, which might be computationally more efficient, is when blanks are mirror images of each other. Another useful domain extension is to consider 2D layout problems, such as are encountered when cutting blanks from sheets. The Minkowski sum is likely to be a useful tool in these problems as well.

246

Journal of Manufacturing Systems Vol. 19/No. 4 2000

References
1. Industry Canada, Industry Overview Reports: SIC-E 3253 - Motor Vehicle Stampings Industry (Ottawa, Ontario, Canada: Nov. 22, 1998). 2. B.T. Cheok and A.Y.C. Nee, Configuration of Progressive Dies, Artificial Intelligence for Engg. Design, Analysis and Mfg. (v12, 1998, pp405-418). 3. B.T. Cheok, K.Y. Foong, and A.Y.C. Nee, An Intelligent Planning Aid for the Design of Progressive Dies, Proc. of Institution of Mechanical Engineers (Part B, v210, 1996), pp25-35. 4. J.C. Choi, B.M. Kim, H.Y. Cho, and C. Kim, A Compact and Practical CAD System for Blanking or Piercing of Irregular-Shaped Metal Products and Stator and Rotor Parts, Intl Journal of Machine Tools and Manufacture (v38, 1998), pp931-963. 5. K. Huang, H.S. Ismail, and K.K.B. Hon, Automated Design of Progressive Dies, Proc. of Institution of Mechanical Engineers (Part B, v210, 1996), pp367-376. 6. H.S. Ismail, S.T. Chen, and K.K.B. Hon, Feature-Based Design of Progressive Press Tools, Intl Journal of Machine Tools and Manufacture (v36, 1996), pp367-378. 7. Z.C. Lin and C.Y. Hsu, An Investigation of an Expert System for Shearing Cut Progressive Die Design, Intl Journal of Advanced Mfg. Technology (v11, 1996), pp1-11. 8. W. Lu, Z. Weidong, and T. Lihua, A CAD/CAM System for Multiple-Step Precision Progressive Dies, Advanced Technology of Plasticity 1993Proc. of 4th Intl Conf. on Technology of Plasticity (1993), pp1710-1715. 9. Y.K.D.V Prasad and S. Somasundaram, CADDS: An Automated Die Design System for Sheet-Metal Blanking, Computing and Control Engg. Journal (v3, July 1992), pp185-191. 10. R. Singh and G.S. Sekhon, A Low-Cost Modeller for TwoDimensional Metal Stamping Layouts, Journal of Materials Processing Technology (v84, 1998), pp79-89. 11. M. Adamowicz and A. Albano, Nesting Two-Dimensional Shapes in Rectangular Modules, Computer Aided Design (v8, 1976), pp27-33 12. A.Y.C. Nee, Computer Aided Layout of Metal Stamping Blanks, Proc. of the Institution of Mechanical Engineers (v198B, n10, 1984), pp187-194. 13. R.R. Martin and P.D. Stephenson, Putting Objects into Boxes, Computer Aided Design (v20, 1988), pp506-514. 14. W.W. Chow, Nesting of a Single Shape on a Strip, Intl Journal of Production Research (v17, 1979), pp305-322. 15. W. Qu and J.L. Sanders, A Nesting Algorithm for Irregular Parts and Factors Affecting Trim Losses, Intl Journal of Production Research (v25, 1987), pp381-397. 16. D. Dori and M. Ben Bassat, Efficient Nesting of Congruent Convex Figures, Communications of the ACM (v27, 1984), pp228-235. 17. F. Karoupi and M. Loftus, Accommodating Diverse Shapes Within Hexagonal Pavers, Intl Journal of Production Research (v29, 1991), pp1507-1519. 18. P. Jain, P. Feynes, and R. Richter, Optimal Blank Nesting Using Simulated Annealing, Trans. of ASME, Journal of Mechanical Design (v114, 1992), pp160-165. 19. V.E. Theodoracates and J.L. Grimsley, The Optimal Packing of Arbitrarily-Shaped Polygons Using Simulated Annealing and PolynomialTime Cooling Schedules, Computer Methods in Applied Mechanics and Engg. (v125, 1995), pp53-70. 20. H.S. Ismail and K.K.B. Hon, New Approaches for the Nesting of Two-Dimensional Shapes for Press Tool Design, Intl Journal of Production Research (v30, 1992), pp825-837. 21. A.Y.C. Nee, A Heuristic Algorithm for Optimum Layout of Metal Stamping Blanks, Annals of the CIRP (v33, 1984), pp317-320. 22. Y.K.D.V. Prasad, S. Somasundaram, and K.P. Rao, A Sliding Algorithm for Optimal Nesting of Arbitrarily Shaped Sheet Metal Blanks, Intl Journal of Production Research (v33, 1995), pp1505-1520. 23. S. Joshi and M. Sudit, Procedures for Solving Single-Pass Strip Layout Problems, IIE Trans. (v26, 1994), pp27-37. 24. J. ORourke, Computational Geometry in C, 2nd ed. (Cambridge, UK: Cambridge Univ. Press, 1998).

25. J. Canny, The Complexity of Robot Motion Planning (Cambridge, MA: MIT Press, 1988). 26. Z. Li and V Milenkovic, Compaction and Separation Algorithms for . Non-Convex Polygons and Their Applications, European Journal of Operations Research (v84, 1995), pp539-561. 27. M. de Berg et al., Computational Geometry: Algorithms and Applications (Berlin: Springer, 1997).

Appendix
Derivation of Strip Pitch [Eq. (4)]

vF=(xF, yF)

Sweepline

vINT=(xINT, yINT)

vS=(xS, yS)

Figure A1 Intersection of Edge with Sweepline

Define the point of intersection between the current edge vSvF and the sweepline as vINT. The equation of the line vSvF is as follows: Ax + By + C = 0 where A = (yF B = (xS C = [(yF yS) xF) yS)xS + (xS xF)yS] (A1)

The equation of the line OvINT (where O is the origin) is as follows: tan( ) x y=0 (A2)

247

Journal of Manufacturing Systems Vol. 19/No. 4 2000

To solve for the intersection, use Cramers rule, as follows:


B 0 1 C = = A B A Btan() tan() 1 A yINT = tan() A C C

Using the identities 1 + tan2( ) = 1/cos( ) and tan( ) = sin( )/cos( ),


p= C Acos() + Bsin()

x INT

(A3)

( yS y F ) x S + ( x F x S ) yS ( yS yF )cos() + ( x F xS )sin()

0 Ctan() = B A Btan() tan() 1

x F yS x S y F = ( yS yF )cos() + ( x F xS )sin()

(A6)

(A4) Authors Biography


Tim J. Nye is an assistant professor in the Mechanical Engineering Dept. at McMaster University (Hamilton, Ontario, Canada). Dr. Nye received his BASc in mechanical engineering from the University of Waterloo, his MSc in industrial and systems engineering from Ohio State University, and his PhD in operations research from the Management Sciences Dept. at the University of Waterloo. His research interests include the application of optimization techniques to manufacturing operations, development of solid freeform fabrication processes for forging and casting, design of manufacturing systems, and examining stochastic, queuing-based lot-sizing models. Dr. Nye is currently a member of SME, NAMRI/SME, ASME, INFORMS, CORS, and PEO.

Strip pitch, p, is as follows:


2 2 p = x INT + yINT

(= Ov )
INT

C 1 + tan 2 () A + Btan()

(A5)

248

S-ar putea să vă placă și