A Proposal for a Carbon Fibre-Manufacturing Life-Cycle Inventory: A Case Study from the Competitive Sailing Boat Industry

Abstract

The competitive sailing boat industry uses carbon fibre for high-performance purposes. Nevertheless, this material is known to cause environmental issues during its manufacturing. We can currently observe, based on the literature, difficulty integrating a reliable, justified, and transparent inventory of carbon-fibre production for LCA applications of high-performance composite materials. The current study aims to gain a better understanding of carbon fibre’s environmental impacts by suggesting a justified, reliable, and transparent inventory, based on the life-cycle assessment methodology. It also aims at providing a LCA of high-performance composites. An EcoInvent flows inventory is suggested, based on the literature presenting primary inventories. It is then discussed in terms of data quality, flows under study, and indicators calculated. Eventually, the inventory is used to assess the environmental impact of carbon fibre-reinforced composites applied to an industrial example representative of the competitive sailing boat industry: a hydrofoil mould. Regarding results on carbon fibres’ scale and impacts, indicators commonly highlighted by the literature, were calculated in this study (GWP = 72 kgCO₂eq and CED = 1176 MJ), as well as other indicators. These indicators are two to five times higher than the inventories suggested in the literature, due to high heat-production value, production scales, or the quality of the fibre under study. The composite scale results show a major contribution from carbon fibre compared to other flows under study, highlighting the need to suggest a reliable inventory of carbon-fibre production.

1. Introduction

In the framework of Paris Agreements [1], it is stated that industrial fields like transport have to reduce their greenhouse gas emissions. Carbon fibre exhibits a weight-saving potential for composite structures due to their highly specific mechanical properties [2]. Hence, their current use can be included in the environmental context for transport applications by reducing the contribution of the use-phase burdens [3].

The competitive sailing boat sector uses carbon fibre as well, for high-performance purposes. Competition rules tend to include environmental aspects in this industry. For example, the International Monohull Open Class Association (IMOCA) and the America’s Cup impose performance of a life-cycle assessment (LCA) on all new sailing boats. Carbon fibre-reinforced plastic (CFRP) moulds have been identified as a strategic boat component to work on to reduce the sector’s environmental impacts.

The carbon-fibre manufacturing process is divided into two phases: the manufacture of the precursor and of the fibre [4]. Here, the PolyAcryloNitrile (PAN) precursor is studied, as it is the most widely used. PAN precursor production is first based on the polymerisation of acrylonitrile [5], obtained from the Sohio process [6]. Then, the so-called dope solution at the end of this step is transferred to a solution containing a solvent and non-solvent through a spinneret. It then undergoes spinning, stretching, and finishing.

Starting from PAN coils, carbon-fibre line production undergoes the following steps. The first step is stabilisation between 200 °C and 300 °C in an oxygen atmosphere. Then, a carbonisation step occurs where PAN fibres are submitted to heating in furnaces between 500 °C and 1600 °C under an inert atmosphere (nitrogen). A temperature of 1600 °C is used for high-strength fibres. If higher-modulus fibres are to be created, one can heat them up to 3000 °C. Finally, the fibres can then undergo a sizing (surface treatment) operation to improve their retention in subsequent processes and their adhesion to the matrix.

All these steps have their own environmental burdens and contribute to a whole environmental impact calculation. Currently, a life-cycle assessment (LCA) is the most effective tool to quantify the environmental impact of products and services following ISO 14044 and 14040 standards. European guides such as the International Life Cycle Data (ILCD) system [7] or the Product Environmental Footprint (PEF) [8] have been created to help spread the methodology. Environmental flows such as electricity, wastes, or materials have to be retrieved throughout the methodology. These flows can then be converted into environmental impacts, with uncertainty depending on the data quality, amongst other choices [9]. Two data typologies can be found across an LCA: primary and secondary data. Primary data are experimental data. Secondary data are data evaluated via calculations, expert estimations, or documentation within the literature. Primary data allow calculation of more accurate results than secondary data.

Two observations can be made based on the results calculated by studies performing LCA on carbon fibre:

• Results related to environmental indicators presented by these studies differ greatly. For example, a 1997 technical report from De Vegt and Haije [10] reported 8 MJ/kg of carbon fibre produced, whereas recent studies indicated around 5600 MJ/kg [4];
• Studies mainly focus on two environmental indicators, energy demand and global warming potential (GWP), while LCA practitioners should maintain several indicators to avoid impact transfers [7].

These two issues are linked to inventories used to perform studies. Indeed, environmental impact gaps are related to flow quantities (i.e., inventories set) under study. Moreover, a multicriteria analysis could be performed based on an inventory used in LCA software. Some studies have suggested aggregated inventories. First, some studies focused on PAN or carbon-fibre production. For example, a recent study from Groetsch et al. [11] has retrieved data experimentally from a carbon-fibre production line, at a laboratory scale. The same author suggested a LCA hypothesis to be set for the analysis of carbon fibre in 2021, for an energy-demand analysis related to a carbon-fibre line [12]. Focusing on articles including environmental impacts of both precursors and carbon-fibre lines, several observations can be made. Inventories are based on articles from the literature. Studies aggregate data and suggest an inventory. This inventory is then applied to calculate environmental impacts to use in a case study, from which arise several issues such as arbitrary choice of the practitioner and higher complexity to evaluate the reliability of inventories. For example, in 2020 Forcellese et al. [13] evaluated impacts related to carbon-fibre manufacturing. This study is based on data from Duflou et al. [5], for PAN manufacturing, and data from Khalil [14] for a carbon-fibre production line. Khalil’s study aggregates several studies but does not give information on the way they made it. Most of the time, case studies are linked to transport or composites in general. For example, Johnson and Sullivan [15] studied carbon-fibre production through a technical report for automotive applications. Scelsi et al. [16] used a carbon-fibre inventory in a case study related to aircraft. Nevertheless, they did not present their inventory. Several issues can be drawn from studies presenting inventories:

First, it is hard to evaluate the extent to which inventories are reliable. Indeed, there is a lack of clarity on the way to aggregate data, as well as data sources. This makes it harder to select an inventory to use in a LCA or evaluate a result’s reliability;

Several methodological choices are still left to the LCA practitioner. For example, data used in environmental databases to perform a study are never highlighted, nor are impacts kept for analysis. This issue affects the ease with which one can perform a LCA and calculate comparability across results;

Finally, none of these studies used their inventories in a case study related to high performance. This type of study could help some industrial sectors like Formula One or competitive sailing boats to include environmental aspects in their practices.

These observations raise a need to suggest a justified, reliable, and transparent inventory in environmental analysis of carbon fibre.

The current study aims to reveal new insights regarding carbon fibre’s environmental impacts based on primary data from literature analysis so that an EcoInvent flows inventory is provided. The suggested inventory is then critically discussed in terms of data quality, while multicriteria analysis following the PEF framework is performed. Finally, the inventory is used to assess the environmental impact of carbon fibre-reinforced plastic (CFRPs) applied to an industrial example representative of the competitive sailing boat industry: a hydrofoil mould.

2. Materials and Methods

Results are presented at two scales: carbon fibre and composite structure. In this section, a hypothesis, materials, and methods related to the composite structure are first presented; then, the hypothesis linked to the study of carbon-fibre manufacturing is presented.

2.1. Materials

A hydrofoil mould manufactured by the company Avel Robotics (Lorient, France) is presented in Figure 1. The hydrofoil mould here presented is not the one under study, for confidentiality purposes. In the current case, data are taken from the manufacture of a hydrofoil mould made of carbon-fibre/epoxy composite, by the company SMM Composites (Lanester, France). First, a polystyrene master is made and placed on a metal frame, which will infuse the composite material that will make up the mould. This carbon skin is then grafted onto a plywood frame and is then removed from the polystyrene master to serve as the foil mould. Two carbon-fibre references have been used for the current hydrofoil, from Zoltek (Bridgeton, MO, USA): X-C-606-1270 mm 50 K and B-C-606-1270 mm 50 K.

2.2. Methodology Set to Study a Composite Structure: The Hydrofoil Mould

The hydrofoil mould is studied based on the LCA methodology. The functional unit is ensuring geometry during the manufacture of a pair of hydrofoils for 4 years. The duration is based on a functional unit suggested for a competitive sailing boat, which represents here the “Vendée Globe” duration. The hypothesis is made that the hydrofoil mould does not fail during this period. The associated reference flow is 1 epoxy/carbon laminate skin to be grafted onto a plywood frame. The vacuum-infusion process under study is shown in Figure 2 as well as the various flows identified with the manufacturer. Several hypotheses have been implemented:

• Inventories not directly related to the quality of the laminate are excluded, such as electricity used to power computers in the factory;
• Consumables, tools, and machinery are excluded;
• VOC emissions have been excluded as they are not measurable;
• Hardeners were excluded from the study as they were not available in the databases;
• Transport has been neglected;
• The manufacture of the master is excluded, to represent a general case where composite parts are manufactured in large series of the same master;
• Weaving of carbon fibres were excluded from the study;
• The two carbon fibre references have been modelled with the same process, as the current literature does not allow this level of detail.

These hypotheses tend to reduce the overall environmental impacts calculated. The last hypothesis could only reduce the environmental impacts of this analysis, as better-quality carbon fibre should have higher impacts than lower-quality ones. Moreover, it should be noted that these hypotheses might reduce the accuracy of the modelling, but it is still a good starting point based on information available.

The “cutoff 0:100” allocation rules were implemented to model the waste generated during the infusion process.

The background processes were modelled using the EcoInvent v3.8 database. Impacts were calculated using OpenLCA v1.10.3 software. All impacts suggested in the method have been presented. Then, three indicators, suggested as the most reliable by the PEF guidelines [8], were kept for a hot-spot analysis:

• Global warming potential: Capacity of a greenhouse gas to influence radiative forcing, expressed in terms of a reference substance (for example, CO₂-equivalent units) and specified time horizon (e.g., GWP 20, GWP 100, and GWP 500, for 20, 100, and 500 years, respectively). It relates to the capacity to influence changes in the global average surface-air temperature and subsequent changes in various climate parameters and their effects, such as storm frequency and intensity, rainfall intensity, frequency of flooding, etc.;
• Ozone depletion: EF impact category that accounts for the degradation of stratospheric ozone due to emissions of ozone-depleting substances, for example long-lived chlorine and bromine-containing gases (e.g., CFCs, HCFCs, and Halons);
• Particulate matter: EF impact category that accounts for the adverse health effects on human health caused by emissions of particulate matter (PM) and its precursors (NO_x, SO_x, and NH₃).

For confidentiality purposes, the inventory used in this study will not be presented, as it is not possible to generate an aggregated inventory in this study. Nevertheless, a carbon-fibre inventory is needed to study this foil mould.

2.3. Methodology Set to Suggest a Carbon-Fibre Inventory

This study is based on the LCA methodology, applied to carbon fibre’s manufacturing process.

2.3.1. Goal and Scope of the Study

The goal of this study is to perform an environmental analysis of the carbon-fibre manufacturing process, based on the current literature. As little information is available about fibre quality in the literature, no functional unit is suggested here. Instead, only a reference flow is studied. The latter is “1 kg of PAN-based carbon fibre”. Boundaries are set to embrace the precursor and fibre manufacturing. As this analysis is intended to be used in a complete LCA of a composite product, neither end-of-life nor usage are considered. No cutoff rules are applied, as this study is intended to aggregate several experimental LCIs. No allocation rules are implemented, as there were no co-products or multifunctionality to take into account with the selected inventories.

2.3.2. Life-Cycle Inventory

The inventory is suggested using current primary and usable literature. This literature has been retrieved by a snowballing method [17] described in a conference paper from Véronique Michaud [18]. Dissertations were not kept for analysis, and only papers presenting LCAs including PAN production and carbon-fibre line production were kept. The articles retrieved from the literature were then sorted as to whether they exhibit primary or secondary data. Articles based on primary data were retained to suggest an inventory and sorted across their usability. These articles are presented in Section 3.1. Articles presenting secondary data were kept for discussion of the environmental impacts calculated in this article.

Each flow presented by these studies was used in an aggregated inventory with the following method. As it was not possible to justify the flow quantities under study, flows have been aggregated in the following manner, based on the methodology suggested by Henriksson et al. [19]. When a flow is suggested once across the studies, its value is kept in the inventory. When two or more flows are suggested, they are averaged in order to create an aggregated value. This implies that the inventory suggested will consist of only two reference studies and that no other flows could be added to the analysis.

Each flow suggested here was modelled within the EcoInvent v3.8 database. Providers from this database are suggested, to help non-experts use the inventory.

2.3.3. Life-Cycle Impact Assessment

The LCIA is performed with OpenLCA, open-source software. In accordance with the PEF guidelines [8], the impact calculation is made with the EF 3.0 method (adapted). All impacts suggested in the method have been presented. Then, impacts indicated as most reliable in the method (and presented in the hydrofoil section) have been selected for a hot-spot analysis added with the cumulative energy demand (CED), which is a commonly used environmental indicator in carbon-fibre studies.

2.3.4. Interpretation of the Results

A hot-spot analysis was performed based on the results derived from the current inventory to identify principal flows responsible for the environmental impacts.

Results of this study at the fibre scale (LCI and environmental indicators) were then discussed based on two analyses regarding aspects that affect greatly the LCA results’ reliability and a comparison between studies. The comparison was made based on the literature from group 2. First, a data quality comparison was performed, based on the data quality rating (DQR) indicator [8]. Studies used in this inventory were compared with other studies based on primary data. Then, a comparison of flows under study was performed. Flows retrieved by studies suggesting primary data were compared with the ones used in this study.

Finally, a comparison of indicator values from the literature was performed. As GWP and CED are often characterised by studies, these two indicators were compared with studies presenting primary and secondary data.

3. Results

3.1. Primary Data Found in the Literature Associated with Carbon-Fibre Production

The literature (academic and technical reports) presenting primary data can be found in Table 1. The literature can be divided into three groups:

Group 1.

Environmental indicators are presented and are based on primary data. They do not give details of the proposed inventories or the assumptions involved in creating the inventory (scope, data collection methods, etc.). These articles have been retained to compare their environmental indicators with the ones from the current study;

Group 2.

More details are given on how their inventories are collected, despite providing few details on the quantification of flows. For example, a recent study by the Japan Carbon Fibre Manufacturers Association (JCMA) [20] joined forces to propose an LCA representative of the Japanese market, in terms of cumulative energy demand (CED), but the quantity of kWh involved with producing precursors or carbon fibre is not provided. These articles are selected to discuss the proposed inventory in terms of data quality and flows studied. Group 2 consists of four references: EcoImpactCalculator (Europe) [21], EcoImpactCalculator (ROK) [22], JCMA [20], and Zhang, 2024 [23]. The first two references relate to simplified LCA software for composite materials. The tool was first developed in Europe, then adapted for use in South Korea. These references are referred to as EUClA and EUClA (ROK). Zhang’s work in 2024 [23] led to an environmental analysis of the fibre manufacturing process in China. This work has been conducted with one of the top carbon-fibre industry chain manufacturers in the country;

Group 3.

Inventories exhibit a usable form, i.e., the quantities of the flows involved are provided, and the assumptions used to determine them are set out, at least in part. These articles are used to suggest an inventory can be used for carbon-fibre manufacturing. The Das [24] and Duflou [5] studies, although old, are currently reference studies on the environmental aspects of carbon fibre. They are the only two studies that present inventory data in a usable form at the present time.

Table 1. Studies from the literature presenting indicators based on experimental inventories from industry or laboratories.

Type of Literature	Author	Year	Group	GWP (kg CO2eq)	Energy Demand (MJ)
Academic literature	[25]	1999	1	-	478
	[26]	2004	1	-	286
	[5]	2009	3	-	-
	[24]	2011	3	31	704
	[27]	2016	1	-	735.20
	[4]	2021	1	-	5586.8
	[23]	2024	2	67.79	-
Technical reports	[28]	2017	1	-	869.95
	[21]	2020	2	-	1041
	[22]	2023	2	-	1046
	[20]	2022	2	-	350.2

An inventory regarding carbon-fibre manufacturing is presented in the following section, based on articles from group 3.

Table 1. Studies from the literature presenting indicators based on experimental inventories from industry or laboratories.

Type of Literature	Author	Year	Group	GWP (kg CO2eq)	Energy Demand (MJ)
Academic literature	[25]	1999	1	-	478
	[26]	2004	1	-	286
	[5]	2009	3	-	-
	[24]	2011	3	31	704
	[27]	2016	1	-	735.20
	[4]	2021	1	-	5586.8
	[23]	2024	2	67.79	-
Technical reports	[28]	2017	1	-	869.95
	[21]	2020	2	-	1041
	[22]	2023	2	-	1046
	[20]	2022	2	-	350.2

An inventory regarding carbon-fibre manufacturing is presented in the following section, based on articles from group 3.

3.2. Carbon-Fibre Manufacturing Inventory

Table 2. Inventory suggested, based on current experiments described in the literature, from Das [24] and Duflou et al. [5]. Five EcoInvent flows are suggested, linked to their providers in the database. Flows are steam, nitrogen, heat, electricity, acrylonitrile, and vinyl acetate.

EcoInvent Flow	Unit	[24]	[5]	Suggested Value	EcoInvent Provider
Steam, in chemical industry	kg	-	33.87	33.90	Market for steam, in chemical industry|steam, in chemical industry|cutoff, S—RoW
Nitrogen, liquid	kg	-	11.52	11.50	Market for nitrogen, liquid|nitrogen, liquid|cutoff, S—RoW
Heat, district or industrial, natural gas	MJ	529	191.47	360.20	Heat production, natural gas, at industrial furnace > 100 kW|heat, district or industrial, natural gas|cutoff, S—RoW
Electricity, medium voltage	kWh	21.92	44.87	33.40	Market for electricity, medium voltage|electricity, medium voltage|cutoff, S—JP
Acrylonitrile	kg	2.09	1.88	2.00	Market for acrylonitrile|acrylonitrile|cutoff, S—GLO
Vinyl acetate	kg	0.018	-	0.02	Market for vinyl acetate|vinyl acetate|cutoff, S—GLO

presents the two usable inventories extracted from the literature to suggest the study’s inventory and EcoInvent flows used in OpenLCA for impact calculation.

Flow quantities were retrieved from the literature and converted into similar units to aggregate them. It was assumed that the natural gas consumed in Das’s [24] study was used to produce heat. EcoInvent assumes that 0.03 m³ of natural gas at high pressure is needed to obtain 1 MJ of heat. This hypothesis was used to suggest a heat-production value per Das’s [24] study. For the purpose of data conversion, a 54 MJ.kg⁻¹ calorific value and a 0.75 kg.m⁻³ density for natural gas were considered.

EcoInvent providers were suggested, with the assumption that the study takes place in Japan, as major carbon-fibre manufacturers are located there. Based on the hypothesis set in the Section 2.3.2, flows suggested are:

• Steam is used during fibre stretching and is an energy flow [4];
• Nitrogen is used as an inert gas in the carbonisation process;
• Natural gas is used as an energy input during carbonisation and also used during precursor production, to heat the product;
• Electricity is used at each step of the process;
• Acrylonitrile (AN) is the initial material in the PAN manufacturing;
• Vinyl acetate is used as a comonomer with AN during wet spinning.

Inventories present differences regarding the heat and electricity flows, but it is not possible to say where these differences come from, as there are few hypotheses presented in the articles. For example, parameters such as fibre quality, line speed, and carbonisation temperature were not presented.

Flows under study and data quality of this inventory, compared to articles from group 2, are then presented, as these aspects are crucial to make LCA analysis reliable and comparable across studies.

3.2.1. Flows under Study

Flows analysed in the current study are compared with other studies using primary data (group 2), as shown in Figure 3. A comprehensive flowchart is presented as well, highlighting all the flows raised across articles in the literature. Also, flows were not linked to processes in the studies. Indeed, the current literature does not give enough detail on such information.

Acrylonitrile was also included. In fact, this flow is always considered, because it is part of the PAN flow in the EcoImpactCalculator [21,22]. Nitrogen was considered by the JCMA study [20], but no other ones. Steam was considered in the JCMA [20] and Zhang [23] studies, but not in the EcoImpactCalculator [21,22]. Vinyl acetate has been included in this inventory, but hot-spot analysis has revealed that it induces negligible impacts.

Some flows could not be included in the inventory presented in Table 2, as they were not presented in the two articles from group 3. These flows are water, solvents, steam, potassium permanganate, ammonium bicarbonate, PAN fibre oil, and solid wastes. Moreover, exhaust gases were not included in the inventory. The latter are ${\mathrm{CO}}_2$ and SO_x for EUClA [21]; ${\mathrm{CO}}_2$, ${\mathrm{NO}}_2$, and ${\mathrm{SO}}_2$ for JCMA [20]; and ${\mathrm{CO}}_2$ and SO_x for EUClA ROK [22]. Five flows were never studied across articles; these flows are helium, compressed air, finish, fresh air, and emulsion.

Hence, the current inventory suggests using 6 flows, but 15 other flows were found across the literature and could not be included in this inventory.

3.2.2. Data Quality

Data quality has been assessed with the DQR approach, based on PEF recommendations [8] and is presented in

Table 3. DQR rating of studies presenting primary data (usable and unusable). P = precision, TeR = technological representativeness, TiR = time-related representativeness, GeR = geographical representativeness, DQR = data quality rating.

Author	P		TeR		TiR		GeR		DQR
	Mark	Remark	Mark	Remark	Mark	Remark	Mark	Remark
[5]	5.00	-	5.00	-	5.00	2009	5	-	5.00
[24]	5.00	-	5.00	-	5.00	2011	5	-	5.00
[21]	2.00	Not independently verified	1.00	At least 3 companies	5.00	2016	5	Europe	3.25
[20]	1.00	External verification	1.00	40% of the Japanese market	5.00	2017	1	Japan	2.00
[22]	2.00	Not independently verified	1.00	At least 3 companies	5.00	2015–2021	1	South Korea	2.25
[23]	5.00	-	1.00	1 company	5.00	2022	1	China	3.00

. When no information is clearly given, the worst mark is given. The GeR rating was given on the basis that the study took place in Asia, in order to remain consistent with the assumptions of the proposed inventory concerning the origin of flows.

All the datasets have the lowest score for temporal correlation, as they all date back more than three administration periods in relation to the PEF report.

Data from Das [24] and Duflou [5] have the lowest possible DQR (poor quality) because little information was given regarding their way of retrieving the data, as well as regarding the fibre quality. Moreover, these studies present old data. EUClA [21] is of fair quality, which is due to the fact that the data are representative of Europe, and not from Asia. The JCMA [20], EUClA ROK [22], and Zhang [23] studies show good data quality. This could be improved by temporal correlation. JCMA [20] presents the best data quality amongst others, with consideration to Asia.

Flows under study as well as data quality will be used to discuss the inventory suggested. This inventory can then be used in LCA software to analyse its associated environmental impacts.

3.3. LCIA Results and Interpretation of Carbon-Fibre Production

The impact calculation has revealed the following indicators, which are shown in

Table 4. Environmental indicators calculated regarding carbon-fibre manufacturing. Indicators under analysis are the ones suggested by the PEF guidelines and the CED.

Impact Category	Reference Unit	Value
Acidification	mol H+ eq	0.25
Climate change	kg CO2-eq	72.32
Ecotoxicity, freshwater	CTUe	841.90
Eutrophication, freshwater	kg P eq	0.01
Eutrophication, marine	kg N eq	0.06
Eutrophication, terrestrial	mol N eq	0.49
Human toxicity, cancer	CTUh	1.62 × 10−8
Human toxicity, non-cancer	CTUh	3.36 × 10−7
Ionising radiation	kBq U-235 eq	3.61
Land use	Pt	72.96
Ozone depletion	kg CFC11 eq	4.44 × 10−6
Particulate matter	disease inc.	1.44 × 10−6
Photochemical ozone formation	kg NMVOC eq	0.13
Resource use, fossils	MJ	1051.02
Resource use, minerals and metals	kg Sb eq	8.87 × 10−5
Water use	m3 depriv.	12.23
CED	MJ	1176

. Moreover, a contribution analysis is shown in Figure 4. First, it can be noted that vinyl acetate used as a comonomer during polymerisation of PAN is negligible across all impact categories. Moreover, heat production is a major source of impact regarding climate change, ozone depletion, and CED. Electricity is a major source of impacts for all of them, except ozone depletion. Acrylonitrile (AN) is a major source of impact regarding particulate matter and CED. Nitrogen can be neglected across all impact categories except particulate matter. Steam is a major source of impact in terms of particulate matter. It cannot be neglected across other categories.

Environmental impacts associated with carbon-fibre production are then included in the hydrofoil analysis.

3.4. Results Associated with Carbon Fibre-Reinforced Composite Structure

Impacts linked to the infusion of carbon-fibre/epoxy laminate have been calculated and presented in

Table 5. Environmental indicators calculated regarding the carbon/epoxy laminate. Indicators under analysis are the ones suggested by the PEF guidelines and the CED.

Impact Category	Reference Unit	Value
Acidification	mol H+ eq	36
Climate change	kg CO2-eq	10,000
Ecotoxicity, freshwater	CTUe	130,000
Eutrophication, freshwater	kg P eq	1.4
Eutrophication, marine	kg N eq	8.3
Eutrophication, terrestrial	mol N eq	61
Human toxicity, cancer	CTUh	2.7 × 10−6
Human toxicity, non-cancer	CTUh	5.2 × 10−5
Ionising radiation	kBq U-235 eq	510
Land use	Pt	10,000
Ozone depletion	kg CFC11 eq	6.6 × 10−4
Particulate matter	disease inc.	2.1 × 10−4
Photochemical ozone formation	kg NMVOC eq	9
Resource use, fossils	MJ	1,500,000
Resource use, minerals and metals	kg Sb eq	1.8 × 10−2
Water use	m3 depriv.	1900
CED	MJ	170,000

. This calculation is based on the hypothesis set Section 2.2 on the hydrofoil mould and the process presented in Figure 2. The hot-spot analysis is shown in Figure 5. First, it can be noted that inert wastes have negligible effects on each impact category.

Then, carbon fibre offers a major contribution to all indicators calculated here (89 to 95%). The epoxy resin is not negligible regarding ozone depletion, but it still has a minor contribution (11%).

4. Discussion

4.1. Carbon-Fibre Inventory

The current results have been compared with those from other studies (in line with the methodology presented), following three aspects: data quality, flows under study, and LCIA results.

4.1.1. Flows under Study

The suggested inventory could benefit from more flows under study. This tends to reduce environmental impacts calculated. Also, no consistency was found between flows analysed. This could justify differences regarding impacts calculated by studies.

For the moment, it is not possible to link the flows in the studies with the processes responsible for these flows. It would therefore not be possible to carry out a contribution analysis of the results of a complete inventory. Nor would it be possible to discuss the values suggested by a complete inventory, as we would not have access to details of the flows by process.

4.1.2. Data Quality

Das’s [24] and Duflou’s [5] studies have the lowest quality data, while EUClA [21,22], JCMA [20], and Zhang [23] have better quality data. Nevertheless, the studies with the best data quality did not present usable inventories. The data quality presented in our inventory is therefore debatable, which means that calculations will contain more uncertainties than the ones already made. Nevertheless, having access to a usable inventory allows performance of a multicriteria analysis, which is not possible with current studies presenting better data quality.

4.1.3. LCIA Results

Environmental indicators calculated from our inventory are compared with the current literature.

Figure 6 gathers data from the studies that were conducted with secondary data. For each study, cumulative energy demand (CED) and global warming potential (GWP) related to 1 kg of CF produced are presented. For the purposes of this comparison, we have retained only articles presenting perimeters that include PAN and fibre production.

Our results are generally two to five times higher than those in the literature based on secondary data. Differences in our results could be caused mainly by data sources used in studies, as well as assumptions set to aggregate data. For example, Hohmann et al. [29] used Gabi’s database to calculate PAN and CF production impacts, but this database is used as a black box. Also, some authors such as Stiller [30], Ghosh et al. [31] or Schnöll et al. [32] had discussions with academics or companies but did not specify how their data was retrieved. For example, if the data were retrieved via expert estimation, quality might be debatable. The most valuable data come from the paper by Meng et al. [29]. They used data sources from Duflou [5] regarding energy, saying that they were close to a confidential dataset. They also used data from Das [24] regarding AN and vinyl acetate quantities, but as has shown in our contribution analysis, these flows are negligible for the GWP impact. They also used a master’s thesis that calculated water, sizing solids, and sulphuric acids in its analysis to add these flows to their inventory. In 1997, De Vegt and Haije [10] calculated the heat-related energy demand for producing carbon fibre, based on the yield of the process, the specific heat of the PAN, and the oxidation and carbonisation temperatures of the fibre. Johnson and Sullivan [15] assessed the impact of carbon-fibre production in 2014. To do this, they first assessed the energy demand for PAN production, based on the different materials involved (ammonia, AN, propylene, and methyl methacrylate), as well as for PAN polymerisation. Energy demand was calculated from the quantities of natural gas, electricity, coal, and oil required for these different elements. The energy demand for the carbon-fibre manufacturing process was taken from the Das study [24]. Forcellese et al. [13] evaluated the impacts of carbon-fibre manufacturing in 2020, based on data from Duflou et al. for PAN production [5] and data from Khalil, 2017, for fibre production [14]. Khalil’s study has not been included here as it presents an inventory relating to fibre manufacture only. Finally, Kawajiri and Sakamoto [34] determined the impact of carbon fibre in 2022 on the basis of Toray patents for PAN production and Harper International data for carbon-fibre manufacture.

Differences within the literature presenting primary data are also presented. The literature presents an energy demand that is 10 to 70% lower than the current study. EUClA [21,22] seems to present an energy demand close to the current article (10% lower), which is surprising, as we used an energy mix from Japan. JCMA [20] presented a 70% lower energy demand. This difference could be linked to the production scale, but we cannot be sure, as this hypothesis was given neither in this report nor in the JCMA study. We can also observe an energy demand three times lower than that in the study of Dér et al. [4]. In fact, this study presented a direct energy demand to produce carbon fibre, which is not a CED. The study of Zhang et al. [23] presents a similar global warming potential (25% lower). Nevertheless, this difference cannot be justified, as flow quantities were not provided in this study.

Here, we cannot precisely determine the reason for the differences between our data and those from the literature, as the choices made to set up the inventories are not clearly presented. Nevertheless, differences with the literature could be justified by the high heat value in the current inventory. Moreover, the current inventory might not benefit from high-scale production. Also, differences with the literature could be justified by the quality of the fibre under study.

4.2. Carbon Fibre-Reinforced Composite Structure

It has been shown here that with our inventory suggestion, carbon fibre has major impacts during manufacturing of carbon-fibre/epoxy laminate. These conclusions also reveal a need to work on the environmental burdens caused by performance aspects. Indeed, depending on the quality of the fibre used, environmental impacts could be reduced, and a compromise between quality and environmental aspects could be reached. This case study could serve as a basis to calculate impacts from other laminates and could lead to a more precise evaluation of a break-even point on vehicles, for example.

5. Conclusions

Carbon fibre-reinforced plastics seem to propose an interesting solution with which to tackle global warming linked to transportation. Moreover, this material is used in the competitive sailing boat industry for high-performance purposes. However, environmental data related to carbon fibre manufacturing are hard to include in environmental assessments. Indeed, we see that two main issues arise after reviewing the current literature. First, results across studies differ greatly. Moreover, current studies do not allow performance of a multicriteria analysis. These issues could be tackled by suggesting a justified, reliable, and transparent inventory in the environmental analysis of manufacturing carbon fibre.

The current study aims to offer new insights regarding carbon fibre’s environmental impacts, based on the LCA methodology. It used primary data from a literature analysis so that the EcoInvent flows inventory was provided. The suggested inventory was then critically discussed in terms of data quality, while a multicriteria analysis following PEF framework was performed. Finally, the inventory was used to assess the environmental impact of carbon fibre-reinforced plastic (CFRPs) applied to an industrial example representative of the competitive sailing boat industry: a hydrofoil mould.

Several conclusions can be drawn from this article:

• Two articles were taken as a basis for suggesting an inventory regarding carbon-fibre manufacturing. The latter suggests using six flows, related to EcoInvent providers;
• The hot-spot analysis has shown that only vinyl acetate has a negligible environmental impact compared to other flows, regarding carbon-fibre manufacturing;
• Data quality analysis and flow analysis showed that this carbon-fibre inventory could benefit from better quality data and more flows under study. However, the inventories suggested in the literature with better data quality and more flows were not usable;
• Indicators calculated in this analysis are two to five times higher than others from the literature. This might be linked to the heat value suggested in the current inventory, different production scales, or quality of the fibre under study. Indeed, a different fibre quality would imply different flows under study. For example, a higher temperature is needed to produce high-modulus carbon fibre, compared to an intermediate modulus. This could affect the heat needed to produce carbon fibre. Nevertheless, in this study, the high heat value could also be justified by the hypothesis that natural gas is used only to produce heat;
• The environmental analysis performed on the hydrofoil mould revealed that carbon fibre has a major contribution across all indicators selected for the hot-spot analysis. This reveals a need to work on the environmental aspects linked to high-performance sports.

Several conclusions can be drawn, in order to help spread quality inventories in the environmental analysis of carbon-fibre manufacturing.

First, more industrial data will have to be gathered in the future. Even if initiatives already exist, discussions could be broadened to include more environmental indicators. The flow analysis performed in this study could help plan data collection. This data collection should be performed process by process, so a contribution analysis could be possible.

In addition, studies performed in the future should clearly mention the mechanical characteristics of the fibre (e.g., quality and number of filaments), as well as those of the production line (e.g., line speed, yields, and production sizes).