Automatic Process Planning and Toolpath Generation of a Multiaxis Hybrid Manufacturing System
Ruan, Jianzhong
Abstract
With the integration of multiaxis layered manufacturing and material removal (machining) processes, a hybrid system has more capability and flexibility to build complicated geometry with a single setup. Process planning to integrate the two different processes is a key issue. In this paper, an algorithm of adaptive slicing for a five-axis Laser Aided Manufacturing Process (LAMP) is summarized that can generate uniform and non-uniform thickness slices. A method to build a non-uniform (thickness) layer that utilizes two processes is presented, and an overall algorithm for integration is described. The newly developed algorithm implemented in the process planning helps the hybrid system build parts more efficiently.
Keywords: Layered Manufacturing, Hybrid System, Process Planning, Toolpath Generation
Introduction
Due to global market competition, manufacturing companies are pressured as never before to develop new products as product life cycles shorten. This trend can be seen in almost all manufacturing companies in the world. The science and technologies that can greatly reduce time to market will be critical for any company to be competitive in the 21st century. Since its appearance in the mid 1980s, Layered Manufacturing (LM) has given industry an approach to achieve the goal of providing better quality products in a shorter time and at a lower cost. This process quickly produces a part by depositing material on substrates, layer by layer, directly from a CAD model.
Recently, the focus of researchers has shifted to metal direct deposition systems to obtain fully functional parts. A high-powered laser is utilized in the process to melt metal powder, layer by layer, to form an expected geometry. Laser-engineered net shaping (LENS) (Keicher et al. 1998) and the directed light fabrication (DLF) system (Milewski et al. 1998) were developed at Sandia National Laboratory and Los Alamos National Laboratory, respectively. Mazumder et al. (1997) has also conducted research on metal-related forming systems. One of the primary disadvantages of a metal deposition system is that the accuracy and surface quality may not be adequate. The surface finish for a LENS fabricated part without additional processing is about 432 µ inches. Although LENS utilizes a laser-glazing technique to improve the surface finish to 74 µ inches, the processing speed is slow.
In a conventional 2.5-D laser deposition process, another major concern for the metal LM system is that support structures are needed to prevent an object from toppling over and to support material that would fall (Allen and Dutta 1995). Often, support materials for functional metal parts are not feasible. Moreover, deposition of the support material for metals leads to poor surface quality at the regions in contact with the support structure, and it increases the building time of the part and necessitates time-consuming post processing (machining or chemical process).
A machining (material removal) process can achieve a much better surface quality than the LM process; therefore, it can be used as a finishing operation to improve overall surface quality from the LM process. In a traditional 2.5-D metal LM system, the building direction is fixed with respect to the workpiece. If the building direction can be changed during the deposition process, some support structures can be eliminated by rotating the workpiece to a different building direction. It will be very helpful to integrate multiaxis deposition and machining processes to obtain the benefits of both processes.
The Laser Aided Manufacturing Processes (LAMP) Laboratory at the University of Missouri-Rolla (UMR) is developing an integrated system as discussed above. With a five-axis deposition process integrated with five-axis machining, some obstacles that occur in a traditional 2.5-D metal LM system can be removed. LAMP integrates the deposition process and the machining process together on a five-axis CNC machine. It includes two major systems: a laser deposition system (Rofin-Sinar 025) and a CNC milling machine system (Fadal VMC-3016L). The laser deposition system and the CNC milling machine work in shifts in a five-axis motion mode. The laser deposition system consists of a laser and a powder feeder. The working area is in a protective argon atmosphere to prevent oxidation. The net shape of the workpiece is formed by multiaxis deposition with layers of nonuniform or uniform thickness. Later, a five-axis CNC machining process is used to obtain an accurate profile of the part. With multiaxis capability, minimal support structures are needed, and the internal structures can be built with integration of the deposition process and machining process. Also, considerable lead time is saved and machining accuracy is achieved.
With an integrated system, the process becomes much more complicated than with regular systems. It is necessary to define motion code and sequence carefully for both processes in order to build parts efficiently. The major objectives of process planning for the LAMP system include: adaptive multiaxis slicing; defining positions of the nozzle and tool for deposition and machining processes, respectively; defining the sequence of two processes; and finding possible building solutions for a given part. The multiaxis adaptive slicing algorithm is discussed in another paper (Zhang and Liou 2001) and summarized here. The purpose of this paper is to discuss the integrated procedures for both processes.
Related Work
Previous work on process planning on the LM system involved little cooperation with the machining process. Most work only relied on 2.5-D adaptive slicing to improve surface quality. However, some systems and research have utilized the machining process in deposition to improve overall accuracy.
The Sanders prototyping machine uses an endmilling cutter to mill off the extra material deposited on each layer to maintain accuracy in the z direction. There is no surface machining to eliminate the step-case effect. The five-axis milling in shape deposition manufacturing (Merz 1994) is utilized to maintain the exterior contour. The layer thickness is uniform and the toolpath is generated for (NC) machining separately. Contour crafting uses an arrangement of trowels to shape the exterior contour of the layer, and material is filled in solid interiors layer by layer (Khoshnevis 1997). Currently, only planar trowels are used to form an approximation of the exterior of the layer.
Pinilla, Kao, and Prinz (1998) proposed a method to decompose a part into several manufacturable components, and every component is built up using the LM process. The machining process is carried on to improve the final part surfaces, which are identified in part decomposition, and the part is transferred between LM and machining workstations. However, no toolpath generation algorithms are discussed.
Kulkarni, Prasharnt, and Dutta (2000) presented a system to integrate layered manufacturing and material removal processes. Single and multipass machining toolpaths are used with the deposition layer with an adaptive thickness together to improve the efficiency and accuracy of the system. The machining toolpath is computed based on cusp height of the deposition layer. Also, a trowel tool is analyzed to sculpt the curve surface for the contour crafting process. This method is limited to 2.5-D problems, and the authors assumed that only stair-stepping caused from the LM process affected the final surface quality.
A five-axis machining process is used to improve the geometry accuracy and overall quality from the SWIFT process (Taylor et al. 2001). Two different algorithms are described for both situations-a simple one is designed for a single-edge contoured edge and a more complex one is presented for five-axis machining within a slice. No integration issues are discussed by the authors.
This paper presents a method that integrates the multiaxis-deposition LM process with the machining (material removal) process, which results in an automatic sequence for both processes and a toolpath for the deposition nozzle and machining tools. The switch between LM and machining is optimized to obtain fast and accurate procedures.
Process Planning Overview
Process planning, simulation, and toolpath generation for the LAMP allow the designer to visualize and perform the part fabrication from the desktop. LAMP process planning uses STL models as input and generates a description that specifies contents and sequences of operations. The objective of the process planning is to integrate the five-axis motion and deposition-machining hybrid processes. The jobs for process planning of the LAMP system involve multiaxis adaptive slicing, 2-D deposition toolpath generation, overhang and hollow structure process, integration between two processes handling non-uniform (thickness) layer building, and overall process sequence. The work presented in this paper is focused on non-uniform layer building and overall process sequence with toolpath generation.
Different from other adaptive slicing algorithms that produce uniform (thickness) layers, the adaptive slicing algorithm used in process planning can result in non-uniform thickness layers. As shown in Figure 1a, the thickness in layer p is consistent for every point on the layer; Figure 1b shows a non-uniform layer whose layer thickness changes with respect to points on the layer. Due to the difficulty of changing powder mass flow during one-layer deposition, it is not feasible to build an adaptive thickness layer with a pure deposition process; therefore, building a non-uniform layer requires cooperation between deposition and machining processes. The ability to build a non-uniform layer allows the LAMP system to build overhang parts without support structures.
The goal for process planning is not only to find a solution to build a part but also to look for an answer to produce it using the least amount of time; therefore, the least amount of switching between the machining process and deposition process the better because each switch requires retreating and relocating the deposition nozzle as well as the machining tool, which may cost extra time. The process planning analyzes the slicing results and compares them to the original geometry in STL format to find an optimal processes sequence. In this procedure, the toolpath for machining and deposition is also generated.
Adaptive Slicing
The details of multiaxis adaptive slicing are presented in another paper (Zhang and Liou 2001). The major algorithm is summarized here. Traditional adaptive slicing algorithms keep the building direction unchanged during the slicing procedure, and layer thickness is kept uniform (Singh and Dutta 2000; Dolenc and Mäkelä 1994). To achieve the benefits of the multiaxis system, the slicing algorithm developed for LAMP is conducted based on local geometry, and the building direction can be changed as needed.
In the multiaxis adaptive slicing process, the cutting planes may not be parallel to each other, as shown in Figure 1b. It can be seen that the building direction and thickness for the current layer are decided by the previous slicing layer. First, a cutting plane is generated based on the previous layer, which is called a slicing guided layer. The slicing step and initial building direction for slicing guided layers are decided by minimizing overhang value computed based on the previous slicing layer. Then, with the resulting slicing loops from the slicing guided layer, every optimal building direction for each small segment on the slicing loops is found. After mapping these directions on a Gaussian map as points, a search for a minimum circle on the map enclosing all these points is conducted. The center of the minimum circle shows the building direction for the current layer. The building direction and slicing step may be modified to fit the maximum layer thickness constraint for the system. The overhang problem is solved in the slicing process by introducing transition walls. The multiaxis slicing results in two different layers, a uniform layer and a non-uniform layer. Figure 2 shows a free-form part and its slicing result.
Integrated Process Planning
In this section, the issues regarding integration between layered manufacturing and machining in the process planning are discussed. Because the multiaxis slicing algorithm is adopted, non-uniform layers may be needed during deposition. One of the key technologies for process planning is to define the machine code to produce a non-uniform layer, which requires cooperation between LM and machining. Another function for process planning is to handle the integration between the machining and the deposition process, which involves optimal sequence defining, collision detection and processing, and toolpath generation for both processes.
Non-Uniform (Thickness) Layer Building
In the deposition process, each layer consists of two slices-an upper slice and a lower slice. The thickness of the traditional slicing layer is uniform; however, for one non-uniform layer the distance between the upper slice and lower slice varies from point to point. Because it is difficult to directly deposit a non-uniform layer, the building process is separated into two procedures: first, a uniform (thickness) layer is deposited; in the next step, the machining process is carried on to craft the shape of a non-uniform layer. The whole process is demonstrated in Figure 3. The process planning module defines the motion code for the deposition and machining processes, respectively. For a non-uniform layer, the only information of geometry is the upper slice and lower slice (Zhang and Liou 2001), which is not sufficient for generating a toolpath for depositing a uniform layer; therefore, the primary task is to generate the geometry of a uniform layer to be deposited. In this paper, this is also called the “nominal” layer or slice. The machining and deposition toolpaths are computed based on that geometry.
Upper Slice Generation (Geometry Generation)
A slice is the intersecting result from a slicing plane and STL model (Zhang and Liou 2001); therefore, a slice can be defined as S = {L^sub 1^, L^sub 2^, …, L^sub n^}, where L^sub i^ is an edge loop and a loop is defined as L = {p^sub 1^, p^sub 2^, …, p^sub n^}, where p^sub i^ is a vertex in the loop. Forming the nominal upper slice for a non-uniform layer is to find the corresponding loops with respect to those in the lower slice. With a nominal slice and lower slice of a non-uniform layer, a uniform layer is formed. From definition of nominal slice, it is clear that all the points on a nominal slice for a uniform layer have the same distance from the lower slice. If “nominal” points on the nominal slice can be found with respect to L^sub i^ in the lower slice, after applying the same method to other loops in the lower slice, all “nominal” points can be generated and the nominal upper slice is formed.
As shown in Figure 4, S^sub low^ represents a lower slice and S^sub upper^ represents an upper slice for a non-uniform layer. In the lower slice, loop L^sub i^ is one of the loops in S^sub low^, and it consists of a set of points, p^sub 1^, p^sub 2^, …, p^sub n^. The maximum and minimum distances between S^sub low^ and S^sub upper^ are defined as d^sub max^ and d^sub min^. N^sub lower^ is defined as the normal of S^sub low^. It is obvious that the “nominal” slice must have a minimum distance, d^sub max^, to the lower slice. If we can find all “nominal” points with distance d^sub max^ to S^sub low^, the “nominal” slice is formed.
There are two different situations for a point p on a slice from the STL model: the point is on an edge of two triangles-two-triangle case; the point is a vertex of a triangle-multi-triangle case, shown in Figure 5. For both cases, all the normals of triangles containing the point can be found. Instead of reconstructing the parametric form of the surface to search for the differentiable normal, the normal is found directly from triangles. Let N^sub 1^,N^sub 2^,…,N^sub n^ be the normals of the triangles containing the point p, N^sub p^ be the normal of point p, and α^sub 1^, α^sub 2^, …, α^sub n^ be the angle between N^sub p^ and N^sub 1^,N^sub 2^,…,N^sub n^. Assuming that the surface before discretion is C^sup 1^ continuous, the constraint arises to reflect the surface normal at point p.
Let α = max(α^sub 1^, α^sub 2^, …, α^sub n^), the constraint for N^sub p^ is that a minimum of α is achieved.
The normal N^sub p^ becomes (N^sub 1^ + N^sub 2^)/2 for the case shown in Figure 5a. It is easy to find a vector satisfying the above constraint for multi-triangle cases. The tangent vector T^sub pi^ can be defined by the edge in the loop associated with the vertex.
As shown in Figure 6a, because the points 1, 2 on the lower slice have different unit normals, different methods are applied to find the corresponding points on the nominal upper slice. After finding all nominal points and connecting them correspondingly, a nominal loop is formed on the nominal slice with a minimum distance to the lower slice. The same method is applied on the other loops on the lower slice. These newly generated loops form the nominal upper slice. Figure 6b shows the geometry of a nominal uniform layer after this operation, and the deposition result is shown in Figure 6a.
The start and end points can be decided by the intersection of the cutting plane and the outermost loop, shown in Figure 7. One singular pass includes fast motion and machining motion of the end mill. The orientation of the machining tool is the same as the normal of the upper slice. Actually, CC points are CL (cutting location) points.
Another constraint is geometry continuity. When geometry continuity is broken, the process is switched from deposition to machining. As shown in Figure 8b, after layer i is built the milling is conducted to machine the top surface of layer i and the side surface of previous layers. The aim for this operation includes: making some datum plane more accurate for the next deposition, and giving the system a break to check the deposition result. The detection of geometry continuity for an STL model is based on two continuous slices because the STL model cannot provide C^sup 1^ information. Two sets of triangles associated with two slices, denoted as ST^sub 1^, ST^sub 2^, can be determined by checking if the triangle contains any vertex in the loops of the slice. For each triangle T^sub i^ in ST^sub 1^, triangles adjacent to it in ST^sub 2^ can be found, and those triangles form another set, denoted as AT^sub i^. The angle between each triangle in AT^sub i^ and T^sub i^ is computed. If an angle is greater than a predefined value that is relative to STL accuracy, the geometry continuity break is detected. Moreover, if AT^sub i^ is empty for a triangle T^sub i^, the geometry continuity break is also detected.
As stated before, the most important goal of the LAMP system is to build parts in the least amount of time. In this hybrid process, although the machining and deposition processes are integrated on one workstation, each switch between two processes will involve loading/unloading the deposition nozzle and machining tool, which takes 30 seconds. Therefore, T^sub c^ is a key factor in determining total build time. Based on the discussion above, the LAMP process planning optimizes the process by minimizing the switching between two different processes and usage of support structures. In other words, once a process is conducted, it will not stop until one of the following conditions occurs.
* One of the constraints (collision or geometry discontinuity) appears.
* 3-D layer deposition is required.
* The process is finished.
The usage of support structures can be examined by selecting base planes of a part.
Machining and Deposition Toolpath Generation
The nature of process planning for LAMP requires automatic generation of the machining toolpath without user interaction. Most of the machining toolpath generation is surface toolpath generation. As mentioned before, STL format geometry is taken as input for the process planning. The strategy and algorithm for generating a surface machining toolpath is based on plane (polygon) classification obtained from the STL model. The surfaces are classified into three different types: convex, concave, and free-form (Ruan and Liou 2003). For each type of plane, a different strategy is used to find the machining toolpath, including machining tool orientation, tool approach direction, toolpath pattern selection, and CL data determination. The generation of a deposition toolpath is a 2-D problem and can be treated the same as the machining process. The method to find the optimal machining toolpath is discussed in Arkin, Held, and Smith (1996). The orientation of the deposition nozzle is always perpendicular to the slice. More details are discussed in Eiamsa-ard, Zhang, and Liou (2001).
Overall Algorithm
The overall integrated process is a slice-based procedure, and motion codes for both processes are generated in proper sequence. The constraints discussed above are basic criteria used to decide whether to switch from the deposition process to the machining process. The whole algorithm is summarized below:
Example
The part building process is demonstrated by the examples discussed as follows. Figure 9a shows a nozzle with a small through hole of diameter 0.5 mm. To accomplish the hole during the entire building process, the drill process is carried out for each layer building, and surface milling is done at the same time. The toolpath is shown in Figure 9b. Figure 10a shows the STL model with two holes. After the model is input into the process planning, the upper hole is filled as shown in Figure 10b. Figure 10c shows the toolpath from process planning, including deposition toolpath and machining toolpath. The processes sequence is defined correctly automatically. For this example, if surface A is chosen as base plane, the overhang branch still can be built without support structures; however, the hole on the base has to be “hidden” from the deposition process, which increases support structures usage. Figure 11 illustrates a 3-D layer-building process. In Figure 11b, a deposition result is shown. The STL model and process planning results are shown in Figure 11a. The side milling and top milling processes are demonstrated in Figures 11c and 11d, respectively. The final part with a total degree of 45 after conducting final machining process is shown in Figure 11e. Compared to regular 2.5-D rapid systems, the LAMP system saves about 80% of the support structures, as shown in Figure 11.
Conclusion
With integration of multiaxis deposition and machining processes on the same workstation, the LAMP system is able to produce complicated geometry, especially the overhang structure with less or no support structures, which saves material cost. The switch between the machining and deposition process is optimized to save operation time/cost. Based on different geometry shapes, usually the LAMP system can save up to 50%~60% of support structures. The surface quality of the final product from LAMP is 4 µ inches.
These discussed methods are used to integrate multiaxis layered manufacturing and the material removal process to improve manufacturing efficiency and accuracy. A multiaxis slicing algorithm producing optimal uniform and non-uniform layers is summarized. In this paper, the non-uniform layer building process and machining toolpath generation are discussed. The overall integration between two processes is described. With these methods, a part can be built with less support structures and effort.
Acknowledgments
This research was supported by National Science Foundation grant number DMI-9871185 and U.S. Air Force Research Laboratory contract #FA8650-04-C-704. This support is greatly appreciated.
References
Allen, Seth and Dutta, Deba (1995). “Determination and evaluation of support structures in layered manufacturing.” Journal of Design and Mfg. (v5), pp153-162.
Arkin, E.M.; Held, M.; and Smith, C.L. (1996). “Optimization problems related to zigzag pocket machining.” 7th Annual ACM-SIAM Symp. on Discrete Algorithms (SODA’96), Atlanta.
Dolenc, A. and Mäkelä, I. (1994). “Slicing procedures for layered manufacturing techniques.” Computer-Aided Design (v26, n2), pp119-126.
Eiamsa-ard, Kunnayut; Zhang, Jun; and Liou, F.W. (2001). “Skeleton-based geometry decomposition for adaptive slicing in a five-axis laser aided manufacturing process system.” Proc. of 12th Annual Solid Freeform Fabrication Symp., Austin, TX, Aug. 6-8, 2001.
Keicher, D.M.; Miller, W.D.; Smugeresky, J.E.; and Romero, J.A. (1998). “Laser engineering net shaping (LENS(TM)): beyond rapid prototyping to direct fabrication.” http://www.highfiber.com/~dmiller/Sanant.htm.
Khoshnevis, B. (1997). “Contour crafting: a new rapid prototyping process.” Proc. of 7th Int’l Conf. on Rapid Prototyping. San Francisco, pp13-22.
Kulkarni: Prasharnt; and Dutta, Debasish (2000). “On the integration of layered manufacturing and material removal processes.” Trans. of the ASME (v122. Feb. 2000), pp100-108.
Mazumder, J.; Choi, J.; Nagarathnam, K.; Koch, J.; and Hetzner, D. (1997). “The direct metal deposition of H13 tool steel for 3-D components.” Journal of Material (May 1997). pp55-60.
Merz, R. (1994). “Shape deposition manufacturing.” PhD thesis. Pittsburgh: Carnegie Mellon Univ.
Milewski, John O.; Thoma, Dan J.; Fonseca, Joe C.; and Lewis, Gary K. (1998). “Development of a near net shape processing method for rhenium using directed light fabrication.” Material and Mfg. Processes (v13, n5). pp719-730.
Pinilla, J.M.; Kao, J.; and. Prinz, F.B. (1998). “Process planning and automation for additive-subtractive solid freeform fabrication.” Proc. of Solid Freeform Fabrication Symp., Austin, TX, pp245-258.
Ruan, J.Z.; Zhang, J.; Liou, F.W. (2001). “Support structures extraction for hybrid layered manufacturing.” Proc. of DETC’01. Pittsburgh, Sept. 9-12, 2001.
Ruan, J.Z. and Liou, F.W. (2003). “Automatic toolpath generation for 5-axis surface machining.” DETC’03. ASME 2003 Design Engg. Technical Conf. and Computers and Information in Engg. Conf.
Singh, P. and Dutta, D. (2000). “Multi-direction slicing for layered manufacturing.” Proc. of DETC’00. ASME 2000 Design Engg. Technical Conf. and Computers and Information in Engg. Conf., Baltimore, MD.
Taylor, James B.; Cormier, Denis R.; Joshi, Sandesh; and Venkataraman, Vivek (2001). “Contoured edge slice generation in rapid prototyping via 5-axis machining.” Robotics and Computer Integrated Mfg. (v17). pp13-18.
Zhang, J. and Liou, F.W. (2001). “Adaptive slicing for a five-axis laser aided manufacturing process.” Proc. of DETC’01, Pittsburgh, Sept. 9-12, 2001.
Jianzhong Ruan, Kunnayut Eiamsa-ard, and F.W. Liou, Dept. of Mechanical & Aerospace Engineering, University of Missouri-Rolla, Rolla, Missouri, USA
Authors’ Biographies
Jianzhong Ruan is currently a post-doctoral researcher in the Dept. of Mechanical & Aerospace Engineering at the University of Missouri-Rolla. He received his bachelor of science degree in 1992 from Shanghai University of Technology. Shanghai, China (now Shanghai University), and MS in 1998 from Zhejiang University, Hangzhou, China, both in mechanical engineering. He received his PhD in 2003 in mechanical engineering from the University of Missouri-Rolla. His research interests include CAD/CAM, rapid prototyping, manufacturing automation, and material modeling.
Kunnayut Eiamsa-ard is currently a PhD candidate in the Dept. of Mechanical & Aerospace Engineering at the University of MissouriRolla. He received his BEng in mechanical engineering from Kasetsart University, Bangkok, Thailand. He also received MS degrees in industrial engineering and mechanical engineering from the University of Pittsburgh and Carnegie Mellon University, respectively. His current research interests are related to layered manufacturing and path planning.
Dr. Frank Liou is a professor in the Dept. of Mechanical & Aerospace Engineering at the University of Missouri-Rolla. He also serves as director of the Interdisciplinary Manufacturing Engineering Program at UMR. As an active researcher, he is a senior research investigator at the Intelligent Systems Center at UMR. His teaching and research interests include CAD/CAM, rapid prototyping, rapid manufacturing, and manufacturing automation.
Copyright Society of Manufacturing Engineers 2005
Provided by ProQuest Information and Learning Company. All rights Reserved