WHITE PAPER COST ESTIMATION OF COMPOSITE PRODUCTION DURING EARLY DESIGN PHASE

Transcription

1 WHITE PAPER Sirris Leuven-Gent Composites Application Lab COST ESTIMATION OF COMPOSITE PRODUCTION DURING EARLY DESIGN PHASE The use of lightweight components stands or falls by the choice of materials. The product value, product costs, production costs, development costs and risks are, however, difficult to estimate when talking about less well-known materials such as composites. Having a clear picture of production costs is important. This is why the SLC-Lab, as well as its partners in the CompositeBoost project, want to pass on the essential tools and methodologies to help designers and OEMs make the right choices. This sixth white paper focuses on the early design phase, as the greater part of the product cost is committed in this stage. The earlier design decisions are made in the concept phase, the greater the impact of these decisions on production cost. Generally, 70-80% of the cost of a product is committed during the early stages of product development. Having the cost commitment curve in mind, meaningful cost estimates are crucial and should be performed as early as possible % of cost Cost Design phase Commited cost Incurred cost Production phase Cost Commitment Curve, adopted from [1] Scope for reduction of production cost Time The development of composite components offers a wide range of material/process combinations, different automation levels and (typically) higher uncertainties than the development of their metal counterparts. In this white paper, a short classification of cost estimation techniques is provided, followed by cost models on three different levels. The models are illustrated in a case study: a bio-based composite table made in RTM-light. 1

2 COST ESTIMATION TECHNIQUES A series of cost estimation techniques was described by Niazi et al. [2]. Qualitative techniques are based on a comparison between already existing products/processes combinations. The past experiences and calculations support the calculation of new products and reduce the difference between planned and actual costs (variance). The qualitative technique can be further subdivided into an intuitive and an analogous technique, both having their advantages and limitations. Quantitative techniques describe the process chain of a product. After the derivation of cost drivers, mathematical models are generated for cost estimation, subdivided into parametric or analytical techniques. A slightly different nomenclature is used by Dan Galorath, founder of a software company, which develops high-fidelity cost estimation software. He introduced and presented eight cost estimation categories [3]. 2

3 MODEL CATEGORY DESCRIPTION ADVANTAGES LIMITATIONS Guessing Off-the-cuff estimates Quick No basis or substantiation. No Process. Usually wrong. Analogy Expert judgment Top-down estimation Bottom-up estimation Design to cost Simple Cost estimation relationships (CERs) Comprehensive parametric models Compare project with similar projects from the past. Consult one or more experts. A hierarchical decomposition of the system into progressively smaller components. Divides the problem into the lowest items. Estimate each item... sum the parts. Uses expert judgment to determine how much functionality can be provided for given budget. Equation with one or more unknowns that provides cost estimates. Perform overall estimate using design parameters and mathematical algorithms. Estimates are based on actual experience. Little or no historical data is needed; good for new or unique projects. Provides an estimate linked to requirements. Complete work breakdown structures (WBS) can be estimated. Estimate always lower than given budget. Cost based on actual data. Models are usually fast and easy to use, and useful early in a program; they are also objective and repeatable. Truly similar projects must exist. Experts tend to be biased; knowledge level is sometimes questionable; may not be consistent. Needs valid requirements. Engineering bias may lead to underestimation. Costs can occur in items that are not considered in the WBS. Little or no engineering basis. Only valid in combination with another model. Simple relationships may not tell the whole story. Historical data may not tell the whole story. Models can be inaccurate if not properly calibrated and validated; historical data may not be relevant to new programs; optimism in parameters may lead to underestimation. In the following, cost models of the two last categories are used and illustrated by the case study: A simple CER model based on the work of Esawi and Ashby [4] A more elaborate CER model which incorporates different process steps for liquid composite moulding and post-processing. This model uses complexity coefficients to accommodate the geometrical complexity of the part and can estimate the tooling cost based on size, complexity, process and material. A high-fidelity parametric model, developed and commercially distributed by Galorath Incorporated. 3

4 Lina case In order to compare these different cost models, the Lina case is introduced. Lina is a small composite table developed in the NIB project Generation Composite. It consists of a non-woven flax, infused with unsaturated polyester resin in an RTM-light process. The overall dimensions are 60 x 70 x 50 cm. Common, estimated values from the prototyping phase were: Mass m = 2 kg Material cost C m = 5 /kg Scrap rate f = 30% Tooling cost C t = 8000 Production quantity n = 5000 Maximum number of parts n t = 1000 Processing time 45 minutes Equipment cost C c = Load factor L = 60% Depreciation time two = 5 years = h Labour and overhead C oh = 50 /h 4

5 FROM LOW TO HIGH-FIDELITY COST MODELS A simplified approach for process cost modelling was developed by Esawi and Ashby [4]. In the paper Cost estimates to guide preselection of processes they calculate the manufacturing cost of a process step by the sum where the material contribution consists of mass m, material cost C m [ /kg] and a scrap rate f, the tooling contribution consists of the tooling cost C t [ ], the production quantity n, the maximum parts per tooling n t, and capital overhead & production rates, the latter containing the production rate n [per h], equipment cost C c, a load factor L [-], depreciation time two [h] and overhead C oh [ /h]. 5

6 Lina case The cost for preforming and infusion of the Lina table in RTM-light is given by The above model could be further refi ned by adding a setup time to the production of each batch, hence defi ning the production rate as the inverse of the production time, which in turn is the sum of time for processing the component and a share of the setup time. With a setup time of 8 hours per individual batch of 1,000 manufactured components, the manufacturing cost slightly increases to: At this early stage of product design, the estimated process time is 45 minutes. In production, the process time could vary between 30 to 60 minutes. Processing time 30 min 36 min 42 min 48 min 54 min 60 min Estimated cost Note that for this type of processing, material cost C m should not only include (composite) raw material. What about gel coats, release agents, consumables such as tubing, buckets, valves, etc.? These types of input are traditionally not included in this simple cost estimation model unless further developed and tailored to specifi c processes. The processing time, particularly for the preparation of the preform, is highly dependent on the topology and the complexity of the component. As shown in the Lina case, the cost estimate may not be suffi ciently accurate at this stage. 6

7 Adding complexity One way to narrow down the process time is adding the product s complexity to the cost model. This was done in the European project Design for Manufacture of Composites (DeMaCo). There, a cost estimator was built for liquid composite moulding processes, such as vacuum resisted resin infusion (VARI), RTM (resin transfer moulding) and RTM-light [5]. The fi rst step was to separate cost shares for material, (manual) preforming, injection and post-processing. In an attempt to add complexity coeffi cients to the preforming cost, an example database was built that linked typical surface complexities to knockdown factors in labour time. Some examples, which range from single curvature/shallow components to double curved components with deep geometric features are shown below. 7

8 The cost model should be simple enough, so that non-experts in composite manufacture are able to use it. Firstly, the processes were divided into process steps and cost drivers were identifi ed. The preforming process chain, for example, consists of steps for pre-cutting the reinforcement, application of a gel coat, preforming and end-cutting of the fi bre package. The mainly manual work is facilitated by the use of scissors and electric cutters, tools and jigs and consumables, such as adhesives, tapes and spray adhesives, whereas the main cost drivers are the number of reinforcement layers, the area, the reinforcement s perimeter and the complexity of the component. Preforming process chain Pre-Cut Gel-Coat Preforming End-Cut Production facility matrix Manual Cutter Manual Hand Lay-up Manual Cutter Electric Cutter Electric Cutter Non-value creating activities Handling Handling Handling Handling Ressources / Investments Scissor Electric- Cutter Gloves Mask Release- Agent Gel-Coat Brushes Gloves Mask Cup Textile Tooling Spray glue Adhesive- Tapes Inserts Scissor Electric- Cutter Gloves Mask Technology chain for preform production 8

9 The collected feedback and experiments within the project served as a basis of the quantitative cost model. Each process step was divided into product, project and investment costs. These individual cost shares were defined as follows: Product costs include all resources that are required at a component level, such as textile reinforcement layers, gel coat, sealant tape, release agent, etc. Project costs include all manufacturing resources at a project level, such as tools and jigs. This model consists of the cost of the material (steel, aluminium, tooling blocks) and the milling costs. If VARI or RTM-light is selected, the cost for tooling includes the milling of a mock-up and the lamination of mould and counter mould. Investment costs include equipment and machines that are depreciated across several projects over a duration of typically five years, such as injection equipment, presses, curing ovens, milling machine, etc. 9

10 Finally, generic data was collected for material cost, process speed (average cutting speed, draping speed, application of gel coats,...), and equipment (typical cost of curing ovens, heat pumps, injection pumps, milling machines,...). Specific material properties or equipment can be entered manually by the user. The tool was developed in Visual Basic for Applications and Microsoft Excel and validated using three demonstrator cases. The figures below show a screenshot of the injection cost module. Injection cost Selected Injection Process? VARI RTM-light RTM Product cost Project cost Resin system:? Polyester resin (UP)_ Primary tooling material:? Aluminium Resin density in g/cm³: 1.2 Surface Complexity (f c ): Resin cost in /kg: 5 edge complexity (f g ): Labour rate in /h: 50 Clear Data for Project Costs Clear Data for Product Costs Calculate Product Costs Unit cost Tooling cost in /unit: Unit cost Mould cost in /unit: Setup cost in /unit: 1.56 Mock-Up cost in /unit: Resin cost in /unit: 1.3 Total project cost in /unit: Labour cost in /unit: Total product cost in /unit: Injection Process Time in min: ?? Calculate Project Costs Machine Machine Machine Machine Machine Machine Machine Machine Clear D ess? VARI RTM-light RTM <- Previous Step: Preforming Calcula Project cost Primary tooling material: Surface Complexity (f c ): edge complexity (f g ):?? Aluminium Investment cost Include in Calculation Do not incude Investment Quantity Name: cost in : per hour: Machine 1 Vacuum System_ Machine 2 Mold Heating_ ? Utility Factor in %: Depreciation Time [years]: 5 5 Overhead in /h: 70 50????? Clear Data for Project Costs Calculate Project Costs Machine 3 sts Unit cost Tooling cost in /unit: 0.74 Machine 4 Machine 5 Mould cost in /unit: 0.06 Machine 6 Mock-Up cost in /unit: 0.09 Machine 7 Total project cost in /unit: 0.89 Machine 8 Clear Data for Investment Costs Calculate Investment Costs Result Investment cost in /unit: 0 Total Injection cost in /unit: Total Injection cost of the lot in : Calculate total Injection Costs Show expert settings Next Step: Finishing -> 10

11 Lina case Applying the Lina case to the DeMaCo cost model, a process time of 37 minutes was estimated, leading to a production cost of per unit. This cost is very close to the estimate of the simplifi ed model above, with the benefi t of refi ned input data regarding shape complexity. In addition, the graphic user interface allows for a fi rst analysis of the main cost drivers of the current design. Total unit cost <- Previous Step: Finishing Calculate total costs Total cost in /unit Total cost of production lot in Total Time/ unit Preforming Injection Finishing Process Time in min 18,65 11,7 6,44 Product Costs 33,14 12,61 6,44 Project Costs 0,89 Investment Costs 0,00 0,00 2,23 Total 33,14 13,50 8,67 Preforming Injection 24% 16% Total 60% Finishing Process time in minutes Preforming Injection Finishing 11

12 Bringing cost estimation to a higher level A much higher level of reliability is afforded by commercially available software. SEER for Manufacturing was developed in the 1990s by Galorath Incorporated. It contains parametric models that enable engineers to model and test manufacturing processes and trade-offs in an early design stage when little detail is known [6]. The models do not only contain composite processes, but a wide range of manufacturing processes, such as: Machining of metals, including milling, turning, drilling, EDM, etc. Fabrication, including spin forming and roll bending Tube fabrication, welding and processing Mould/cast/forge/powdered metals, including injection moulding, thermoforming, casting and sintering Assembly, including fasteners, welding, brazing and adhesive bonding PC boards and electrical assembly Finish and heat treatments, including wet chemical coating, powder coating, etc. The composite modules are divided as follows: Composites o Lay-up o Filament Winding o Pultrusion o Composite Spray Detailed composites o Hand Lay-up o Automated Tape Lay-up o Resin Transfer Moulding o Liquid Resin Infusion o Resin Film Infusion o Use of ovens and/or autoclave o Finishing, trimming, and non-destructive testing 12

13 The parametric models are built from sets of mathematical equations. All these models were made using actual, industrial data. Some of the models, for example the Detailed composites module, were written in collaboration with production companies (in this case Airbus UK) and rely on underlying kinematic process models. The mathematical relationships between input and output data are referred to as Cost Estimation Relationships (CERs). Unlike the simplified CERs of the cost models presented previously, much more input data is possible (and required) in SEER-MFG. Modelling is done using a graphical user interface, in which the work breakdown structure, input, output reports and charts are displayed. Graphical user interface of SEER-MFG (courtesy of Galorath Inc.) 13

14 Lina case The user interface of SEER-MFG is intuitive and straightforward to learn. With some prior experience in using the software, the production cost of a case as shown here can be modelled from scratch in less than half an hour. The output of SEER-MFG is given in terms of labour cost (including shares for set-up and recurrent cost), material cost and tooling. With an estimated labour of 43 minutes and a total cost of 56.19, this estimate is again very close to the foregoing results. Minutes/ Unit Cost/ Unit Cost for 5,000 Units LABOUR TOTAL , Set-up labour Direct labour , ADDITIONAL COST , Material Tooling TOTAL COST ,

15 CONCLUSION A classification of cost estimation techniques was shown, along with their advantages and disadvantages. Subsequently, cost models on three different levels were introduced which allow for the estimation of composite manufacturing cost at an early design stage. The models were illustrated using a case study: a bio-based composite table made in RTM-light. The advantage of the first model was its simplicity and the option to tailor it to a wide range of manufacturing processes. Its main limitation was the difficulty to correctly estimate the process time. The second model, developed specifically for liquid composite moulding processes, incorporated the time for preforming and infusion by the application of complexity factors in combination with geometrical input. In addition, the mould cost could be estimated based on complexity factors and historical data. The parametric model of SEER-MFG included sophisticated mathematical models. The software shows a great level of detail, ranging from tooling, process setup, processing to post-processing steps. It was shown that the difference between the highest and the lowest estimates was less than 16%, which is a good result in this phase of product development. Which level of reliability is necessary to estimate the production cost at an early design phase? As much as required, but as little as possible. The answer is not straightforward and depends on the specific need of the design team, the complexity of the product and the number and type of processes involved. In addition, the lack of information in this phase can challenge the use of high fidelity models, particularly for novel product developments outside the company s field of expertise. 15

16 AUTHOR Markus Kaufmann, PhD (Sirris) SOURCES [1] Ullman D, The Mechanical Design Process, 4th ed., McGraw-Hill, New York, [2] Niazi, A et al. Product Cost Estimation: Technique Classification and Methodology Review, Journal of Manufacturing Science and Engineering, 126, pp , [3] Galorath D: Why Can t People Estimate: Estimation Bias and Strategic Mis- Estimation, Webinar 2014 [4] Esawi and Ashby: Cost estimates to guide pre-selection of processes; Materials & Design, pp , 2003 [5] Kaufmann M et al: Design for Manufacture of Composites (DeMaCo), Final Report, [6] SEER by Galorath Inc. 16

17 This white paper was written within the scope of the project CompositeBoost. COMPOSITEBOOST CompositeBoost is a collaborative project involving the Sirris SLC-Lab, UGent and KU Leuven. Based on six highly relevant issues, the project partners want to use these essential tools and methodologies to allow designers and OEMs to make the right choices. Masterclasses, demonstrations and exploratory case studies will help transform the composites processor into a reliable production company and partner. This means that our companies will retain their competitiveness over foreign competitors. Markus Kaufmann, PhD (Sirris) is the program manager of the composites division of Sirris and has been head of the SLC-Lab since April Before that he acquired experience in design and cost estimation of composite structures. Markus is responsible for coordinating CompositeBoost. Linde De Vriese (Sirris) is the SLC-Lab s team member specialising in material characterisation, press forming of thermoplastic composites, and bio-composites. Linde received her Master s in Materials Engineering at KU Leuven in 2010, specialising in polymers and composite materials. 17

18 Bart Waeyenbergh (Sirris) works at SLC-Lab on prototyping, product development, mould design and processing of thermoset composites. He graduated with a Master s in industrial sciences at Group T in Leuven in 2008, specialising in advanced manufacturing. Tom Martens (Sirris) is the senior technician at SLC-Lab, with 18 years of experience in the plastics industry. He is responsible for the production of prototypes and demonstrators in both thermoplastics and thermoset composites. Katleen Vallons, PhD (KU Leuven) is a post-doctoral researcher at SLC-Lab. She has worked with the Composite Materials Group at KU Leuven since 2005, mostly on projects in collaboration with industrial partners. Her expertise is in the material behaviour of composites. Geert Luyckx, PhD (UGent) is a post-doctoral researcher at SLC-Lab. He has worked at UGent, Mechanics of Materials and Structures, since 2003 and is involved in the experimental validation of new optical measuring technologies for measuring shape distortion in composite components. 18

19 PARTNERS SIRRIS LEUVEN-GENT COMPOSITES APPLICATION LAB Celestijnenlaan 300C 3001 Heverlee blog.sirris.be Sirris Leuven-Gent Composites Application Lab