- Email: [email protected]
Cost Modelling and Sensitivity Analysis of Wire and Arc Additive Manufacturing
Available online at www.sciencedirect.com
ScienceDirect Procedia Manufacturing 11 (2017) 650 – 657
27th International Conference on Flexible Automation and Intelligent Manufacturing, FAIM2017, 27-30 June 2017, Modena, Italy
Cost modelling and sensitivity analysis of wire and arc additive manufacturing C. R. Cunningham*, S. Wikshåland, F.Xu, N. Kemakolam, A.Shokrani, V. Dhokia, and S.T. Newman Department of Mechnical Engineering, University of Bath, Bath, BA2 7AY, UK
Abstract With the proliferation and diversification of metal additive manufacturing (AM) processes in recent years, effective decision tools for process selection are of increasing importance. This paper presents a novel time activity based cost model for Wire-Arc Additive Manufacturing (WAAM), an emergent metal AM technology. The full process chain is modelled and a tool-path based deposition cost employed. The results show that WAAM has significant potential as a cost-effective manufacturing approach compared to other AM and conventional CNC machining methods. Sensitivity analysis identifies indirect costs as a key cost driver generally, with parts per build plate, and shielding cost having significance for smaller and larger parts respectively. © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license © 2017 The Authors. Published by Elsevier B.V. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-reviewunder underresponsibility responsibility of the scientific committee the International 27th International Conference on Flexible Automation and Peer-review of the scientific committee of theof27th Conference on Flexible Automation and IntelligentManufacturing Manufacturing. Intelligent Keywords: Wire and Arc Additive Manufacturing; Cost Modelling; WAAM; Additive Manufacturing
1. Introduction Additive Manufacturing (AM) has become the subject of a substantial amount of research due to promising industrial potential, with AM products and services set for sales of $26 billion worldwide by 2021[1]. The concept of creating metal products additively has seen an increased uptake in recent years. AM provides opportunities for
* Corresponding author. Tel.: +44-1225-386934; fax: +44-1225-386928. E-mail address: [email protected]
2351-9789 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the 27th International Conference on Flexible Automation and Intelligent Manufacturing doi:10.1016/j.promfg.2017.07.163
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
reduced weight and cost savings and has been especially prevalent in the aerospace industry where there have been instances of conventional manufacturing processes being replaced by metal AM processes. Processing limitations have led to the diversification of metal additive manufacturing techniques especially within the categories of Powder Bed Fusion (PBF), Binder Jetting and Direct Energy Deposition (DED) [2], each with their distinct benefits. Understanding the performance of metal AM processes in terms of cost effectiveness is essential for effective deployment and further uptake of AM. Cost effectiveness is especially important as a measure because industrial standards of quality control are rarely achieved in metal AM. The combination of knowledge of process capability and effective cost modelling can provide a valuable insight into the potential cost escalation or “real cost” of a particular AM process. Wire and Arc Additive Manufacturing (WAAM), is a DED approach that is seeing considerable interest due to its ability to create large metal products faster than powder-based alternatives, with deposition rates up to 4 kg/hr [3]. The process uses an electrical arc as a heat-source to melt feedstock metal wire, which is deposited onto a base plate [4]. The WAAM process comprises successive cycles of melting, depositing and cooling to result in a near-net shape deposit. Cold-work through rolling may be subsequently employed to improve microstructural properties and relieve residual stresses induced by thermal cycles incurred through the deposition process [5][6]. WAAM with an integrated roller consists of an arc torch is shown in Figure 1 [5]. To achieve end-product status, heat treatment for residual stress relief and/or to develop mechanical properties and finish machining is required [8]. Compared to powder based AM methods, WAAM offers several distinct benefits. In certain materials, directed deposition and cheaper wire feedstock mean that cost be reduced significantly compared to PBF. Further, processing issues such as powder agglomeration, recycling are overcome. Concerns with development of full density and microstructure of non-heat treatable materials in PBF and Binder Jetting may be overcome by the inter-pass rolling which is easily implemented in WAAM. To supplement these benefits, effective cost modelling of the WAAM process is required. This paper introduces a method of generating a detailed Activity Based Cost model for WAAM. A literature review of the current cost models developed in metal additive manufacturing is provided in section 2. The method adopted for cost model development is explained in section 3. The results of cost model and sensitivity analysis are presented in section 4. The outcomes and advantages and disadvantages of this method are discussed in section 5, with final conclusions in section 6.
Figure 1 Schematic of Wire and Arc Additive Manufacturing with integrated rolling [5]
2. Literature review of cost modelling in wire and arc additive manufacture Cost is usually the key point for decision making and cost modelling followed by a break-even analysis is a key method for determining manufacturing process selection. As AM technology has matured it has become possible to compare cost effectiveness with the results achieved by traditional technology. A consequence of process maturation, is that processing uncertainty is reduced allowing increasing accurate cost models to be developed. Cost models used for traditional production methodologies focus on material and labor costs, while modern automated manufacturing processes need cost models able to consider the high impact of investments and overheads Ruffo et
651
652
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
al. [9] and Baumers et al. [10] attempted to quantify this effect for laser sintering through activity based costing where each activity comprises direct and indirect costs. Direct costs comprised only materials used and indirect allocations were costs of machine, overhead, and administration and were time based. Baumers et al, [10] demonstrated that average production cost is dependent on the full capacity of the build plate. Maximum utilization reduces the effect of indirect costs by amortization over the highest number of parts. Both models, however, did not consider the effects of post-processing. Lindemann et al, [11] was one of the first to integrate these costs and found them to be the third most significant cost driver, behind material and investment. Schröder et al, [13] developed an extensive cost model to compare Stereolithography, Fused Deposition Modelling, Selective Laser Melting, Electron Beam Melting and Laser Cladding processes. The most cost-influencing factor from sensitivity analysis was the investment costs of the machine for all the manufacturing processes proposed. Due to the lower maturity compared to powder based metal AM, limited publications referring the cost modelling of the wire based methods and WAAM especially exist. In 2015, Martina et al. [3] developed a time activity based cost model. This simplified deposition times using volume characteristics taken from CAD models in combination with Buy-to-Fly (BTF) ratio, the weight ratio of raw material used for a component and the weight of the component itself. Using this model Martina et al. compared WAAM to traditional CNC machining. The approach is a robust early stage cost estimate tool, however, many of the activity costs were dependent on industry modifiers to approximate time input. Another simplification was that the deposition and post processing machining times were volume-based neglecting the impact of deposition paths and part complexity. Additionally, many process aspects were not considered, such as setup, inspection, rolling, and heat treatment despite the potential to be a substantial cost driver in the WAAM process. The aim of this research is to incorporate the full WAAM process chain to create a detailed cost model. Additionally, deposition paths tool will be used to accurately approximate deposition times. WAAM specific cost drivers will be identified through sensitivity analysis for case study components. Providing ranges of possible values, or an uncertainty to the input values to the indirect and direct cost of activities will generate less subjective and more robust results than those formed by singular inputs generated from expert experience. 3. Cost modelling and sensitivity analysis of the WAAM process Using a time-based Activity Based Costing approach [13], a detailed cost model has been designed and realised. The activities of the WAAM production process chain were identified according to the major production processes involved, and can be listed as substrate preparation, set up, deposition, cooling, rolling, heat treatment, machining, substrate removal and inspection (see figure 2a). Wire deposition, cooling and rolling are repeated as cyclical processes for each layer until the entire part geometry has been achieved.
Figure 2 a) The WAAM cost activities b) Cost model methodology for generating activity cost based on activity time and cost rate.
The activity cost rates were based on the summation of the direct and indirect costs for each activity. Sales and overhead costs such as administration, management and facility wide excluded according to best practice [14]. Combined with activity time, an activity cost can be generated as shown in figure 2b. The WAAM machine data and
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
user inputs required to calculate the direct and indirect costs were collected through literature and business bestpractice. For this cost model, machine data was defined as fixed operating parameters of the machine between set ups. For example, coolant flow rate is one machine data parameter used in calculation of machining cost. User inputs were defined as variables much more likely to change between set ups, for example wire gauge and material. The direct costs were generated from this information using the latest market prices for the consumables mentioned. Indirect costs were mainly machine data generated e.g. labour, and the hardware cost is fixed. However, actual implementation of this cost model relies on activity cost rates and times properly sourced by the specific company employing the model using their own production variables. A comparison of WAAM total production cost for the two case study components shown in figure 3 was made to two alternative powder based metal AM processes. The machine data and user inputs used for the direct costs of WAAM activity is presented in table 1. a
b
Figure 3 WAAM Cost Model Case Study Parts a) Turbine part b) X-shape part
Substrate preparation times were generated based on finish machining a 520x320x35mm billet to size using a face milling material removal rates (MMR’s). Time inputs for WAAM set up, rolling, cooling were generated modifying factors were sourced from literature based on work at by Williams et al. at Cranfield University making aerospace parts in Titanium Alloy Ti6Al4V, equating cooling times to deposition times [3]. This is a limitation as this would change based on the WAAM deposition material. Therefore, to ensure accuracy the cost model was fixed to process the commonly used Titanium Alloy Ti6Al4V. Wire deposition times were generated on deposition path based time inputs to the cost model were obtained for two case study components. The tool paths of deposition for both case study components were able to be sourced through process planning software by Autodesk which included motion paths in which the welding torch was not actively depositing. The active deposition and motion times were calculated based on travel speeds of 0.2m/min and 2m/min respectively. Using wire gauge of 1.2mm and wire feed speed of 3.6m/min this corresponds to 1.1kg/hr. These production variables were developed by Williams et al.[3] and have been shown to achieve a BTF ratio of 1.5. A tool path error of 20% is possible due to various factors such as waiting times, software reaction delays and machine kinematics and dynamics and to take a conservative approach the worst case scenario was included for the comparison analysis [15]. Heat treatment times were acquired from the literature on standard times for Ti6Al4V and a nominal 120 minutes with 20 minutes ramping time was given [16]. Machining times were calculated using the BTF ratio of 1.5. Similarly, to substrate preparation, machining times were calculated based on cutting tool metal removal rate’s (MRR’s), tool change and set up time. Estimates of inspection times were based on user inputs by the authors of 30 minutes and 90 minute set up time. Further assumptions that were necessary to simplify the model design and operation are listed as follows. Indirect costs were based on the estimations of the labour and machine costs for each activity. Assuming an annual capacity utilisation of 46% based on 252 days of operation and 16 hours, machine uptime daily it was possible to calculate the indirect depreciation costs. Labour input to each activity was approximated and also included as an indirect cost. It was assumed that the WAAM machine requires constant supervision, however this is a conservative estimation as the process matures and becomes more automated.
653
654
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657 Table 1 User and machine data inputs used in the activity based costing the WAAM process for direct costs WAAM Activity
Substrate Preparation
WAAM deposition
Heat Treatment
Machining
Cost Inputs
User Input
Machine Parameter
Set Up Loading Machining Tooling and consumables Unloading Substrate purchase
Substrate size Substrate material Substrate cost/kg Cutting tool cost No of parts per build
Cutting tool life Material Removal Rate Set up time Loading time Coolant flow rate and cost/litre CNC power rating Non-cutting motion time
Active Deposition Motion Cooling Rolling Wire
Autodesk motion and deposition tool paths Cooling and rolling time Wire cost/kg Inert gas cost/kg
Heat treatment
Heat treatment time Ramping Time
Set Up Loading Machining Tooling and consumables Unloading
BTF Ratio
Substrate Removal
Set up EDM consumables
EDM time
Inspection
Inspection
Set up Inspection time
Travel speed Wire feed Speed Inert gas flow rate
Furnace Power Rating Cutting tool life Material Removal Rate Set up time Loading time Coolant flow rate CNC power rating Non-cutting motion time EDM wire consumption rate and cost/kg EDM feed rate EDM Power Rating -
Comparison of total production cost of the two case study components was made to two alternative powder based metal AM processes, Electron Beam Melting (EBM) and Direct Melting Laser Sintering (DMLS) and conventional CNC machining. For the AM processes, this was performed using a summary metrics developed by Baumers et al [10]. This provided an equivalent cost per kilogram of product for each alternative process. The cost rates were updated to 2017 Pound Sterling (£) [18] as shown in table 2. Comparison was made for WAAM BTF ratios of 5, 10, 15, and 20. Inspection costs were omitted from calculations as similar costs were not available for the other processes. Table 2 Equivalent cost per cubic centimeter in EBAM and DMLS with currency conversion rate 2010-2017 [17] AM System
Total cost per cm3
Electron Beam Melting
£2.84
Direct Metal Laser Sintering
£7.44
Currency conversion rate for 2010-2017 1.187
A sensitivity analysis was carried out to identify the key cost drivers of the WAAM process. This was possible by applying maximum and minimum ranges to the machine and user inputs considering what values may realistically be found in industry and accounting for price fluctuations. Industry indexes were used to collect average values for variables and rates. To identify the most important variables of the model, a One-Factor-at-a-Time (OFAT) sensitivity analysis approach was taken [18].This would be a step in verifying the results of the cost comparison as well as a start in identifying the variables where variable uncertainty would have the highest cost impact. Sensitivity indices were calculated using an equation:
655
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
(1)
As case study component 1 (Figure 2a) is larger than component 2 (Figure 2b) the difference in size will provide an indication of the sensitivity of the cost drivers to part size. The processing time inputs were varied by 20% for deposition and machining activities. The machine data and user inputs selected for sensitivity analysis and the applied maximum and minimum values are presented in table 2. Table 2 Sensitivity analysis variables selected from machine data and user inputs and the applied ranges Range
Range
Indirect Cost Sensitivity Analysis Variables
Min
Max
Direct cost Sensitivity Analysis Variables
Min
Machine Uptime (hrs)
0
24
Machining tool life (mins)
5
200
No of workers per shift
1
5
Machining tool cost (£)
20
100
Labour salary (£/annum)
22,000
40,000
Coolant cost (£/litre)
0.1
1
WAAM machine cost (£)
326,370
2,500,000
Coolant flow rate (litre/min)
0.1
30
CNC machine cost (£)
120,000
300,000
Wire cost (£/kg)
100
250
Heat treatment machine cost (£)
60,000
300,000
Number of parts per build
1
10
Depreciation years
1
10
Inert gas flow rate (litre/min)
50
400
Inert gas cost (£/canister)
60
350
Max
4. Results The cost of the Propeller and X-part were estimated by the cost model to be £18,135 – £22,419 and £1,986 – £2,211 respectively dependent on the tool path uncertainty range. The results from a cost comparison of the case study parts against two additive processes and conventional CNC machining can be seen in table 3. Table 3 Comparison of total WAAM cost of production of the two case study components with alternative manufacturing processes
Manufacturing Process
AM
Conventional CNC Machining
Case Study 1 Propeller
Case Study 2 X-Part
Cost
Reduction (%)
Cost
Reduction (%)
WAAM
£18,359
-
£1,703
-
Electron Beam
£33,362
45%
£2,123
20%
Direct Metal Laser Sintering
£86,267
79%
£5,489
69%
BTF 5
£18,732
2%
£1,703
0%
BTF 10
£38,166
52%
£3,687
54%
BTF 15
£57,549
68%
£5,483
69%
BTF 20
£76,983
76%
£7,329
77%
Note the price of each component is lower due to the necessary exclusion of inspection cost to make a fair comparison to the Baumers et al. model [10]. The results of the sensitivity analysis for the five highest ranking variables are shown in table 4 for each case study part. Sensitivity analysis performed on ±20% variation in processing times for deposition and machining showed that deposition had a higher impact. An increase in cost of approximately ±5% and ±11% was realised for the X-Part and Propeller respectively for change in deposition process times, compared to machining which only caused ±1% change in the cost. Set up costs had minimal effect for both case study parts ranging between a maximum of 3.4% for X-part deposition to a minimum of 0.1% for propeller deposition and other activities.
656
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657 Table 4 Five highest ranking variables identified through sensitivity analysis for the case study parts
Sensitivity Ranking
Case Study 1 - Propeller
Case Study 2 X-Part
Variable
(%)
Variable
(%)
1
Daily uptime
71.2
Daily uptime
88.6
2
Shield gas flow rate
43.5
Depreciation years
72.0
3
Shield gas cost
42.5
WAAM machine cost
44.0
4
WAAM machine cost
39.3
Heat treatment time
39.0
5
Depreciation years
31.9
Parts per build
28.0
5. Discussion The cost model developed showed that the WAAM process outperforms all the other manufacturing processes both case study parts, even with CNC machining operating at a material removal rate close to 9.3 kg/hr. This shows that WAAM process displays a promising level of cost effectiveness. However, a limitation is the volumetric and MRR simplification in calculating WAAM machining time. Not accounting for motion, part complexity or machining strategy means that WAAM cannot yet be conclusively said to outperform CNC and requires further evidencing. WAAM appears to be highly cost effective compared to the AM processes investigated. As the AM process cost models assumed densely packed parts covering most of the available build volume, actual costs for single parts are likely be higher. Further, as WAAM is a less mature technology, likely to be decreasing cost of investment will improve cost effectiveness in the future. The sensitivity analysis showed that both parts were most sensitive to changes in machine uptime indicating the significance of indirect costs. Case study part 2, the smaller X-part was dominated by indirect costs. Parts per build were significant and it would be beneficial to amortise set up costs, heat treatment and depreciation costs over more parts. This corresponds with the sensitivity analysis performed by Schröder et al, [12], possibly due to similarity in part size. In contrast, the larger case study 1 propeller, was more sensitive to direct material related variables. In particular, the inert gas flow rate and cost per litre was of high impact. Previously, it was thought that this was only a minor part of the total material costs [3][19]. However, the sensitivity ranking showed the shielding inert-gas flow rate and cost having larger impacts than both wire material cost fluctuations and labour costs. This is due to the significantly longer deposition times and highlights that further research into minimising shielding costs would be highly beneficial for the production of large parts. One of the advantages of this model is the accuracy introduced to deposition times represented by real tool paths. However, due to the linked deposition rate and BTF, it limits the applicability of the results to Ti6Al4V, medium to large components. To identify the limits of cost effectiveness of WAAM in general compared to other processes, this needs to be performed for a range of materials and part sizes. Generally, in AM, there is need for work towards the quantification of ‘ill’ defined costs such as inspection. Inclusion of inspection in the process chain of this cost model is an improvement, however, the assumption that this is fixed at 90 minutes set up and 30 minutes processing is an over-simplification. This is because it is influenced by the process capability of the WAAM process for specific materials and component designs. Special processing methods may be required for example heating, inclusion of sacrificial features or tooling has to ensure excessive residual stresses, deflection or cracking does not occur. The cost model in this study can act as a base for establishing the cost of WAAM production in many other materials and component designs. It is envisioned that this knowledge, paired with growing understanding of process capability, will allow better quantification of ill-defined costs and allow more realistic and accurate comparisons between WAAM and alternative manufacturing processes in the future. 6. Conclusions In this paper, a WAAM cost model has been designed and realised to quantify the cost effectiveness of WAAM production. The cost model, for the first time, includes the full process chain for the WAAM process and includes tool path based deposition time to improve the accuracy of the deposition time estimation and account for the
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
various tool path strategies that may be adopted in the manufacture of a part. The results show significant costs savings can be made for two case study Ti6Al4V parts. A cost reduction of 20-45% is found compared to electron beam additive manufacturing and 69-79% reduction for direct metal laser sintering. Compared to conventional CNC machining a breakeven point is found at a BTF ratio of 5. However, an average cost reduction of 53% can be achieved by a BTF ratio of 10. Using this model, a sensitivity analysis identified the key cost drivers in the WAAM process and has shown that indirect costs have a substantial impact for both large and small components. Shielding cost was highlighted as an area of particular impact for large components produced. Further work will focus on systematically expanding the WAAM cost model for other materials, part sizes, deposition rates and BTF ratios. It is envisioned this will establish the limits of cost effectiveness of the WAAM process to enable optimal process selection and potentially highlight new areas of opportunity. Acknowledgements The authors would like to acknowledge support from EPSRC case studentship No. 1780168 with Renishaw plc, and the Innovate UK project entitled RAWFEED No. 101663 as well as support from Autodesk, Airbus and the Welding Engineering and Laser Processing Centre at Cranfield University. References [1] T. Wohlers, Wohlers report 2016. WOHLERS Associates, 2016. [2] ASTM Standard Terminology for Additive Manufacturing Technologies, F2792-12A, 2013. [3] S. Williams and F. Martina, ‘Wire + arc additive manufacturing vs. traditional machining from solid: a cost comparison’, http://waammat.com/documents/waam-vs-machining-from-solid-a-cost-comparison, April 2015. [4] J. Mehnen, J. Ding, H. Lockett, and P. Kazanas, ‘Design study for wire and arc additive manufacture’, Int. J. of Product Development, vol. 19, No. 2, pp. 2-20, 2014. [5] P. A. Colegrove, H. E. Coules, J. Fairman, F. Martina, T. Kashoob, H. Mamash, L. D. Cozzolino, ‘Microstructure and residual stress improvement in wire and arc additively manufactured parts through high-pressure rolling’, J Mater Process Technol, vol. 213, no. 10, p. 1782-1791, 2013. [6] J. Donoghue, A. A. Antonysamy, F. Martina, P. A. Colegrove, S. W. Williams, P. B. Prangnell, ‘The effectiveness of combining rolling deformation with Wire-Arc Additive Manufacture on beta-grain refinement and texture modification in Ti-6Al-4V’, Materials Characterization, vol. 114, pp. 103–114, Apr. 2016. [7] S. W. Williams, F. Martina, A. C. Addison, J. Ding, G. Pardal, P. Colegrove, ‘Wire + Arc Additive Manufacturing’, Mater Sci and Tech, vol. 32, no. 7, pp. 641–647, 2016. [8] P. A. Colegrove, A. McAndrew, J. Ding, S. Williams, ‘Systems architecture for large scale wire + arc additive manufacture’, presented at the 10th International Conference on Trends in Welding Research, Tokyo, Japan, 2016. [9] M. Ruffo, C. Tuck, and R. Hague, ‘Cost estimation for rapid manufacturing - Laser sintering production for low to medium volumes’, presented at the Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 220, no. 9, pp. 1417-1427. 2006. [10] M. Baumers, C. Tuck, R. Wildman, I. Ashcroft, E. Rosamond, and R. Hague, ‘Combined buildtime, energy consumption and cost estimation for direct metal laser sintering’, presented at the Twenty Third Annual International Solid Freeform Fabrication Symposium, 2012. [11] C. Lindemann, U. Jahnke, and M. Moi, ‘Analyzing product lifecycle costs for a better understanding of cost drivers in additive manufacturing’, presented at the Twenty Third Annual International Solid Freeform Fabrication Symposium, 2012. [12] M. Schröder, B. Falk, and R. Schmidt, ‘Evaluation of Cost Structures of Additive Manufacturing Processes Using a New Business Model’, presented at the 7th Industrial Product-Service Systems Conference, Proced CIRP, no. 30, pp. 311–316, 2015. [13] R. Kaplan and S. R. Anderson, Time-driven activity-based costing: a simpler and more powerful path to higher profits. 2013. [14] G. Costabile, M. Fera, F. Fruggiero, A. Lambiase and D. Pham, ‘Cost models of additive manufacturing: A literature review’, Int J of Industrial Engineering Computations, vol. 8, pp. 263-282, 2017. [15] K. Hamilton, ‘Toolpath limitations’, private communication, Autodesk, Birmingham UK, April 2017. [16] G. Welsch, R. Boyer, and E. W. Collings, Materials properties handbook: titanium alloys, ASM International, 1993. [17] Triami Media BV, ‘Historic Inflation Great Britain - CPI Inflation’ [online]. Available: http://www.inflation.eu, 2017. [18] A. van Griensven, T. Meixner, S. Grunwald, T. Bishop, M. Diluzio, and R. Srinivasan, ‘A global sensitivity analysis tool for the parameters of multi-variable catchment models’, J of Hydrology, vol. 324, no. 1, pp. 10-23, 2006. [19] Y. Zhai, Early Cost Estimation for Additive Manufacture, MSc Thesis, Cranfield University, 2012.
657
ScienceDirect Procedia Manufacturing 11 (2017) 650 – 657
27th International Conference on Flexible Automation and Intelligent Manufacturing, FAIM2017, 27-30 June 2017, Modena, Italy
Cost modelling and sensitivity analysis of wire and arc additive manufacturing C. R. Cunningham*, S. Wikshåland, F.Xu, N. Kemakolam, A.Shokrani, V. Dhokia, and S.T. Newman Department of Mechnical Engineering, University of Bath, Bath, BA2 7AY, UK
Abstract With the proliferation and diversification of metal additive manufacturing (AM) processes in recent years, effective decision tools for process selection are of increasing importance. This paper presents a novel time activity based cost model for Wire-Arc Additive Manufacturing (WAAM), an emergent metal AM technology. The full process chain is modelled and a tool-path based deposition cost employed. The results show that WAAM has significant potential as a cost-effective manufacturing approach compared to other AM and conventional CNC machining methods. Sensitivity analysis identifies indirect costs as a key cost driver generally, with parts per build plate, and shielding cost having significance for smaller and larger parts respectively. © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license © 2017 The Authors. Published by Elsevier B.V. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-reviewunder underresponsibility responsibility of the scientific committee the International 27th International Conference on Flexible Automation and Peer-review of the scientific committee of theof27th Conference on Flexible Automation and IntelligentManufacturing Manufacturing. Intelligent Keywords: Wire and Arc Additive Manufacturing; Cost Modelling; WAAM; Additive Manufacturing
1. Introduction Additive Manufacturing (AM) has become the subject of a substantial amount of research due to promising industrial potential, with AM products and services set for sales of $26 billion worldwide by 2021[1]. The concept of creating metal products additively has seen an increased uptake in recent years. AM provides opportunities for
* Corresponding author. Tel.: +44-1225-386934; fax: +44-1225-386928. E-mail address: [email protected]
2351-9789 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the 27th International Conference on Flexible Automation and Intelligent Manufacturing doi:10.1016/j.promfg.2017.07.163
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
reduced weight and cost savings and has been especially prevalent in the aerospace industry where there have been instances of conventional manufacturing processes being replaced by metal AM processes. Processing limitations have led to the diversification of metal additive manufacturing techniques especially within the categories of Powder Bed Fusion (PBF), Binder Jetting and Direct Energy Deposition (DED) [2], each with their distinct benefits. Understanding the performance of metal AM processes in terms of cost effectiveness is essential for effective deployment and further uptake of AM. Cost effectiveness is especially important as a measure because industrial standards of quality control are rarely achieved in metal AM. The combination of knowledge of process capability and effective cost modelling can provide a valuable insight into the potential cost escalation or “real cost” of a particular AM process. Wire and Arc Additive Manufacturing (WAAM), is a DED approach that is seeing considerable interest due to its ability to create large metal products faster than powder-based alternatives, with deposition rates up to 4 kg/hr [3]. The process uses an electrical arc as a heat-source to melt feedstock metal wire, which is deposited onto a base plate [4]. The WAAM process comprises successive cycles of melting, depositing and cooling to result in a near-net shape deposit. Cold-work through rolling may be subsequently employed to improve microstructural properties and relieve residual stresses induced by thermal cycles incurred through the deposition process [5][6]. WAAM with an integrated roller consists of an arc torch is shown in Figure 1 [5]. To achieve end-product status, heat treatment for residual stress relief and/or to develop mechanical properties and finish machining is required [8]. Compared to powder based AM methods, WAAM offers several distinct benefits. In certain materials, directed deposition and cheaper wire feedstock mean that cost be reduced significantly compared to PBF. Further, processing issues such as powder agglomeration, recycling are overcome. Concerns with development of full density and microstructure of non-heat treatable materials in PBF and Binder Jetting may be overcome by the inter-pass rolling which is easily implemented in WAAM. To supplement these benefits, effective cost modelling of the WAAM process is required. This paper introduces a method of generating a detailed Activity Based Cost model for WAAM. A literature review of the current cost models developed in metal additive manufacturing is provided in section 2. The method adopted for cost model development is explained in section 3. The results of cost model and sensitivity analysis are presented in section 4. The outcomes and advantages and disadvantages of this method are discussed in section 5, with final conclusions in section 6.
Figure 1 Schematic of Wire and Arc Additive Manufacturing with integrated rolling [5]
2. Literature review of cost modelling in wire and arc additive manufacture Cost is usually the key point for decision making and cost modelling followed by a break-even analysis is a key method for determining manufacturing process selection. As AM technology has matured it has become possible to compare cost effectiveness with the results achieved by traditional technology. A consequence of process maturation, is that processing uncertainty is reduced allowing increasing accurate cost models to be developed. Cost models used for traditional production methodologies focus on material and labor costs, while modern automated manufacturing processes need cost models able to consider the high impact of investments and overheads Ruffo et
651
652
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
al. [9] and Baumers et al. [10] attempted to quantify this effect for laser sintering through activity based costing where each activity comprises direct and indirect costs. Direct costs comprised only materials used and indirect allocations were costs of machine, overhead, and administration and were time based. Baumers et al, [10] demonstrated that average production cost is dependent on the full capacity of the build plate. Maximum utilization reduces the effect of indirect costs by amortization over the highest number of parts. Both models, however, did not consider the effects of post-processing. Lindemann et al, [11] was one of the first to integrate these costs and found them to be the third most significant cost driver, behind material and investment. Schröder et al, [13] developed an extensive cost model to compare Stereolithography, Fused Deposition Modelling, Selective Laser Melting, Electron Beam Melting and Laser Cladding processes. The most cost-influencing factor from sensitivity analysis was the investment costs of the machine for all the manufacturing processes proposed. Due to the lower maturity compared to powder based metal AM, limited publications referring the cost modelling of the wire based methods and WAAM especially exist. In 2015, Martina et al. [3] developed a time activity based cost model. This simplified deposition times using volume characteristics taken from CAD models in combination with Buy-to-Fly (BTF) ratio, the weight ratio of raw material used for a component and the weight of the component itself. Using this model Martina et al. compared WAAM to traditional CNC machining. The approach is a robust early stage cost estimate tool, however, many of the activity costs were dependent on industry modifiers to approximate time input. Another simplification was that the deposition and post processing machining times were volume-based neglecting the impact of deposition paths and part complexity. Additionally, many process aspects were not considered, such as setup, inspection, rolling, and heat treatment despite the potential to be a substantial cost driver in the WAAM process. The aim of this research is to incorporate the full WAAM process chain to create a detailed cost model. Additionally, deposition paths tool will be used to accurately approximate deposition times. WAAM specific cost drivers will be identified through sensitivity analysis for case study components. Providing ranges of possible values, or an uncertainty to the input values to the indirect and direct cost of activities will generate less subjective and more robust results than those formed by singular inputs generated from expert experience. 3. Cost modelling and sensitivity analysis of the WAAM process Using a time-based Activity Based Costing approach [13], a detailed cost model has been designed and realised. The activities of the WAAM production process chain were identified according to the major production processes involved, and can be listed as substrate preparation, set up, deposition, cooling, rolling, heat treatment, machining, substrate removal and inspection (see figure 2a). Wire deposition, cooling and rolling are repeated as cyclical processes for each layer until the entire part geometry has been achieved.
Figure 2 a) The WAAM cost activities b) Cost model methodology for generating activity cost based on activity time and cost rate.
The activity cost rates were based on the summation of the direct and indirect costs for each activity. Sales and overhead costs such as administration, management and facility wide excluded according to best practice [14]. Combined with activity time, an activity cost can be generated as shown in figure 2b. The WAAM machine data and
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
user inputs required to calculate the direct and indirect costs were collected through literature and business bestpractice. For this cost model, machine data was defined as fixed operating parameters of the machine between set ups. For example, coolant flow rate is one machine data parameter used in calculation of machining cost. User inputs were defined as variables much more likely to change between set ups, for example wire gauge and material. The direct costs were generated from this information using the latest market prices for the consumables mentioned. Indirect costs were mainly machine data generated e.g. labour, and the hardware cost is fixed. However, actual implementation of this cost model relies on activity cost rates and times properly sourced by the specific company employing the model using their own production variables. A comparison of WAAM total production cost for the two case study components shown in figure 3 was made to two alternative powder based metal AM processes. The machine data and user inputs used for the direct costs of WAAM activity is presented in table 1. a
b
Figure 3 WAAM Cost Model Case Study Parts a) Turbine part b) X-shape part
Substrate preparation times were generated based on finish machining a 520x320x35mm billet to size using a face milling material removal rates (MMR’s). Time inputs for WAAM set up, rolling, cooling were generated modifying factors were sourced from literature based on work at by Williams et al. at Cranfield University making aerospace parts in Titanium Alloy Ti6Al4V, equating cooling times to deposition times [3]. This is a limitation as this would change based on the WAAM deposition material. Therefore, to ensure accuracy the cost model was fixed to process the commonly used Titanium Alloy Ti6Al4V. Wire deposition times were generated on deposition path based time inputs to the cost model were obtained for two case study components. The tool paths of deposition for both case study components were able to be sourced through process planning software by Autodesk which included motion paths in which the welding torch was not actively depositing. The active deposition and motion times were calculated based on travel speeds of 0.2m/min and 2m/min respectively. Using wire gauge of 1.2mm and wire feed speed of 3.6m/min this corresponds to 1.1kg/hr. These production variables were developed by Williams et al.[3] and have been shown to achieve a BTF ratio of 1.5. A tool path error of 20% is possible due to various factors such as waiting times, software reaction delays and machine kinematics and dynamics and to take a conservative approach the worst case scenario was included for the comparison analysis [15]. Heat treatment times were acquired from the literature on standard times for Ti6Al4V and a nominal 120 minutes with 20 minutes ramping time was given [16]. Machining times were calculated using the BTF ratio of 1.5. Similarly, to substrate preparation, machining times were calculated based on cutting tool metal removal rate’s (MRR’s), tool change and set up time. Estimates of inspection times were based on user inputs by the authors of 30 minutes and 90 minute set up time. Further assumptions that were necessary to simplify the model design and operation are listed as follows. Indirect costs were based on the estimations of the labour and machine costs for each activity. Assuming an annual capacity utilisation of 46% based on 252 days of operation and 16 hours, machine uptime daily it was possible to calculate the indirect depreciation costs. Labour input to each activity was approximated and also included as an indirect cost. It was assumed that the WAAM machine requires constant supervision, however this is a conservative estimation as the process matures and becomes more automated.
653
654
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657 Table 1 User and machine data inputs used in the activity based costing the WAAM process for direct costs WAAM Activity
Substrate Preparation
WAAM deposition
Heat Treatment
Machining
Cost Inputs
User Input
Machine Parameter
Set Up Loading Machining Tooling and consumables Unloading Substrate purchase
Substrate size Substrate material Substrate cost/kg Cutting tool cost No of parts per build
Cutting tool life Material Removal Rate Set up time Loading time Coolant flow rate and cost/litre CNC power rating Non-cutting motion time
Active Deposition Motion Cooling Rolling Wire
Autodesk motion and deposition tool paths Cooling and rolling time Wire cost/kg Inert gas cost/kg
Heat treatment
Heat treatment time Ramping Time
Set Up Loading Machining Tooling and consumables Unloading
BTF Ratio
Substrate Removal
Set up EDM consumables
EDM time
Inspection
Inspection
Set up Inspection time
Travel speed Wire feed Speed Inert gas flow rate
Furnace Power Rating Cutting tool life Material Removal Rate Set up time Loading time Coolant flow rate CNC power rating Non-cutting motion time EDM wire consumption rate and cost/kg EDM feed rate EDM Power Rating -
Comparison of total production cost of the two case study components was made to two alternative powder based metal AM processes, Electron Beam Melting (EBM) and Direct Melting Laser Sintering (DMLS) and conventional CNC machining. For the AM processes, this was performed using a summary metrics developed by Baumers et al [10]. This provided an equivalent cost per kilogram of product for each alternative process. The cost rates were updated to 2017 Pound Sterling (£) [18] as shown in table 2. Comparison was made for WAAM BTF ratios of 5, 10, 15, and 20. Inspection costs were omitted from calculations as similar costs were not available for the other processes. Table 2 Equivalent cost per cubic centimeter in EBAM and DMLS with currency conversion rate 2010-2017 [17] AM System
Total cost per cm3
Electron Beam Melting
£2.84
Direct Metal Laser Sintering
£7.44
Currency conversion rate for 2010-2017 1.187
A sensitivity analysis was carried out to identify the key cost drivers of the WAAM process. This was possible by applying maximum and minimum ranges to the machine and user inputs considering what values may realistically be found in industry and accounting for price fluctuations. Industry indexes were used to collect average values for variables and rates. To identify the most important variables of the model, a One-Factor-at-a-Time (OFAT) sensitivity analysis approach was taken [18].This would be a step in verifying the results of the cost comparison as well as a start in identifying the variables where variable uncertainty would have the highest cost impact. Sensitivity indices were calculated using an equation:
655
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
(1)
As case study component 1 (Figure 2a) is larger than component 2 (Figure 2b) the difference in size will provide an indication of the sensitivity of the cost drivers to part size. The processing time inputs were varied by 20% for deposition and machining activities. The machine data and user inputs selected for sensitivity analysis and the applied maximum and minimum values are presented in table 2. Table 2 Sensitivity analysis variables selected from machine data and user inputs and the applied ranges Range
Range
Indirect Cost Sensitivity Analysis Variables
Min
Max
Direct cost Sensitivity Analysis Variables
Min
Machine Uptime (hrs)
0
24
Machining tool life (mins)
5
200
No of workers per shift
1
5
Machining tool cost (£)
20
100
Labour salary (£/annum)
22,000
40,000
Coolant cost (£/litre)
0.1
1
WAAM machine cost (£)
326,370
2,500,000
Coolant flow rate (litre/min)
0.1
30
CNC machine cost (£)
120,000
300,000
Wire cost (£/kg)
100
250
Heat treatment machine cost (£)
60,000
300,000
Number of parts per build
1
10
Depreciation years
1
10
Inert gas flow rate (litre/min)
50
400
Inert gas cost (£/canister)
60
350
Max
4. Results The cost of the Propeller and X-part were estimated by the cost model to be £18,135 – £22,419 and £1,986 – £2,211 respectively dependent on the tool path uncertainty range. The results from a cost comparison of the case study parts against two additive processes and conventional CNC machining can be seen in table 3. Table 3 Comparison of total WAAM cost of production of the two case study components with alternative manufacturing processes
Manufacturing Process
AM
Conventional CNC Machining
Case Study 1 Propeller
Case Study 2 X-Part
Cost
Reduction (%)
Cost
Reduction (%)
WAAM
£18,359
-
£1,703
-
Electron Beam
£33,362
45%
£2,123
20%
Direct Metal Laser Sintering
£86,267
79%
£5,489
69%
BTF 5
£18,732
2%
£1,703
0%
BTF 10
£38,166
52%
£3,687
54%
BTF 15
£57,549
68%
£5,483
69%
BTF 20
£76,983
76%
£7,329
77%
Note the price of each component is lower due to the necessary exclusion of inspection cost to make a fair comparison to the Baumers et al. model [10]. The results of the sensitivity analysis for the five highest ranking variables are shown in table 4 for each case study part. Sensitivity analysis performed on ±20% variation in processing times for deposition and machining showed that deposition had a higher impact. An increase in cost of approximately ±5% and ±11% was realised for the X-Part and Propeller respectively for change in deposition process times, compared to machining which only caused ±1% change in the cost. Set up costs had minimal effect for both case study parts ranging between a maximum of 3.4% for X-part deposition to a minimum of 0.1% for propeller deposition and other activities.
656
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657 Table 4 Five highest ranking variables identified through sensitivity analysis for the case study parts
Sensitivity Ranking
Case Study 1 - Propeller
Case Study 2 X-Part
Variable
(%)
Variable
(%)
1
Daily uptime
71.2
Daily uptime
88.6
2
Shield gas flow rate
43.5
Depreciation years
72.0
3
Shield gas cost
42.5
WAAM machine cost
44.0
4
WAAM machine cost
39.3
Heat treatment time
39.0
5
Depreciation years
31.9
Parts per build
28.0
5. Discussion The cost model developed showed that the WAAM process outperforms all the other manufacturing processes both case study parts, even with CNC machining operating at a material removal rate close to 9.3 kg/hr. This shows that WAAM process displays a promising level of cost effectiveness. However, a limitation is the volumetric and MRR simplification in calculating WAAM machining time. Not accounting for motion, part complexity or machining strategy means that WAAM cannot yet be conclusively said to outperform CNC and requires further evidencing. WAAM appears to be highly cost effective compared to the AM processes investigated. As the AM process cost models assumed densely packed parts covering most of the available build volume, actual costs for single parts are likely be higher. Further, as WAAM is a less mature technology, likely to be decreasing cost of investment will improve cost effectiveness in the future. The sensitivity analysis showed that both parts were most sensitive to changes in machine uptime indicating the significance of indirect costs. Case study part 2, the smaller X-part was dominated by indirect costs. Parts per build were significant and it would be beneficial to amortise set up costs, heat treatment and depreciation costs over more parts. This corresponds with the sensitivity analysis performed by Schröder et al, [12], possibly due to similarity in part size. In contrast, the larger case study 1 propeller, was more sensitive to direct material related variables. In particular, the inert gas flow rate and cost per litre was of high impact. Previously, it was thought that this was only a minor part of the total material costs [3][19]. However, the sensitivity ranking showed the shielding inert-gas flow rate and cost having larger impacts than both wire material cost fluctuations and labour costs. This is due to the significantly longer deposition times and highlights that further research into minimising shielding costs would be highly beneficial for the production of large parts. One of the advantages of this model is the accuracy introduced to deposition times represented by real tool paths. However, due to the linked deposition rate and BTF, it limits the applicability of the results to Ti6Al4V, medium to large components. To identify the limits of cost effectiveness of WAAM in general compared to other processes, this needs to be performed for a range of materials and part sizes. Generally, in AM, there is need for work towards the quantification of ‘ill’ defined costs such as inspection. Inclusion of inspection in the process chain of this cost model is an improvement, however, the assumption that this is fixed at 90 minutes set up and 30 minutes processing is an over-simplification. This is because it is influenced by the process capability of the WAAM process for specific materials and component designs. Special processing methods may be required for example heating, inclusion of sacrificial features or tooling has to ensure excessive residual stresses, deflection or cracking does not occur. The cost model in this study can act as a base for establishing the cost of WAAM production in many other materials and component designs. It is envisioned that this knowledge, paired with growing understanding of process capability, will allow better quantification of ill-defined costs and allow more realistic and accurate comparisons between WAAM and alternative manufacturing processes in the future. 6. Conclusions In this paper, a WAAM cost model has been designed and realised to quantify the cost effectiveness of WAAM production. The cost model, for the first time, includes the full process chain for the WAAM process and includes tool path based deposition time to improve the accuracy of the deposition time estimation and account for the
C.R. Cunningham et al. / Procedia Manufacturing 11 (2017) 650 – 657
various tool path strategies that may be adopted in the manufacture of a part. The results show significant costs savings can be made for two case study Ti6Al4V parts. A cost reduction of 20-45% is found compared to electron beam additive manufacturing and 69-79% reduction for direct metal laser sintering. Compared to conventional CNC machining a breakeven point is found at a BTF ratio of 5. However, an average cost reduction of 53% can be achieved by a BTF ratio of 10. Using this model, a sensitivity analysis identified the key cost drivers in the WAAM process and has shown that indirect costs have a substantial impact for both large and small components. Shielding cost was highlighted as an area of particular impact for large components produced. Further work will focus on systematically expanding the WAAM cost model for other materials, part sizes, deposition rates and BTF ratios. It is envisioned this will establish the limits of cost effectiveness of the WAAM process to enable optimal process selection and potentially highlight new areas of opportunity. Acknowledgements The authors would like to acknowledge support from EPSRC case studentship No. 1780168 with Renishaw plc, and the Innovate UK project entitled RAWFEED No. 101663 as well as support from Autodesk, Airbus and the Welding Engineering and Laser Processing Centre at Cranfield University. References [1] T. Wohlers, Wohlers report 2016. WOHLERS Associates, 2016. [2] ASTM Standard Terminology for Additive Manufacturing Technologies, F2792-12A, 2013. [3] S. Williams and F. Martina, ‘Wire + arc additive manufacturing vs. traditional machining from solid: a cost comparison’, http://waammat.com/documents/waam-vs-machining-from-solid-a-cost-comparison, April 2015. [4] J. Mehnen, J. Ding, H. Lockett, and P. Kazanas, ‘Design study for wire and arc additive manufacture’, Int. J. of Product Development, vol. 19, No. 2, pp. 2-20, 2014. [5] P. A. Colegrove, H. E. Coules, J. Fairman, F. Martina, T. Kashoob, H. Mamash, L. D. Cozzolino, ‘Microstructure and residual stress improvement in wire and arc additively manufactured parts through high-pressure rolling’, J Mater Process Technol, vol. 213, no. 10, p. 1782-1791, 2013. [6] J. Donoghue, A. A. Antonysamy, F. Martina, P. A. Colegrove, S. W. Williams, P. B. Prangnell, ‘The effectiveness of combining rolling deformation with Wire-Arc Additive Manufacture on beta-grain refinement and texture modification in Ti-6Al-4V’, Materials Characterization, vol. 114, pp. 103–114, Apr. 2016. [7] S. W. Williams, F. Martina, A. C. Addison, J. Ding, G. Pardal, P. Colegrove, ‘Wire + Arc Additive Manufacturing’, Mater Sci and Tech, vol. 32, no. 7, pp. 641–647, 2016. [8] P. A. Colegrove, A. McAndrew, J. Ding, S. Williams, ‘Systems architecture for large scale wire + arc additive manufacture’, presented at the 10th International Conference on Trends in Welding Research, Tokyo, Japan, 2016. [9] M. Ruffo, C. Tuck, and R. Hague, ‘Cost estimation for rapid manufacturing - Laser sintering production for low to medium volumes’, presented at the Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 220, no. 9, pp. 1417-1427. 2006. [10] M. Baumers, C. Tuck, R. Wildman, I. Ashcroft, E. Rosamond, and R. Hague, ‘Combined buildtime, energy consumption and cost estimation for direct metal laser sintering’, presented at the Twenty Third Annual International Solid Freeform Fabrication Symposium, 2012. [11] C. Lindemann, U. Jahnke, and M. Moi, ‘Analyzing product lifecycle costs for a better understanding of cost drivers in additive manufacturing’, presented at the Twenty Third Annual International Solid Freeform Fabrication Symposium, 2012. [12] M. Schröder, B. Falk, and R. Schmidt, ‘Evaluation of Cost Structures of Additive Manufacturing Processes Using a New Business Model’, presented at the 7th Industrial Product-Service Systems Conference, Proced CIRP, no. 30, pp. 311–316, 2015. [13] R. Kaplan and S. R. Anderson, Time-driven activity-based costing: a simpler and more powerful path to higher profits. 2013. [14] G. Costabile, M. Fera, F. Fruggiero, A. Lambiase and D. Pham, ‘Cost models of additive manufacturing: A literature review’, Int J of Industrial Engineering Computations, vol. 8, pp. 263-282, 2017. [15] K. Hamilton, ‘Toolpath limitations’, private communication, Autodesk, Birmingham UK, April 2017. [16] G. Welsch, R. Boyer, and E. W. Collings, Materials properties handbook: titanium alloys, ASM International, 1993. [17] Triami Media BV, ‘Historic Inflation Great Britain - CPI Inflation’ [online]. Available: http://www.inflation.eu, 2017. [18] A. van Griensven, T. Meixner, S. Grunwald, T. Bishop, M. Diluzio, and R. Srinivasan, ‘A global sensitivity analysis tool for the parameters of multi-variable catchment models’, J of Hydrology, vol. 324, no. 1, pp. 10-23, 2006. [19] Y. Zhai, Early Cost Estimation for Additive Manufacture, MSc Thesis, Cranfield University, 2012.
657