Design for Manufacturability Forward or Reverse?

Transcription

1 Invited Keynote Paper 3 rd International and 24 th AIMTDR 2010, AU College of Engineering, Andhra University, Visakhapatnam, December 2010 Design for Manufacturability Forward or Reverse? Dr. B. Ravi, Professor Mechanical Engineering Department Indian Institute of Technology Bombay, Mumbai b.ravi@iitb.ac.in, Phone: Abstract The narrow focus of product engineers on functionality alone renders most parts difficult to manufacture. To achieve the desired quality, manufacturing engineers usually tweak the part design, thereby increasing its weight, cost, and lead time. This could be avoided if product engineers optimize the part design considering both functionality and manufacturability. While the methodology for functionality evaluation and improvement is well defined and suitable tools are available, the same is not true regarding manufacturability. As a result, we observe high rates of rejection, repair or additional machining, especially in metal cast parts, where there is large gap between manufacturing intent and feasibility. In this work, we present a 4-level system for manufacturability evaluation and improvement applicable to gravity-cast parts produced in sand and metal molds. The levels correspond to the design of part, tooling, methods and process parameters. Each subsequent level provides more information and feedback regarding part quality and cost to the designer, though requiring more time and domain knowledge than the previous level. A project data management system integrates the different tasks and enables collaboration between casting lifecycle engineers. A real-life industrial example is presented to illustrate the benefits of such a system, and to show how it can change the way castings are designed and manufactured in future. Keywords: Casting, CAD/CAM, Design for Manufacturability, Simulation, Quality, Cost. 1. Introduction Most parts are designed for manufacture, not for manufacturability. Product designers in OEM firms determine a suitable combination of part material composition, geometric features and manufacturing tolerances to achieve the desired functionality. They are usually not aware of the extent to which the designed features of the part affect quality and cost issues later (Fig. 1). The part design is passed to a tier-1 or tier-2 supply firm, where production engineers try to achieve the desired quality through appropriate design of tooling and process parameters, and minor changes to part design as well. In the case of cast parts, this includes applying draft to faces along draw direction, plugging drilled holes, increasing fillet radius, and padding thin walls. These changes increase the weight of as-cast parts by 10-15% compared to the original design. Machining the additional volume leads to an (unnecessary) increase in cost. Further, 8-10% of castings in jobbing foundries are rejected, recycled or repaired, implying further (avoidable) costs.

2 The abovementioned wastage of resources could be avoided by predicting the influence of part design on quality and cost, and modifying the design to achieve the desired manufacturability without compromising the required functionality. We refer to this as forward DFM to be practiced by OEM firms, in contrast with the reverse DFM that is currently practiced by component supply firms. Since forward DFM has to be carried out by product designers, they need to be equipped with the right knowledge and software tools. Fig.1: Effect of original and revised part design features on quality and cost In this paper, we review the current way of casting development and related quality issues to understand why there is very little practice of forward DFM in industry. Then we present our work in developing a comprehensive framework for casting DFM that connects part design, tooling design, methods design and process planning levels. It enables evaluation of product quality and cost at each level, and plows the information back to product engineers to facilitate DFM. Initial results generated for an industrial casting are presented to illustrate the working and benefits of the system. 2. Productivity versus Quality Increasing emphasis on core competence has led to a scenario where parts are designed in an OEM firm, tooling are developed in a tool room, castings are produced in a foundry, rough machined in a machine shop, inspected and heat treated by special units, then delivered to the OEM or a Tier-1 supplier for finish machining and assembly. Over time, each firm in the casting supply chain adopted new technologies and methodologies to maximize their resource utilization, continually reducing production cycle time and costs. This has yielded a significant jump in productivity. Today, an automobile firm creates new vehicle designs every year, that too, with ever smaller team sizes. A modern die casting unit churns out thousands of castings per day a fairly intricate casting (like an aluminum-alloys crank case) costs only a few hundred rupees.

3 The jump in productivity along with tougher inspection standards comes with a major side effect: high rejection rates, especially in the case of castings. If quality is defined as the conformance of ascast part with respect to the original design from the OEM within a band of tolerances as narrow as for machined parts, then virtually no casting will pass. Let us briefly review the quality issues related to three types of conformance: geometry, integrity and property (Fig. 2). Fig.2: Various conformance criteria for as-cast part with respect to designed part Geometry conformance is affected by the series of transformations from part to pattern, and further on to mold cavity, as-cast part, rough-machined part, and finally finish-machined part. The OEMs usually prepare the drawings or 3D models of only the finished part; the casting supplier is forced to recreate the model of the as-cast part and tooling. Various allowances (shrinkage, distortion, machining, etc.), draft, fillets, and plugging of drilled holes can add as much as 10-15% to the original part volume. Individual features, for example, a particular hole diameter or wall thickness may need to be drastically modified (as much as half or double the original value) to facilitate manufacturability. Integrity conformance can be defined as absence of defects. The defects can be characterized by: (a) location - internal or surface; (b) size - micro, meso or macro; (c) stage - casting or machining; (d) repairing - repairable or not repairable; and (e) criticality - low or high. A more useful classification scheme is in terms of the physical phenomenon involved: (i) mold cavity creation, (ii) metal melting & pouring, (iii) casting solidification, and (iv) fettling and grinding. Mold cavity creation includes molding and core-making up to their assembly, leading to defects like flash, mismatch and scab. Melting and pouring related defects include: incomplete filling (cold shut and misrun), gaseous inclusions (gas porosity and blow holes), and solid inclusions (sand and slag). Solidification related defects include shrinkage (cavity, porosity, corner, centerline and sink) (Fig.3), cooling stress (hot tear and distortion) and swell. Fettling and grinding can lead to defects like cracks and poor appearance. The defects are often diagnosed incorrectly. While it is not surprising that shrinkage porosity may be difficult to distinguish from gas porosity, identifying a shrinkage cavity as blow hole (quite common in industry) is not pardonable! Wrong diagnosis can lead to wrong treatment.

4 Fig.3: Shrinkage defects (left to right): cavity, porosity, centerline, corner and sink Property conformance is affected by alloy composition and manufacturing process. The related defects may be classified as: incorrect composition (ex. segregation), and inadequate properties (ex. poor tensile strength and hard spots). When the specifications are narrow (ex. smaller allowable range of composition or property values), or localized (ex. test bars to be taken from a specific section) then achieving a good conformance of properties between the cast part and designed part becomes even more difficult. In plastic molding and metal forming there is much better conformance between designed and molded parts. Part requirements (like wall thickness) are compatible with the process capabilities, resulting in less than 1% rejections. In contrast, a recent survey of about 200 foundries throughout India, representing all major metals, processes, capacities and applications showed that integrity defects alone account for an average 7.4% rejections (at casting plus machining stages). Knowing that some defective castings go un-detected or under-reported, and by including property defects, the average rejection rate is estimated to be 10-15%. Considering that most castings are hardly designed for manufacturability, this is actually good news. The credit for preventing a much higher rejection rate (that would otherwise be expected) goes to casting engineers. They tweak the part design, produce acceptable quality castings with the existing process (with all the limitations and constraints), and take the additional cost burden of machining the as-cast parts to bring them closer to the original design. Let us briefly review some of the key reverse DFM practices in industry. 3. Design for Manufacturability In Reverse Foundry engineers spend considerable time fixing casting designs to reduce shrinkage defects. Most of the alloys have a volumetric contraction of 3-7% from their liquidus to solidus temperature. Since the solidification of a casting progresses from its surface to interior, the last freezing region or hot spot compensates for the rest of the casting, and ends up with a shrinkage porosity defect. In a cubic casting of 100 mm side, the size of shrinkage defect, concentrated in a sphere, can be over 30 mm! To prevent shrinkage defects, foundry engineers manipulate the solidification of the casting from thin to thicker sections and finally toward feeder(s). Feeders are essentially late-solidifying reservoirs of liquid metal that compensate the volumetric contraction of the casting during solidification. This shifts the shrinkage defects to feeders, which are later fettled and recycled. The volume of feeders can be significant, reducing the yield by as much as 30-40%; fettling of feeders (by impact and grinding) can lead to cracks and poor appearance. Hence minimizing the number and volume of feeders is an important secondary objective.

5 For effective feeding, there must be controlled directional solidification from thin to thicker to thickest sections of a casting, and finally to feeders. This is achieved through the location, shape and size of feeders, as well as feeder connection to casting (neck), and feed-aids (like insulating or exothermic sleeves and covers. Fig. 4 shows simplified representations of three common configurations of section thickness encountered in metal castings. The first one shows the thickest region at one end, feeding the intermediate thickness region, which in turn feeds the thinnest section at the other end. Connecting a feeder to the thickest section thus takes care of feeding the entire casting, making this an ideal configuration. Fig.4: Section thickness configurations affect directional solidification and feeding The second configuration shows the thinnest section in the middle, leading to a major hot spot on one side and a minor hot spot on the other side. While the major hot spot can be fed by a feeder connected to the thickest section, this feed metal cannot reach the minor hot spot at the other end due to early solidification of intervening thin region. There are three foundry solutions to prevent the defect at the minor hot spot: (i) add another feeder for the minor hot spot, (ii) place a chill near the minor hot spot, or (iii) add an insulation padding around the intervening thin section. All three involve additional manufacturing cost. The defect can also be prevented by three part design changes: (i) reduce the thickness of the section containing the minor hot spot, (ii) add fins to the minor hot spot, to (iii) increase the section thickness of the intervening region. Foundries often resort to the last solution, followed by machining to bring the thickness back to original design. The third configuration shows the thickest section in the middle, which contains a single major hot spot that feeds both adjacent sections, and ends up with a big shrinkage defect. Assuming top feeding is disallowed due to some other constraint (say, top surface is curved and feeder will be difficult to fettle, or company name is molded on the top surface), there is no place to connect a side feeder to feed the hot spot. Foundries try to reduce such defects by connecting a larger feeder to the second thickest section, coupled with placing a chill at the bottom of the thickest section. These measures fail to eliminate the defect entirely, especially when the distance between the feeder and hot spot is large. Ideally, such problems could be visualized at the design stage itself, and appropriate changes made to part design, meeting the functional requirements in some other way.

6 The reason forward DFM is not practiced in industry is simple: predicting the effect of a design change on product quality (foresight) is far more difficult than shop-floor manufacturing trials and analysis (hindsight). Designers lack manufacturing knowledge, and manufacturing engineers lack design rights. It is no surprise that DFM is practiced in reverse. Late detection of defects that are traced back to poorly designed part features, resulting in avoidable costs and lead time for fixing such defects, are no longer acceptable. At the same time, intense pressure from competitors and customers is forcing OEM firms to share the responsibility for product quality and explore new avenues for cost reduction. Short-term price negotiations with suppliers are therefore giving way to long-term technological collaborations. Both sides are beginning to appreciate the importance of collaborative design and optimization of part, tooling and process. Unfortunately, there are no tools for early evaluation of the manufacturability of cast parts, suitable for product engineers. This gap is sought to be addressed by our R&D work. 4. Multi-Level Framework for Forward DFM The casting DFM framework comprises four levels corresponding to part design, tooling design, methods design and process planning (Fig.5). Compared to the current way of casting development practiced in industry, there are three distinguishing features: Fig.5: Multi-level framework for casting DFM with feedback loops

7 1. The flow of information from one level to subsequent level is seamless; the combined output of previous levels becomes the input to the subsequent level. 2. Each level has a local optimization loop, supported by appropriate tools, to determine the values of design parameters corresponding to that level that will achieve the best possible quality at the least possible cost. 3. There is an additional feedback loop at the end of each level, connecting back to part design, allowing part designers to better predict part quality and cost, and modify the part design to achieve the best overall quality and cost. Major design parameters at each level, that significantly affect casting quality and cost, are briefly mentioned here. Part Design Level: Parameters related to wall thickness, junctions and through holes can be linked to part quality; part volume and overall shape complexity can be linked to manufacturing cost. Cast metal/alloy composition and manufacturing tolerance affect the selection of process parameters and finally affect all conformances: geometry, integrity and property. Tooling Design Level: Major design parameters at this level include draw direction, parting line, undercuts, draft allowance, cored features and mold cavity layout. All these mainly affect the geometry conformance and tooling cost. Methods Design Level: This includes feeding and gating system design, which have a significant effect on integrity and property conformance, as well as yield and therefore manufacturing cost. Process Planning Level: Major control parameters at this level that affect casting quality include furnace charge mix, melt treatments, pouring temperature, pouring time, and cooling time, along with ambient conditions (temperature and humidity). Over the last ten years, we have developed and integrated different pieces of the above framework, utilizing the knowledge gained from research work and experience gained from industry interactions. Some portions of the framework are still under development, testing or improvement. To ensure that the entire system is of practical use, the following guidelines have been set forth: 1. The overall goal is to ensure that the castings are right first time (fewer trials), and right every time (consistent quality) at the least possible cost. 2. Part designers with very little process knowledge and casting engineers with only basic computer skills should be able to use the system. 3. Overall computation time should be minimized by automatic good-first design suggestions, reusing the data (inputs and outputs), and using efficient algorithms. A preliminary implementation of the proposed framework in a computer program is now available. At the part design level, thickness checks enable visualizing thin and thick sections, and locations with rapid thickness changes. These enable preliminary checks of part manufacturability, essentially by comparing part design parameters with corresponding process capability values. At the tooling design level, parting line, cores and mold cavity layout can be semi-automatically designed and

8 analyzed. The methods design level has been implemented in detail, allowing semi-automatic design and 3D modeling of feeding and gating system, followed by mold filling and casting solidification, to predict quality issues more accurately (compared to part design level). The cost model similarly gives more accurate estimations as the design progresses from part to process level. At the part level, only the direct metal cost is known. At the next stage, tooling cost can be estimated. After methods design, the energy costs can be added to the estimate. After process design, labour and overhead costs can be added to the total estimate. The working and benefits of the system are illustrated with an industrial example of a ductile iron gear case casting of overall size is 371 mm and weighing 5.76 kg (Fig. 6). Wall thickness analysis shows a rapid transition of thickness from about 35 mm to 4 mm within a short distance of about 40 mm, signifying a high thickness gradient that can possibly cause hot tears. This is also indicated by the part-process compatibility checks. The minimum thickness value (vertical black line) lies at the lower limit of process capability (green band); the minimum thickness should be increased if possible, to avoid unfilling defects. The major cored feature is semi-automatically identified by specifying the three openings (top, side and bottom). Fig.6: Part design checks: shape complexity, thickness, cored features and process compatibility Tooling design involves checking the part orientation, specifying the parting line, designing the cores, and optimizing the cavity layout (Fig. 7). The cores are automatically designed by converting the cored hole to a solid body, and adding appropriate prints (core supports) at the openings. The user can modify the diameter or length of the core print if needed. The combination of mold size and number of cavities is optimized considering the weight ratio of part metal to mold material. A ratio of 1:2 to 1:4 is considered good; a higher ratio (more cast metal) can lead to solidification problems, and a lower ratio (more mold material) lead to poor utilization of mold material.

9 Fig.7: Tooling design: parting line, cores, cavity layout and material ratio check Fig.8: Methods design (feeders and gating) and casting simulation

10 The methods design is verified by simulation of mold filling and casting solidification. In mold filling, the impact of melt jet on mold wall (or cores) is evaluated to ascertain the possibility of mold erosion and thereby sand inclusions (Fig. 8). The total filling time is evaluated against the ideal filling for the given cast metal, weight, wall thickness and pouring temperature (fluidity); if the simulated filling is too short, it indicates the possibility of turbulence-related defects; too long filling indicates the possibility of unfilling defects. In such a case, the gating design is semi-automatically optimized for Casting solidification simulation gives the temperature profile, gradient map and solidification time. These are useful to predict shrinkage related defects, based on which the feeding system design is optimized to achieve the desired quality at the highest possible yield. Finally, the casting costs are estimated in terms of tooling, direct metal, indirect materials (mold, core, feedaids, etc.), energy, labour and overheads (Fig. 9). This is very useful for comparing different layouts (combination of part, tooling and methods design) to select the most frugal one. Fig. 10 shows the actual tooling fabricated for this casting, and Fig. 11 shows the defects observed in actual casting, matching the predictions by simulation. The defects can be reduced by optimization of methods design, and fully eliminated by 4-level DFM as proposed in this work. The final solution is not shown here to protect the interest of the manufacturer. Fig.9: Tooling cost estimation, and comparison of all costs Fig.10: Actual tooling fabrication and casting with feeding and gating system

11 Fig.11: Casting defects as predicted, that could have been avoided by DFM The total time for design optimization is minimized by suggesting good-first design parameters at each level, thereby reducing the number of design iterations. The parameter values are obtained using analytical models available in literature, supported by empirical and phenomological models developed by our long association with the casting industry. These models are being continuously refined and improved based on industry feedback. The use of Gradient Vector Method for casting solidification simulation enabled reducing the computation time to an order of magnitude less (compared to other numerical methods), without compromising the accuracy of visualizing hot spots (temperature peaks), feed paths (maximum gradient lines), and shrinkage defects. Starting from the part model check, all the steps shown above, including methods design, casting simulation and cost estimation were completed within one hour. The need for very few inputs, coupled with the ease and speed of use, makes this technology eminently suitable for DFM applications by product engineers in OEM firms. 5. Summary and Conclusion A comprehensive framework for DFM of cast parts has been presented. It integrates the local optimization loops at part design, tooling design, methods design and process planning levels by seamless flow of design data from preceding levels to subsequent levels, and incorporates feedback loops from each level connecting back to part design. This facilitates global optimization of all design parameters to achieve the desired quality at the minimum cost. Thus the reverse DFM (looking back, or hindsight) that is practiced in industry is transformed into forward DFM (looking ahead, or foresight). Potential manufacturing issues can be predicted early in product lifecycle, and prevented by suitable changes to design parameters. The casting project data as well as materials and process databases are stored in XML format. Further, the project data is automatically compressed and archive. These measures facilitate webbased data exchange and collaboration. Eventually, the applications also can be moved to the web, enabling what is now referred to a cloud computing. Our work is throwing up many new challenges, which we hope will motivate the new generation researchers to come back and help the mother of all industries.

12 References 1. Durgesh Joshi and B. Ravi, Early Castability Evaluation using Analytical Hierarchy Process, International Journal of Advanced Manufacturing Technology, Feb 2010, DOI: /s Durgesh Joshi and B. Ravi, Classification and Simulation based Design of 3D Junctions in Castings, Transactions of American Foundry Society, 117, 2009, pp Milind M. Akarte and B. Ravi, Casting Product-Process-Producer Compatibility Evaluation and Improvement, International Journal of Production Research, 45(21), , K. Subburaj, Sandeep Patil and B. Ravi, Voxel-based Thickness Analysis of Intricate Objects, International Journal of CAD/CAM, 6(1), , R.G. Chougule and B. Ravi, Casting Cost Estimation in an Integrated Product and Process Design Environment, International Journal of Computer Integrated Manufacturing, 19(7), , R.G. Chougule and B. Ravi, Variant Process Planning of Castings Using AHP-based Nearest Neighbor Algorithm for Case Retrieval, International Journal of Production Research, 43(6), , R.G. Chougule, M.K. Jalan and B. Ravi, Casting Knowledge Management for Concurrent Casting Product Process Design, Transactions of the AFS, 112, , M.M. Akarte and B. Ravi, Casting Data Markup Language for Web-based Collaborative Engineering, Transactions of the AFS, 110, , B. Ravi, R.C. Creese and D. Ramesh, Design for Casting A New Paradigm to Prevent Potential Problems, Transactions of the AFS, 107, B. Ravi and M.N. Srinivasan, Features-Based Castability Evaluation, International Journal of Production Research, 33(12), , Author Profile: Dr. B. Ravi is a Professor of Mechanical Engineering at IIT Bombay, and Founder-Coordinator of OrthoCAD Network Research Centre. His areas of research include casting simulation, medical implant development, and product lifecycle engineering, in which he has guided seven Ph.D. and over 40 Master students. His work is published in 180 technical papers (including 40 in international journals), a book on Metal Casting: Computeraided design and analysis (PHI), and a chapter in the book Bio-Materials and Prototyping Applications in Medicine (Springer). Dr. Ravi is a sought-after speaker and consultant, and has assisted over 50 organizations ranging from small foundries to global MNCs. The casting software developed by his team has helped improve the quality and yield of several hundred industrial castings. He twice represented India at World Foundry Congresses: Dusseldorf in 1989, and Philadelphia in He is a Member of ASME Bio Manufacturing Committee and Fellow of the Institution of Engineers. He is on the review boards of several international journals in engineering, medicine and computer science, and has chaired several international and national conference sessions. He has won several awards, including the WTI Foundation award in 2003 for his draft manufacturing policy The Past, Present and Future of Indian Manufacturing Industry, given by former Prime Minister Mr. A.B. Vajpayee. Dr. Ravi holds an engineering degree from National Institute of Technology, Rourkela in 1986, followed by Masters and Ph.D. from Indian Institute of Science, Bangalore in He has held visiting appointments at Purdue University and West Virginia University in USA. He is currently establishing an e-foundry network for affordable casting simulation through public-private partnerships.