Carbon fiber production costing: a modular approach

Abstract

Carbon fiber is expected to increase its importance as a lightweight substitute material. The significant market potential in the automotive industry is strongly dependent on carbon fiber cost. A decrease of about 50% from the present cost level of polyacrylonitrile (PAN)-based carbon fibers is needed. This ambitious target is only achievable with high-cost transparency in the production chain. This paper reviews published cost models and identifies the need for a consistent methodology and transparency over input parameters. A new cost model with a modular structure covering all process steps by cost type and disclosing all process parameter assumptions is introduced. The most cost-intense steps are polymerization, including raw material (25%), carbonization (22%) and fiber washing after coagulation (19%). Most important cost types are energy (34%), raw material (19%) and capital costs for equipment (18%). The high-cost share of 54% for carbon fiber PAN precursor is consistent with most reviewed models.

Keywords

carbon fiber precursor polyacrylonitrile manufacturing cost modeling cost structure

Carbon fiber is one of the most promising materials for lightweight applications in the automotive industry. Production volume was estimated at 67,071 metric tons in 2012 and is forecasted to have significant growth of 8% per year on average until 2020 to 121,896 metric tons.¹

Carbon fiber made from polyacrylonitrile (PAN) was first produced in the 1950s.² A precursor fiber is wet-spun from PAN and comonomers. The as-spun fiber is washed and stretched, dried, a surface preparation is applied and it is wound onto a bobbin. The finished precursor then is thermally stabilized in an oven and finally converted in a furnace process into carbon fiber by pyrolysis. During this process it loses more than 50% of its weight, resulting in a fiber consisting of at least 88–99% carbon.²

Today PAN-based precursors are used for 86% of the carbon fiber in the market and are forecasted to be used for 92% in 2017.³ The precursor has been identified as a main cost driver for carbon fibers in the past. A cost model from Kline & Company allocates 51% of the cost to the precursor material used for conversion.⁴

Carbon fiber reinforced plastic (CFRP) has a high potential as a lightweight material in the automotive industry. It already has a predominant status as the preferred lightweight material in the aerospace industry. Latest models in civil aviation, such as the Boeing Dreamliner and Airbus A350, are made of up to 52% CFRP.³ Also for fast rotating applications, such as wind turbine blades, and manually operated tools, such as sports goods, its strength and modulus advantage over metals is exploited. However, the costs for CFRP components are too high today for a breakthrough in the automotive industry.

Research activities currently mainly focus on low-cost carbon fiber. So-called large tow fibers with more than 24,000 single filaments are generating scale effects and are suitable for automotive applications. The cost for 1 kg of such carbon fiber is around EUR 18 today.⁵ The willingness to pay in the automotive industry, however, is around EUR 10 per kg (USD 5–7 per lb),⁶ resulting in a needed cost decrease of about 50%.

A comparison of cost structures published by leading carbon fiber producers in their annual reports is shown in Table 1. Cost of goods sold – which is mainly manufacturing costs – make between 76% and 88% of the total costs. This means that for decreasing costs of carbon fiber, manufacturing costs have to be reduced significantly.

Table 1.

Cost structure of carbon fiber producers^7–13

	Cytec	Hexcel	Mitsubishi Rayon	SGL	Toho Tenax	Toray	Zoltek
FY	2012	2012	2009	2012	2007	2012	2012
Cost of goods sold	76%	88%	81%	81%	83%	85%	88%
R&D	3%	3%	4%	3%	17%	15%	4%
Distribution	10%	10%	15%	11%			8%
Administration	9%	10%	15%	6%			8%
Others	2%	–1%	0%	–1%			0%

Lower carbon fiber cost is only achievable when different cost levers are pulled. A high-cost transparency in state-of-the-art manufacturing processes is needed for identifying these levers, for assessing and prioritizing cost-reducing process innovations and tracking their impact over time.

Unfortunately, there have been only a few publications on the cost structure of carbon fiber. All reviewed models are either missing parts of the process chain, transparency in underlying assumptions and calculation methods or have an obviously outdated parameterization.

The objective of this paper is to present a carbon fiber cost model that covers the whole process chain from precursor production to final carbon fiber. This paper sets the stage for future carbon fiber production cost research by introducing a clear methodology and disclosing the full set of parameters. Scientist and industry participants are enabled to evaluate the cost impact of process improvements based on this paper. A modular structure will qualify the model to include disruptive changes in the production process from new precursors.

Review of existing models

The cost structure of carbon fiber manufacturing has been analyzed in the past.

At the Massachusetts Institute of Technology (MIT) a spreadsheet methodology for production cost modeling was introduced by Poggiali in 1985.¹⁴ Based on this, Goss¹⁵ shows a detailed cost structure of carbon fiber, presenting all input parameters and calculation methods. The model covers the process chain from finished precursor to carbon fiber. All cost types are covered except for a missing split of equipment and building investment and precursor manufacturing steps.

At Oak Ridge National Laboratories (ORNL) the cost structure of carbon fiber has been analyzed since the beginning of low-cost carbon fiber research in the late 1990s. The first results by Das and Cohn show results without disclosing the underlying assumptions and calculations.^16,17

In 2004 a cost model was developed by Kline & Company for the Automotive Composites Consortium.⁴ The model was not published and results are only shown by ORNL in various presentations.⁶ This model again covers the process chain for precursor conversion to carbon fiber without detailing precursor production.

ORNL published results from a new cost model in 2012.¹⁸ This also starts with a finished precursor as input. Process steps are shown in more detail and costs are disaggregated by more types, such as energy and material.

Equipment manufacturer Trützschler Man-made Fibers also published a cost structure in 2012.¹⁹ It is based on the results from ORNL and own calculations. Vague ranges of values for a production line are disclosed for selected parameters only. This is the only reviewed model giving some transparency in cost structure by type for polymerization and precursor production, but it is missing the granularity by process step.

Recently the conversion furnace manufacturer Harper International published results from their proprietary cost model.^20–22 It again focuses on the conversion processes from PAN precursor into carbon fiber. There are several undisclosed input parameters and there is high transparency in energy consumption in the process steps.

In Table 2 the results of costs by cost type are summarized for the reviewed publications. All process consumables are summarized under material except for the precursor. A clear trend of increasing component of precursor cost can be observed, with the exception of the results published by Harper International. While the precursor accounted for about one third of the carbon fiber costs in 1986, they make up more than half today. According to the reviewed models, the nominal costs for carbon fiber have decreased by ∼50% since the 1980s and show some stagnation in a corridor between USD 18 and 25 per kg over the last decade. Real numbers reflecting US inflation rates show an even stronger decrease by ∼75% since the 1980s and corridor of between USD 20 and 27 per kg over the last decade with a slightly decreasing trend.²³ Possible underlying effects for the cost decrease are increased energy efficiency in solvent recovery and thermal treatment processes, scale effects in equipment and labor and the learning curve theory.

Table 2.

Cost structure by type in reviewed models

	Goss¹⁵ (1986)	Das and Cohn^16,17 (2001)	Kline⁴ (2004)	Harper²² (2011)	Trützschler Man-made Fibers¹⁹ (2012)	Das and Warren¹⁸ (2012)
Precursor	34%	44%	51%	31%	51%	54%
Energy	7%	5%	18%	12%	18%	8%
Material	12%	6%	18%	2%	n/A	4%
Labor	4%	9%	10%	21%	10%	10%
Equipment	32%	21%	21%	18%	12%	21%
Building	32%	21%	21%	16%	12%	21%
Maintenance	7%	15%	n/A	n/A	9%	3%
Insurance	2%		n/A	n/A
Taxes	2%		n/A	n/A
Carbon fiber costs in $/kg	42.8	18.3	21.8	19.1	24.7	21.1
Inflation in % to 2013²³	113%	33%	25%	5%	2%	2%
Inflation-adj. in $/kg	91.4	24.4	27.2	20.1	25.2	21.5

The comparison of the reviewed models along the process chain in Table 3 shows that the dominant cost component is added until the precursor is stabilized and carbonized. Up to this point, between 77% and 90% of the total costs have occurred in the different models.

Table 3.

Cost structure by process step in reviewed models

	Goss¹⁵ (1986)	Das and Cohn^16,17 (2001)	Kline⁴ (2004)	Harper²² (2011)	Trützschler Man-made Fibers¹⁹ (2012)	Das and Warren¹⁸ (2012)
Precursor	34%	47%	51%	31%	51%	54%
Pretreatment	3%	47%	0%	69%	49%	6%
Stabilization	20%	16%	16%			13%
Carbonization	23%	26%	23%			8%
Surface treatm.	4%	2%	4%			3%
Sizing	7%	2%	4%			4%
Winding	5%	3%	6%			9%
Other	4%	4%	0%			3%

However, several inconsistencies can be observed. The model of Das and Warren¹⁸ shows a relatively low-cost component of 21% for stabilization and carbonization. Goss,¹⁵ Das and Cohn^16,17 and Kline & Company⁴ have 39–43% of the total costs allocated to these process steps. A significant spread can also be observed for the cost component of surface treatment and sizing (4–11%) and for winding and handling (3–9%).

Differences in the energy cost component in a range of 5–12% of total fiber costs might be based on different energy price assumptions that are not undisclosed in several models. A broad range can be observed worldwide with USD .03 per kWh reported from carbon fiber manufacturer SGL’s Moses Lake plant up to EUR .09 per kWh for the German industry average in 2012.^24,25

As the calculation method and parameterization of the reviewed models is undisclosed, a deeper analysis of cost differences is difficult. As Table 4 shows, differences in the precursor cost component for the model published by Harper International can be traced back to the input of USD 6 per kg carbon fiber for PAN precursor material. This seems to be much lower than in all other models. For inflation-adjusted costs all models except for the ones published by Goss and Harper show a cost corridor for precursor and pretreatment of USD 11.4–13.9 per kg carbon fiber.

Table 4.

Precursor costs in reviewed models

In $/kg CF	Goss¹⁵ (1986)	Das and Cohn^16,17 (2001)	Kline⁴ (2004)	Harper²² (2011)	Trützschler Man-made Fibers¹⁹ (2012)	Das and Warren¹⁸ (2012)
Precursor	14.5	8.6	11.1	6.0	12.5	11.5
Pretreatment	1.1	8.6	11.1	n/A	n/A	1.3
Inflation-adjusted
Precursor	31.0	11.4	13.9	6.3	12.7	11.7
Pretreatment	2.3	11.4	13.9	n/A	n/A	1.3

When reviewing the existing models the need for higher transparency in calculation methods, input factors and precursor production costs becomes obvious. The aim of this paper is to introduce a modular model covering the process chain from polymerization to finished carbon fiber, defining a reference process and disclosing all relevant parameterizations.

This will allow researchers as well as producers to analyze sensitivities and to calculate the cost impact of process innovations.

Reference process

For modeling the production cost, a reference process for a 24k PAN-based carbon fiber is defined. The tow size of 24,000 filaments is at the lower range of large tow fibers. It is well suited for applications such as the automotive industry, wind turbines and sports. All process steps and important parameters of the reference process are shown in Figure 6 for precursor production and in Figure 7 for conversion into carbon fiber in the Appendix.

As raw material for the PAN precursor, 95 mol% acrylonitrile (AN), 4 mol% methyl acrylate (MA) and 1 mol% itaconic acid (ITA) are polymerized. In literature there is a variety of compositions. According to Morgan,² at least 85% of AN is needed. Masson proposes a mix of 90–94% AN, 6–9% MA and 0–1% ITA.²⁶ MA is an ideal neutral comonomer for AN due to its similar polarity ensuring a homogeneous polymerisation.² ITA is the most effective carboxylic acid in reducing the exothermicity, more effective than alternatives such as methacrylic acid, acrylic acid and acrylamide.²

Dimethyl sulfoxide (DMSO) is chosen as an organic solvent for the wet spinning process. It has a concentration of 75% in the spinning dope. There are different solvents that can be used for PAN-based precursor spinning. According to Masson,²⁶ carbon fiber market leader Toray¹ is using DMSO. Concentration for DMSO is in the 75–80% range reported by Morgan² and Masson.²⁶

Before spinning, the dope is filtered in a continuous process through a stainless steel woven wire mesh filter with mesh size of 5 µm, as disclosed by the Pall Corporation.²⁷

For wet spinning, the reference process in the cost model is set to a spinning speed of 5 m/min at 80℃. The spinneret nozzles have a diameter of 71 µm. This is consistent with the parameters reported by Morgan² and Masson²⁶ of 3–16 m/min and 50–250 µm. Morgan² indicates the temperature at the spinneret in the wide range of 25–120℃.

For coagulation and washing the DMSO concentration and temperature of the four consecutive baths is set to 60%/20%/10%/1% and 25℃/40℃/70℃/95℃, respectively. Trützschler Man-made Fibers¹⁹ discloses a four-step washing with concentrations of 60%/30%/10%/∼0%. Morgan² reports 0–50℃ for the coagulation bath, Masson²⁶ gives 60% and 45℃ for dimethyl formamide (DMF) as solvent. For the washing bath, a temperature of slightly below 100℃ is the optimum.²⁶

The organic solvent DMSO for the spinning dope is mostly recovered from the coagulation and washing baths. Solvent recovery processes are described by ATOFINA²⁸ and Smallwood.²⁹ First, the DMSO concentration is increased up to 40–70%, for example by evaporation. The second and third steps are fractional distillations in which DMSO and water are split. After distillation about 1% of DMSO remains in the water and can be disposed.²⁸

Stretching is performed in three steps with a total stretching factor of 10 in hot water at a temperature of 95℃. Each single step has a stretching factor of $\sqrt[3]{10} = 2.15$ , which is below a maximum of 3 disclosed by Tanaka and Yamamoto.³⁰ Morgan² reports a maximum total stretching factor of 12 and Masson a range of 3–12.²⁶ The temperature of the water has to be above the wet glass transition temperature (T_g) of about 65℃ according to Masson²⁶ and Morgan² and is the higher the better according to Tanaka and Yamamoto.³⁰

For preparation, the precursor fiber is pulled through a bath, where 2 wt% of silicone oil is added. Morgan² reports that sorbitan esters of long chain fatty acids, polyoxyethylene derivatives or silicones can be used. Tanaka and Yamamoto³⁰ disclose a process where 0.5–2 wt% of silicone oil is added in a bath of 35℃ or less.

The fiber is dried at 180℃, which is in the preferred range of 160–200℃ reported by Tanaka and Yamamoto.³⁰

The PAN-based precursor is finally wound up with a speed of 50 m/min. The line consists of 114 tows and has a nameplate output of 6600 t per year, which is determined to feed a conversion process with a nameplate output of 3000 t per year of carbon fiber. Trützschler Man-made Fibers¹⁹ discloses a range of 40–70 m/min for wet-spun precursors on a typical line with up to 96 tows and 4000 t per year.

The precursor filaments have a linear mass density of 1.0 dtex. This is in the range given by Morgan² (0.8–1.22 dtex), Trützschler Man-made Fibers¹⁹ (1.0–1.6 dtex) and Ise et al.³¹ (0.7–1.0 dtex).

The carbon fiber conversion process follows in a decoupled production line. The PAN-based precursor material is first unwound from non-driven creels and pulled through unheated water to ensure that any dirt from handling and transport is removed.

Stabilization is done in air at 275℃ for 90 min. Three ovens are connected in series to increase heat stepwise to 275℃. This is in line with the oven process described by Morgan² with a series of air heated zone of 220–270℃. Trützschler Man-made Fibers¹⁹ discloses a two-oven process with 200–300℃.

The fiber is carbonized in a 3 min two-step furnace process under nitrogen atmosphere. The first step is at 900℃ and the second is up to 1400℃, according to Morgan.²

Surface treatment is conducted in an electrolytic process where NaOH is used as an electrolyte. This is consistent with what Morgan² describes as the convenient process. The electrolyte should be water soluble.² Goss¹⁵ describes that NaOH is a suitable electrolyte.

An epoxy resin sizing is applied to protect the filaments and improve the fiber-matrix adhesion. Morgan² describes that water-based emulsions are used as sizings and that it is preferred to use the same chemical class as the ultimate polymer matrix.

The finished carbon fiber is collected on a winder with 10 m/min. The conversion consists of two lines with 1500 t per year nameplate capacity and 286 tows each. This is consistent with the range given by Trützschler Man-made Fibers¹⁹ of 2000 t per year as the maximum nameplate line capacity with up to 400 tows and a speed of up to 12 m/min.

Method

The new carbon fiber cost model has a modular design with independently configurable process steps. The model covers the process chain from polymerization to winding of finished carbon fiber, ensuring a holistic cost perspective on carbon fiber manufacturing.

The modular model allows analyzing changes to the defined reference process regarding variations of input parameters and adaptations to the process chain. Each module can be replaced or taken out in case cost impact, for example for alternative thermal conversion of PAN precursors or new precursors such as lignin, is quantified.

Process steps for PAN-based carbon fiber production are deducted from the reference process that has been defined based on the literature, patents and knowledge at the Institut für Textiltechnik of RWTH Aachen University. This ensures that all process steps are included and important parameters are made transparent.

As shown in Figure 1, the model includes all relevant types of manufacturing costs. These are costs for raw material, process material, direct labor and labor overhead, capital costs for equipment and the plant building, maintenance, insurance and taxes.³² Manufacturing costs do not include costs for research and development, procurement or distribution.³³

Figure 1.

Modular cost model concept.

Input factors are split into those that are general and those that are process step (module) specific. General factors are shown in Table 5. They include data for capacity, labor, process material, energy, capital costs, building invest, life expectancy for equipment and building and other costs.

Table 5.

General input parameters

Category	Parameter	Unit
Capacity	Production days	days/year
	Process availability	%
	Filaments per tow	k
	Precursor nameplate output	kg/year
	Conversion nameplate output	kg/year
	Precursor lines	#
	Conversion lines	#
	Precursor tows per line	#
	Conversion tows per line	#
	Precursor filament	dtex
	Precursor stretching factor	#
	Conversion stretching factor	#
	Precursor end line speed	m/min
	Precursor spinning speed	m/min
	Precursor spinneret diameter	µm
Labor	Working days	days/year
	Shift time	hours/shift
	Shifts	shifts/day
	Man–year line worker	EUR/year
	Man–year line engineer	EUR/year
	Overhead	%
Raw material	Acrylonitrile (AN)	EUR/kg
Raw material	Methyl acrylate (MA)	EUR/kg
	Itaconic acid (ITA)	EUR/kg
	Percentage of weight AN	%
	Percentage of weight MA	%
	Percentage of weight ITA	%
Process material	Process water	EUR/kg
Process material	Dimethyl sulfoxide (DMSO)	EUR/kg
	Nitrogen	EUR/m³
	Acetone	EUR/kg
	Silicone oil agent	EUR/kg
	Sizing resin	EUR/kg
	NaOH	EUR/kg
	Packaging precursor	EUR/unit
	Packaging carbon fiber	EUR/unit
Energy	Electricity	EUR/kWh
Capital	Weighted average cost of capital	%
Building	Space invest	EUR/m²
Life expectancy	Equipment	years
Life expectancy	Building	years
Other	Maintenance	% of invest
	Insurance	% of invest
	Property taxes	% of invest

Table 6 shows specific input factors by process step. As they vary for each process step represented by an independent module, these factors have to be defined individually. Yield is split into chemical and mechanical, where chemical yield represents the material losses or gains due to conversion processes and mechanical yield is a factor including waste due to process instabilities. Energy consumption is split into basic demand and an efficiency factor depending on the equipment involved. Raw material for polymerization and process materials required in the production steps are separated. Energy and material consumptions are measured per mass of material input into the process step. This ensures that these parameters are independent from yield in the process step.

Table 6.

Specific input parameters by module

Category	Parameter	Unit
Energy	Electricity	[kWh/kg]
	Electricity efficiency	[%]
Raw material	Acrylonitrile (AN)	[kg/kg]
	Methyl acrylate (MA)	[kg/kg]
	Itaconic acid (ITA)	[kg/kg]
Process	Process water	[l/kg]
material	Dimethyl sulfoxide (DMSO)	[kg/kg]
	Nitrogen	[m³/kg]
	Silicone oil agent	[kg/kg]
	Sizing resin	[kg/kg]
	NaOH	[kg/kg]
	Precursor packaging	[Units/kg]
	Carbon fiber packaging	[Units/kg]
Labor	Direct workers	[Workers per line]
	Direct engineers	[Engineers per line]
Capital	Equipment invest	[Mio. EUR/line]
	Plant surface	[m²/line]

Capital cost for equipment and building are calculated as annuity over the expected lifetime:

Annualcost = Invest \times \frac{r \times (1 + r) n}{(1 + r) n - 1}

where r is the weighted average capital cost (WACC) and n is the lifetime.³⁴

Sensitivity of carbon fiber costs on energy price, AN price, WACC and process availability is analyzed.

Results

Table 7 shows the assumed values for all general input parameters. Without doubt this parameterization can only be a snap-shot of current industry averages and has to be adapted to the individual location.

Table 7.

Assumptions for general parameters

Category	Parameter	Unit	Assumption
Capacity	Production days	days/year	292
	Process availability	%	80%
	Filaments per tow	k	24
	Precursor nameplate output	kg/year	6,601,355
	Conversion nameplate output	kg/year	3,000,000
	Precursor lines	#	1
	Conversion lines	#	2
	Precursor tows per line	#	114
	Conversion tows per line	#	286
	Precursor filament	dtex	1.0
	Precursor stretching factor	#	10
	Conversion stretching factor	#	1
	Precursor end line speed	m/min	50
	Precursor spinning speed	m/min	5
	Precursor spinneret diameter	µm	71
Labor	Working days	days/year	230
	Shift time	hours/shift	8
	Shifts	shifts/day	3
	Man–year line worker	EUR/year	48000
	Man–year line engineer	EUR/year	100,000
	Overhead	%	60%
Raw material	Acrylonitrile (AN)	EUR/kg	1.70
Raw material	Methyl acrylate (MA)	EUR/kg	1.50
	Itaconic acid (ITA)	EUR/kg	1.60
	Percentage of weight AN	%	95%
	Percentage of weight MA	%	4%
	Percentage of weight ITA	%	1%
Process material	Process water	EUR/kg	0.002
Process material	Dimethyl sulfoxide (DMSO)	EUR/kg	0.92
	Nitrogen	EUR/m³	0.05
	Acetone	EUR/kg	0.80
	Silicone oil agent	EUR/kg	3.00
	Sizing resin	EUR/kg	4.00
	NaOH	EUR/kg	0.35
	Packaging precursor	EUR/unit	15.00
	Packaging carbon fiber	EUR/unit	0.40
Energy	Electricity	EUR/kWh	0.080
Capital	Weighted average cost of capital	%	7.5%
Building	Space invest	EUR/m²	1100
Life expectancy	Equipment	years	25
Life expectancy	Building	years	40
Other	Maintenance	% of invest	3%
	Insurance	% of invest	0.5%
	Property taxes	% of invest	1.0%

All assumptions for process step specific parameters are shown in Table 12 in the Appendix.

Yield by process step is shown in Table 8. In precursor production the chemical yield is >100% because ∼3% of DMSO remains in the filaments after washing and ∼2% of weight is added during surface preparation with silicone oil. For conversion steps the overall chemical yield is parameterized with 48%. Combined with the mechanical yield the conversion process shows a total yield of 45.4%. This means that 2.2 kg of precursor material is needed to produce 1 kg of good carbon fiber material in the model.

Table 8.

Yield by process step in percent

Process step	Chemical	Mechanical	Total yield
Subtotal precursor	104.5	94.6	98.9
Polymerization & dope prep.	399.0	99.5	397.0
Filtration	99.0	100.0	99.0
Spinning	100.0	99.0	99.0
Coagulation	62.5	99.5	62.2
Washing	45.5	99.5	45.3
Stretching	100.0	99.0	99.0
Preparation	102.0	99.5	101.5
Drying	91.2	99.5	90.7
Winding/handling	100.0	99.0	99.0
Subtotal conversion	48.0	94.7	45.4
Pretreatment	100.0	98.7	98.7
Stabilization	96.0	98.7	94.8
Carbonization	49.0	98.7	48.4
Surface treatment	100.0	99.5	99.5
Sizing	102.0	99.5	101.5
Winding/handling	100.0	99.5	99.5
Total	50.1	89.6	44.9

The total cost of carbon fiber manufacturing is calculated bottom-up at EUR 19.64 per kg. Table 9 shows the absolute values split by process step and cost type. This contains EUR 10.54 for precursor material and EUR 9.10 for conversion from precursor into carbon fiber.

Table 9.

Costs overview by process step and cost type in EUR/kg carbon fiber

In EUR/kg carbon fiber
Process step	Raw material	Process materials	Labor	Equipment	Building	Energy	Maintenance	Insurance	Taxes	Total
Subtotal precursor	3.76	0.94	0.43	1.38	0.11	3.18	0.50	0.08	0.17	10.54
Polymerization & dope preparation	3.76	0.38	0.14	0.34	0.03	0.00	0.13	0.02	0.04	4.86
Filtration	–	–	0.04	0.07	0.01	0.05	0.03	0.00	0.01	0.20
Spinning	–	–	0.05	0.09	0.00	0.06	0.03	0.01	0.01	0.25
Coagulation	–	0.00	0.05	0.03	0.00	0.14	0.01	0.00	0.00	0.26
Washing	–	0.17	0.07	0.58	0.03	2.48	0.21	0.03	0.07	3.65
Stretching		–	0.02	0.10	0.01	0.30	0.04	0.01	0.01	0.49
Preparation	–	0.14	0.02	0.03	0.00	0.01	0.01	0.00	0.00	0.23
Drying	–	–	0.02	0.05	0.00	0.12	0.02	0.00	0.01	0.22
Winding/handling	–	0.23	0.02	0.07	0.01	0.01	0.03	0.00	0.01	0.38
Subtotal conversion	–	0.78	0.91	2.20	0.41	3.50	0.87	0.15	0.29	9.10
Pretreatment	–	0.00	0.06	0.07	0.03	0.02	0.03	0.01	0.01	0.23
Stabilization	–	–	0.28	0.69	0.18	0.75	0.29	0.05	0.10	2.34
Carbonization	–	0.58	0.28	0.89	0.06	1.94	0.32	0.05	0.11	4.23
Surface treatment	–	0.02	0.06	0.21	0.05	0.31	0.08	0.01	0.03	0.76
Sizing	–	0.08	0.06	0.14	0.05	0.32	0.06	0.01	0.02	0.73
Winding/handling	–	0.10	0.17	0.21	0.05	0.16	0.08	0.01	0.03	0.80
Total	3.76	1.72	1.33	3.58	0.52	6.68	1.37	0.23	0.46	19.64

Raw material costs of AN, MA and ITA represent the largest cost component. They account for EUR 3.76 of the precursor costs. Energy is the second with EUR 3.50 for conversion and EUR 3.18 for precursor production.

Table 10 shows the cost breakdown in percent of the total carbon fiber manufacturing costs. The model results in a cost component for PAN precursor of 54% and 46% for conversion into carbon fiber.

Table 10.

Costs overview by process step and cost type in percent of carbon fiber costs

In % of carbon fiber costs
Process step	Raw material	Process materials	Labor	Equipment	Building	Energy	Maintenance	Insurance	Taxes	Total
Subtotal precursor	19.2	4.8	2.2	7.0	0.6	16.2	2.6	0.4	0.9	53.7
Polymerization & dope preparation	19.2	2.0	0.7	1.8	0.2	0.0	0.6	0.1	0.2	24.7
Filtration	–	–	0.2	0.4	0.0	0.3	0.1	0.0	0.0	1.0
Spinning	–	–	0.3	0.4	0.0	0.3	0.2	0.0	0.1	1.3
Coagulation	–	0.0	0.3	0.2	0.0	0.7	0.1	0.0	0.0	1.3
Washing	–	0.9	0.4	3.0	0.1	12.6	1.1	0.2	0.4	18.6
Stretching	–	–	0.1	0.5	0.1	1.5	0.2	0.0	0.1	2.5
Preparation	–	0.7	0.1	0.2	0.0	0.0	0.1	0.0	0.0	1.2
Drying	–	–	0.1	0.3	0.0	0.6	0.1	0.0	0.0	1.1
Winding/handling	–	1.2	0.1	0.4	0.0	0.1	0.1	0.0	0.0	1.9
Subtotal conversion	–	4.0	4.6	11.2	2.1	17.8	4.4	0.7	1.5	46.3
Pretreatment	–	0.0	0.3	0.4	0.2	0.1	0.2	0.0	0.1	1.2
Stabilization	–	–	1.5	3.5	0.9	3.8	1.5	0.2	0.5	11.9
Carbonization	–	2.9	1.5	4.6	0.3	9.9	1.6	0.3	0.5	21.5
Surface treatment	–	0.1	0.3	1.1	0.2	1.6	0.4	0.1	0.1	3.9
Sizing	–	0.4	0.3	0.7	0.2	1.6	0.3	0.1	0.1	3.7
Winding/handling	–	0.5	0.9	1.1	0.2	0.8	0.4	0.1	0.1	4.1
Total	19.2	8.7	6.8	18.2	2.6	34.0	7.0	1.2	2.3	100.0

Overall, energy makes the highest cost component with about 34%. Second is the precursor raw material costs with about 19% and third the capital costs for equipment with about 18%.

The process step with the highest cost component (25%) is the polymerization and dope preparation in the beginning of precursor production. Here, raw material cost is the main driver. Second is the carbonization step (22%) during carbon fiber conversion. High energy consumption for the furnace, nitrogen as inert gas and high equipment investment can be identified as main causes. Third is the fiber washing after coagulation (19%) in the precursor wet-spinning process with high energy consumption for solvent recovery and heating of baths, as well as high investment for the solvent recovery unit causing significant costs.

Carbon fiber manufacturing costs have been analyzed for sensitivities. The energy price shows a high impact with EUR .83 per kg in carbon fiber cost per EUR .01 per kWh change in energy price. As the range of industrial energy prices worldwide varies significantly, this should be considered in the choice of site location. The results are shown in Figure 2. Standard parameterization in the model is EUR .08 per kWh.

Figure 2.

Sensitivity to energy price.

AN price has an influence of EUR .02 per kg in carbon fiber cost per EUR .01 per kg change in AN price. As the AN price is strongly correlated to oil price, a cost pressure can be expected in the future from this raw material. Carbon yield is assumed to be stable in the sensitivity analysis according to data in Table 9. The results are shown in Figure 3. Standard parameterization in the model is EUR 1.70 per kg.

Figure 3.

Sensitivity to acrylonitrile price.

The plant availability has an influence of EUR .08 per kg in carbon fiber cost per 1% change in availability. This can be equalized to capacity utilization, as only fixed costs occur during downtimes in the model. The results are shown in Figure 4. Standard parameterization in the model is 80% availability.

Figure 4.

Sensitivity to availability.

Capital costs for equipment and plant building depend on capex and the yearly interest rate. The latter is assumed to be the producers’ WACC in the model, which has an influence of EUR .37 per kg in carbon fiber cost per 1% change in WACC. The results are shown in Figure 5. Standard parameterization in the model is 7.5% WACC.

Figure 5.

Sensitivity to weighted average capital cost (WACC).

Conclusions and discussion

The cost structure in the presented model is compared to the reviewed models in Table 11. The high share of precursor costs in recent models can also be observed in the presented model. The low cost for PAN precursor in the model by Harper International²² is not reproduced. While Harper has USD 6 per kg carbon fiber, the presented model comes to the result of EUR 10.54 per kg, which is more than double. Indications that the cost effect of carbonization has been underestimated in the model of Das and Warren¹⁸ can been confirmed.

Table 11.

Comparison of results

	Goss¹⁵ (1986)	Das and Cohn^16,17 (2001)	Kline⁴ (2004)	Harper²² (2011)	Trützschler Man-made Fibers¹⁹ (2012)	Das and Warren¹⁸ (2012)	This paper (2013)
Precursor	34%	47%	51%	31%	51%	54%	54%
Pretreatment	3%	47%	51%	69%	49%	6%	1%
Stabilization	20%	16%	16%			13%	12%
Carbonization	23%	26%	23%			8%	21%
Surface treatm.	4%	2%	4%			3%	4%
Sizing	7%	2%	4%			4%	4%
Winding	5%	3%	6%			9%	4%
Other	4%	4%	0%			3%	0%

Due to missing information about parameterization, a detailed discussion about the reviewed models is impossible. The presented model with all input parameters disclosed shall support a more fact-based discussion about carbon fiber cost structure in the future.

A strong sensitivity of carbon fiber price to energy price has been identified. Cost optimization should consider the choice of site with low energy prices and/or process optimization for lower energy consumption.

Outlook

The model will be extended to new precursor materials as a substitute to PAN. The deviating production processes will be modeled with new modules. This will allow measuring the impact of the alternate raw materials along the whole process chain by cost type.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Appendix

References

Sloan J. Market outlook: surplus in carbon fiber's future? High Perform Compos 2013; 21(2): 46–52.

Morgan

. Carbon fibers and their composites, Boca Raton, FL: Taylor & Francis, 2005.

Carson EG. The future of carbon fibre to 2017. In: gocarbon conference, London, 2012.

Kline & Company. Cost assessment of a new carbon fiber manufacturing process. Unpublished draft report submitted to Automotive Composites Consortium, March 2004.

Heuss R, Müller N, van Sintern W, et al. Lightweight, heavy impact: how carbon fiber and other lightweight materials will develop across industries and specifically in automotive, 2012 McKinsey & Company, Advanced Industries, February 2012. http://www.mckinsey.com/~/media/mckinsey/dotcom/client_service/Automotive%20and%20Assembly/PDFs/Lightweight_heavy_impact.ashx (accessed 25 September 2014).

Warren DC. Low cost carbon fiber overview. Oak Ridge National Laboratory, Oak Ridge, Tennessee, May 2011.

Cytec Industries Inc. 2012 annual report. Woodland Park, USA, 2013.

Hexcel Corporation. Annual report 2012. Stamford, USA, 2013.

Mitsubishi Rayon Co., Ltd. 2010 annual report. Tokyo, Japan, 2010.

10.

SGL Carbon SE. Geschäftsbericht 2012. Wiesbaden, Germany, 2013.

11.

Toho Tenax Co., Ltd. 2007 annual report. Tokyo, Japan, 2007.

12.

Toray Industries, Inc. 2012 annual report. Tokyo, Japan, 2012.

13.

Zoltek Companies, Inc. 2012 annual report. St. Louis, USA, 2012.

14.

Poggiali B. Production cost modeling: a spreadsheet methodology. Master's Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1985.

15.

Goss I. Production functions and cost models for fibers in advanced composite materials. Master's Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1986.

16.

Das S. The cost of automotive polymer composites: a review and assessment of DOE's lightweight materials composites research. Energy Division of Oak Ridge National Laboratory, Oak Ridge, TN, January 2001.

17.

Madia WJ. Microwave-assisted plasma carbon fiber manufacturing: 2001 R&D 100 award entry form. Oak Ridge National Laboratory, Oak Ridge, TN, 2001.

18.

Das S and Warren D. Technical cost modeling -life cycle analysis basis for program focus. Oak Ridge National Laboratory, Oak Ridge, TN, May 2012.

19.

Trützschler Man-made Fibers. New prospects for the manufacturing of carbon fibers. Dresden, 2012.

20.

Blackmon RM and Carmina K (eds). Low cost carbon fiber: on the verge of realization, SAMPE 2013 proceedings, Long Beach, CA, 6–9 May 2013, pp.363–386.

21.

Blackmon RM. Cost Structure and process efficiency for carbon fibre conversion. Harper International, Low Cost Carbon Fibre for Massive Automotive Application, Liege, November 2012.

22.

Harper International. Carbon fiber line capacity: historical perspective. Harper International, GO Carbon Fiber/IntertechPira, 2009.

23.

United States Department of Labor. Consumer price index from 1913 to present, http://www.bls.gov/cpi/#tables (table 24, 2014, accessed 25 September 2014).

24.

Eurostat. Electricity prices for medium size industries, http://epp.eurostat.ec.europa.eu/tgm/refreshTableAction.do?tab=table&plugin=1&pcode=ten00117&language=en (2014, accessed 25 September 2014).

25.

Rix J. Blog: inside a carbon-fibre factory: we get a look around BMW’s new project ‘i’ carbon-fibre plant in Washington, http://www.autoexpress.co.uk/bmw/61595/blog-inside-a-carbon-fibre-factory (2012, accessed 19 June 2013).

26.

Masson

. Acrylic fiber technology and applications, New York: M. Dekker, 1995.

27.

Pall Corporation. Carbon fiber and pan precursor production, http://www.pall.com/pdfs/Fuels-and-Chemicals/PPG-5b.pdf (2005, accessed 16 July 2013).

28.

ATOFINA. DMSO recycle product information, www.arkema-inc.com/literature/pdf/358.pdf (n.d., accessed 17 July 2013).

29.

Smallwood

. Solvent recovery handbook, 2nd ed. Boca Raton, FL: Blackwell Science; CRC Press, 2002.

30.

Tanaka F and Yamamoto Y. Oil agent for precursor fiber of carbon fiber, carbon fiber and production method of the carbon fiber. Patent 2008/0152574 A1, USA, 2008.

31.

Ise M, Nakayama I and Endi M. Carbon fiber, process for production of polyacrylonitrile-base precursor fiber for carbon fiber production, and process for production of carbon fiber. Patent 8,137,810 B2, USA, 2012.

32.

Männel

. Handbuch Kostenrechnung, Wiesbaden: Gabler, 1992.

33.

Eichmann

. Industrielle Fertigungskostenrechnung: Erfassung und Verrechnung der Fertigungskosten des Industriebetriebs, Lohmar: Eul, 2001.

34.

Neufville R de, Clark JP and Field FR III. Introduction to technical cost modeling concepts and illustrations. MIT Vorlesungsskript, Cambridge, MA, 2013.