Over the years, Victrex’s (Cleveleys, U.K.) LMPAEK polymer has increasingly become a versatile enabler in the realm of thermoplastic composites (TPC), intersecting with projects that push boundaries in welding, forming and lightweight structures.

From NASA’s in-space manufacturing to cryogenic fuel lines and helicopter components, LMPAEK’s low-melt properties facilitate efficient processing and high performance in demanding environments. These efforts highlight a shift toward sustainable, high-rate production without compromising structural integrity.

→ The short is accessible at CompositesWorld’s YouTube channel.

→ Full-length video is accessible via the ThermoForged YouTube channel.

Aluminum and furan and polyester alternative by NPSP. Source | Bambooder Biobased Fibers B.V. 

These electric vehicle (EV) components show how aluminum can be replaced by bamboo fiber-reinforced alternatives, including 100% bio-based resins (black part, center) or partly bio-based polyester resins (far right).

The black part is carbon-negative (≈ -2 kilograms CO_₂ per kilogram of product) while both alternatives cut weight by 50% versus the original aluminum part with the same performance, including EMC shielding and heat resistance.

The components were made possible using BambooSheet Pro 200-400 LF by Bambooder Biobased Fibers B.V. (Amsterdam, Netherlands) with compounding and molding by NPSP B.V. (Delft, Netherlands).

That is why Cambium caught my attention.

Cambium is backed by more than 20 different funding sources. Source | Getty Images (top) and Yannick Willemin (bottom)

The El Segundo, California-based advanced materials company announced a $100 million Series B in January 2026, led by 8VC Just weeks earlier, Cambium had announced the acquisition of U.K.-based SHD Group, the established advanced composites manufacturer now operating as “SHD, a Cambium company.”

Taken separately, each announcement was noteworthy. Together, they suggest something more strategic.

Cambium has positioned itself as an advanced materials innovator working on difficult problems in defense, aerospace and other high-performance sectors. Publicly, the company emphasizes AI-enabled materials discovery and the creation of new monomers and polymers, not just reformulations of existing chemistries. SHD, by contrast, brings an established prepreg platform, a global footprint and decades of accumulated know-how in advanced composite materials manufacturing. Cambium says the acquisition gives the combined company an innovation and aerospace qualified manufacturing footprint spanning the U.S., the U.K. and Slovenia.

This is where the story becomes more interesting than a classic startup funding round. At JEC World, Brett Schneider, now president of SHD and Cambium Composites, described the logic in unusually direct terms: “How do you scale a materials company when in this industry it’s fairly asset-heavy at times?” His answer was equally direct: “Putting the two companies together under Cambium’s new manufacturing vertical for composites just made a ton of sense and accelerates commercialization of new chemistries”

That line deserves attention because it reframes the role of fundraising. In many startup stories, capital is presented as validation. Here, it looks more like industrial design.

Schneider is a credible guide to that logic. He brings three decades of experience in advanced materials, automation and industrial leadership. Before joining Cambium’s board in 2025, he held leadership roles at Hexcel, served as chairman and CEO at Carbon Conversions and worked as an operating partner at Arsenal Capital. In other words, he not only understands how composites businesses are run. He understands how industrial platforms are built.

That may explain why the Cambium-SHD construct feels unusual, but also smart.

In our discussion, Schneider described how the relationship began through toll manufacturing and then evolved as the teams recognized an unusually strong cultural and strategic fit. “It became obvious that we should just put the two companies together,” he said.

“Culture” can sound soft in dealmaking language, but in composites it matters. Materials businesses are not software businesses. They depend on tacit know-how, process discipline, customer trust and the ability to move from development to production without compromising quality. That is also why the transaction appears to be about more than adding manufacturing capacity.

Brett Schneider (left) discusses how to scale an AI materials company during an interview on The Composites Catalyst. Source | Yannick Willemin

Fairmat’s FairFactory production site employs robotics- and AI-optimized cutting, ply placement and tracking to produce high-quality recycled carbon fiber products to meet industry needs. Source (All images unless otherwise specified) | Fairmat

“We didn’t launch Fairmat as a recycling company. Yes, we are a material provider that is using waste as a primary source of incoming raw materials, but we are a material provider first and foremost,” emphasizes Benjamin Saada, CEO and founder of Fairmat (Paris, France), a startup that supplies composite materials manufactured from recycled carbon fiber (rCF).

Saada, who previously co-founded the Paris-based aircraft seat manufacturer Expliseat, founded Fairmat in 2020. “We launched this company to replace aluminum,” he explains. “If you look at the specifications of Fairmat’s material, it’s roughly the specifications of aluminum.” He knew that carbon fiber-reinforced polymer (CFRP) composites provided the needed mechanical properties, but recognized the higher cost of carbon fiber versus aluminum — and its reliability on petroleum as a precursor material — begged for a better solution.

“We started thinking about how we could use composite waste to create a new material that is at least as good as aluminum if not better, and no more expensive. We ended up developing a completely new process of creating materials,” Saada says.

This process is simple at its highest level: Turning high-quality CFRP scrap into new, reusable materials that retain as much of the original material’s properties as possible. Most recycling technologies in this landscape either involve a chemical- and/or heat-based fiber reclamation process or mechanically grind up the full CFRP scrap — all of which typically results in very small pieces that are then compounded for use in injection molding.  

Fairmat’s technology (detailed more below), however, is centered on two main points: retaining as much of the length and performance of the original fibers as possible to create new materials with high value, and doing this through a digitally traceable, data-optimized process.

Fairmat decided on the size and format of its standard chips and other products in an effort to preserve as long of a fiber length as possible and to function as efficiently as possible on current equipment.

The company’s cornerstone product is its chips. These are single-ply, cured and precision-cut from woven or unidirectional (UD) CFRP, in standard sizes of 10 × 4-centimeter rectangles or 6 × 6-centimeter squares. The chips can be provided as they are, or further manufactured into one of Fairmat’s other products — single-ply, wider FairPatch or longer FairStrip elements, multi-layer custom Fairmat Plate laminates, or sheets and rolls of layup-ready FairPly. The company’s latest product, FairBoard, diverges from the chip concept for the first time, with the manufacture of structural panels made from recycled pultruded profiles.

Fairmat decided to work with chips as its baseline product “because we wanted to be able to control the length of the fiber that we keep in the mix,” Saada explains. “If you use a grinding process, for example, you’re breaking the fiber. The decision was all about providing our customers with as high performance and useful of a material as we could.”

A demonstration of the FairTrack system during a facility tour of the FairFactory. This data storage and AI-based optimization system is called the “brain of the factory.”  

The key, Saada explains, is finding a way to manage material input diversity. “If you cannot manage this, you cannot guarantee any kind of quality. Because of this constraint, it was obvious to me that digital data management was the only way to scale this solution. We started simple, just by recording everything in order to find a way to standardize our hardware, machines and our process.” This then expanded to today’s in-house software system FairTrack, which enables real-time data tracking, connects all machines within the plant and provides AI-enabled correction for cutting and ply placement.

This level of digitization is not yet common in composite materials and parts production facilities, and thus impressive. “The fact the products are recycled is a cherry on top,” notes Saada, “but the goal was always about answering industry needs for high-performance and lightweight materials that aren’t dependent on natural resources.”

For source material, high-quality, high-quantity manufacturing scrap from the aerospace industry was the natural place to start. One of the startup’s  first partnerships was with the European branch of carbon fiber manufacturer Hexcel Corp. (Stamford, Conn., U.S.) to recycle prepreg cutoffs. Fairmat’s Nantes production site is a former Hexcel plant that the company originally leased in 2022 as part of this partnership and now owns. Since then, Fairmat has announced partnerships with multiple other carbon fiber manufacturers or OEMs that supply the company with high-quality scrap materials, including Airbus, Dassault Aviation, Siemens Gamesa, Syensqo and Tarmac Aerosave.

Today, Fairmat employs about 100 people and operates four sites —  its Paris headquarters and R&D lab; a main production and development site in Nantes, France; a second production site in Salt Lake City, Utah, U.S.; and a new development site in Danyang, China. According to Saada, about 70% of the company’s business is in Europe and 30% in the U.S.

CW recently had the opportunity to participate in a tour of the Nantes production site to learn more about Fairmat’s process of manufacturing its rCF materials.

The FairFactory: From scrap to chips to final products

The 2,000-square-meter Nantes production site is called the FairFactory and employs 54 people on-site — more than half of which are engineers managing the software and mechatronics — operating 25 robots and a full-scale process from recycling to materials manufacturing to quality inspection and new development.

Currently, Fairmat mostly processes carbon fiber/epoxy prepreg scrap from aerospace manufacturing. The company is also working on processes incorporating other feedstocks, and more recently, has started receiving profiles sourced from wind energy spar caps made from carbon fiber and either vinyl ester (VE) or polyester (PE).

Incoming materials are sorted by type (top) and tested to ensure they meet quality standards and that data is then stored in the FairTrack digital system (bottom).

Julien Pascal, head of sales at Fairmat and one of the tour leaders, explains that incoming scrap feedstock materials are stored in a small accessory unit outside the main building. “When we start working with a new supplier, the materials first need to go through a range of quality tests and are qualified at the lab in Paris,” Pascal explains.

A small test lab on-site at Nantes also houses equipment for measuring incoming material parameters like fiber density and thickness as needed. Mechanical tests like bending and tensile tests on final parts or chips can also be performed on equipment here, and Fairmat plans to soon add additional machines for increased test capabilities.

“All of the data of course goes into the FairTrack system to ensure it’s all traced in the system,” Pascal emphasizes.

Preparing the materials: Curing and cutting

First, films are removed from prepreg waste. Then the first process step for waste feedstock is a curing step via a France Etuves (Chelles, France) oven, which cures the pieces of material together into one single-ply sheet.

The main process steps for Fairmat’s chips are layup and curing of like scrap materials to create single-ply laminates (top) followed by cutting into chips or sheets that will be used to make other material products (bottom).

Out of the oven, the materials are sorted by a technician into material type — woven fabric versus UD, and different types of carbon fiber — into bins, and tagged with printed tracking labels that will travel with them and be scanned at each step through the rest of the process.

The cured single-ply sheets are then moved to one of five Gunnar (Altstätten, Switzerland) Novex AI-optimized, automated cutting machines to be cut into one of the company’s standard chip size parameters or a customer-specific custom patch pattern.

A sixth, newer cutting machine is set apart for R&D. This station is equipped with a projector and camera system that project the cutting pattern onto the material. The system is controlled by in-house-created cutting software that generates and sends data into FairTrack to measure cutting process efficiency in real time.

Fairmat’s materials are already manufactured from prepreg offcuts, so ensuring that as much material as possible is used to make chips is top priority.

This cutting efficiency is key, Pascal explains — the goal is to reuse as much of the repurposed material as possible. The small amount of scrap that is left is typically rerouted back into the initial processing steps to be made into new chips.

At this point, chips are ready for final inspection and shipment to customers. Or, they can be moved to another area of the FairFactory for manufacture into the company’s FairPly product.

Manufacturing FairPly

A roll or sheet of FairPly is essentially composed of two elements — a single layer of rCF/epoxy chips bonded to a supportive substrate material, that is typically a glass or carbon fabric. Substrates are first prepped with a layer of liquid PE to serve as a binder and adhesive for the chips.

A series of four robotic cells house two Kuka (Augsburg, Germany) pick-and-place robots supporting each layup platform on which the substrate is placed. The robots then automatically place chips in programmed patterns onto the substrate, and ultimately bond them to the PE-based substrate via a cold atmosphere plasma process.

The FairPly material (top) manufacturing area (shown in middle image during FairFactory tour) comprises pick-and-place robot cells (bottom) that are optimized via FairTrack for precise layup.

As the robotic cell continues working, Saada explains, “Quality control is ensured at every step of the robot’s process.” Cameras at the top of each cell provide FairTrack with constant visual information such as size and angles of each chip, while sensors on the robotic arms relay information such as chip weight — this data is being used by the system to constantly adjust the cell’s activity to account for any small inconsistencies in the recycled materials.

“FairTrack is the brain of the factory,” Saada emphasizes.

The orientation, angle and shape in which the chips are placed is vital to the performance of the overall FairPly product. The standard 6 × 6 and 10 × 4-centimeter chip sizes were developed to make themost efficient and highest-performance use of the fibers — “and to retain as much fiber length as possible while using as much waste as possible,” Pascal says.

Final FairPly sheets finish curing at room temperature, and then can be rolled and shipped as off-the-shelf products or finished per customer specifications.

While the timing for each sheet depends on specific size and specifications, approximately 77,000 chips are processed into FairPly per week on this line.

Finally, on the main factory floor beside the cutting machine area, a product inspection station is set up for final inspection of representative chip samples and finished custom materials before packaging and shipment. Size, angles, fiber alignment and other measurements are recorded via FairTrack. At the time of the FairFactory tour, customer shoe inserts were being readied in this area for final inspection.

New and in development: FairBoards and pipes, Infinite Recycling

The tour ends in the development room — this is not Fairmat’s full R&D site, which is located at its headquarters in Paris, but an area in which new machines and technologies can be evaluated for use on the FairFactory floor.

At the time of this tour, the company set up several demonstrations showcasing developments both from the Nantes site and its other facilities.

New product capabilities. Claire Schune, product manager, introduces the company’s latest product, FairBoard, which right now is being produced out of the Fairmat facility in China, though there is are small development presses at the Nantes site and the Paris R&D lab.

Why the Chinese facility? “The full-sized press we acquired is in China, so logistically it makes sense to do the development work in China,” Saada explains. The goal is to bring the technology to Nantes as well.

Unlike the company’s typical single-ply chip-based products, FairBoards are multilayer panels sourced from pultruded profiles. Pictured samples were on display at JEC World 2026. Source | CW

These panels, manufactured currently up to 4 × 1.6 meters in size, start from a base of rCF/PE or carbon fiber/VE waste from pultruded spar caps sourced from the wind energy industry. Fairmat has developed and patented a low-energy method for cutting these spar caps to size, after which the materials are laid up between layers of glass fabric and new liquid epoxy. The board is then pressed in a closed mold. “The glass fabric and epoxy serve as extra reinforcement to the recycled material. It fills in any cracks in the pultruded waste and keeps the board leak-tight,” Schune says. Epoxy was chosen over PE or VE for its higher mechanical strength and fatigue resistance to meet customer performance standards, while still being compatible with the original resin and working as an adhesive to bond the assembly together. After molding, the FairBoards are trimmed and can be painted.

In Nantes, the Fairmat team is working on automating and optimizing this process, with a development-scale robotic cell that is similar to the FairPly cells. “FairBoards are basically just bigger chips, in a way,” Schune says. Similarly, in the development cell, AI-enabled cameras are being optimized to scan and correct the layup process, and a robotic arm operates a roller for laying up the glass fabric layers. An AI-optimized, robotic resin deposit system is also in development.

The final FairBoard panels can be used for a variety of end-use applications, including countertops and wall or ceiling panels.

Fairmat continues optimizing its products for the end user. New surface treatment options allow for more flexibility to mold into pipes, angled bars or other shapes. Source | CW

In addition to FairBoard, the company also recently launched a version of its Fairmat Plates that are optimized with a special surface treatment for molding into pipes or other tube-shaped parts. These were on display in the days before CW’s visit to the FairFactory at JEC World 2026.

“Our surface treatment creates a more flexible version of the plates and chips, and there are no limits to the types of applications possible,” Saada says.

Quality inspection. The Fairmat team is also constantly working on further optimizing its quality inspection system, including its processing software for analyzing camera images of chips ply-by-ply to inspect for gaps, overlaps and distance between chips. “We always want to continue optimizing and improving the precision of our robots,” Pascal says.

Fairmat has patented a process for using cold atmospheric plasma systems for both bonding its chips together into its FairPly products, and for reversing this process to remove resin and separate the individual chips out again.

Infinity Recycling. Michael Gaultois, chief science officer, and Pauline Prouilhac, materials engineer, present Fairmat’s Infinity Recycling process, which is a patented technology that uses a cold atmospheric plasma system, optimized to fit Fairmat’s needs, to break the bond between Fairmat chips without damaging the chips themselves.

“We ultimately want to be able to not only make materials from rCF feedstocks from our supply partners, but to collect parts made from our FairPly materials back, remove the new resin that was added in the FairPly process in between the chips and give these parts another life,” Gaultois explains.

Rackets, skis, shoes and more

Some sporting goods customers purchase Fairmat materials to mold into new parts themselves, or Fairmat can provide custom materials (like the pictured shoe inserts).

After leaving the FairFactory, Fairmat’s customers further mold these materials into a variety of end-use parts, especially in the sporting goods sector. Today, recycled chips and plates can be found in skis, tennis rackets, padel rackets, running shoe inserts, sports helmets and more.

The company is moving toward applications in new markets like construction pipes and brackets, and drone airframes.

New investment, partnerships and next steps

Saada acknowledges that for all companies that work in the recycled materials space, funding and investment are constant challenges. “There are so many initiatives being launched that it’s somewhat diluting the investment and the energy. While we obviously need several technologies for recycling, right now the industry really needs to focus on the few technologies that make the most sense.”

That said, Fairmat has seen success both in attracting investments and other partnerships, including most recently the announcement of €10 million in equity from the Circular Plastics Fund, managed by Infinity Recycling (Rotterdam, Netherlands), in addition to an announced €51.5 million raised in early 2025.

So far, most Fairmat rCF material end users are in the sporting goods market, but the company is making headway into new applications in construction and beyond. Source | CW

Securing long-term feedstock supply and interest in recycled products is also a challenge to all in the rCF space. For example, the main drivers to recycle manufacturing scrap or adopt rCF into new applications are still coming from private actors versus regulatory incentives. While there are a growing number of companies on board, including aircraft OEMs like Airbus and Boeing, “I think the authorities need to push the market a little bit in this direction,” Saada says.

In March 2026, coinciding with JEC World, Fairmat announced six new partnerships expanding both its feedstock supply base and new application areas for its materials. These include an expanded partnership to recycle prepreg scrap feedstock from Syensqo; an R&D initiative to explore new approaches for recovering materials from end-of-life aircraft parts with Airbus; and deals to supply rCF materials for use in construction applications (Etex), racket sports equipment (Babolat) and orthotic footwear (Launchpad O&P and Billy Footwear).

“We’ve always positioned ourselves as not only a recycling company, but as a high-performance material supplier for sporting goods,” Saada says. “Now we’re aiming to replicate what we’ve done in sporting goods to other sectors like construction, and to deploy our materials into applications with industry leaders. We are very proud and excited of this first step with Etex in Europe, but we know we can do much more.”

Along with growing its feedstock and customer bases, Fairmat aims short-term to focus on the launch of its FairBoard products and longer-term, to continue expanding its footprint internationally, particularly in China and the U.S.

Saada notes, “For many years, most of the green innovation was more expensive, and we [as a society] started to associate those two characteristics. But that is not what Fairmat is about. My strongest conviction is that economic rationality is also ecological rationality. At the end of the day, if you reduce CO₂ emissions, you save energy. If you save energy, the product is ultimately cheaper. Economical and ecological innovation are naturally aligned. At Fairmat, we’ve created a solution that is good, technically smart, economically viable and scalable. The incentive for the customer is the quality, the cost and the performance of our product.”

While comprising honeycomb metal, this airplane wing flap showcases the image detail produced by the AdaptixNDT3D. Source (All Images) | Adaptix Ltd.

For as long as composite components have been made at scale, the dominant quality model has been the same: manufacture the part, then inspect it. End-of-line nondestructive testing (NDT)  has been the primary mechanism for separating conforming product from scrap for decades and, when components were relatively small and volumes were low, the cost of late-stage rejection was commercially manageable.

However, composite structures are now becoming larger, more structurally critical, complex and produced in higher volumes across a wider range of industries than in the past: wind turbine blades regularly exceed 80 meters; composite-intensive aircraft programs target increasingly challenging  production rates; the automotive sector is scaling carbon and glass fiber composite structures out of low-volume niches into mainstream architectures — and with that expansion comes a real quality issue. If a defect is only discovered at the end of the production process, the full cost of manufacturing that part has already been spent.

Where a delamination or fiber misalignment identified during layup might cost relatively little to address, the same defect found after curing, or even later in the manufacturing process, can cost significantly more. In many cases, the answer is simply to scrap the component. With composite components often being energy-intensive to produce and difficult to recycle, this will also have impact upon sustainability issues.  

Existing methods and challenges

Existing composite NDT techniques are not necessarily ineffective in themselves, but they have been designed for a different era.

Ultrasonic testing (UT) remains the primary quality assurance method for composite parts, especially in the inspection of aerospace primary structures (such as wings and spars). It requires a skilled operator and is relatively slow, but automated phased-array systems have advanced considerably. The Welding Institute’s (TWI, Cambridge, U.K.) robotic inspection systems, for example, have demonstrated a significant reduction in inspection time for curved carbon fiber composite parts compared to manual methods. But even automated UT is a contact or near-contact method, and demands careful coupling of media, surface preparation and setup time. In short, it is optimized for discrete inspection, not live production environments.

Infrared thermography and shearography offer rapid, noncontact coverage of large surface areas and genuine potential for in-line use with composites. However, their limitation is depth sensitivity, diminishing significantly with thicker structural laminates.

Alternatively, computed tomography (CT) can provide a reconstructed 3D image of the part being inspected, including significant 3D detail throughout its geometry. However, CT has traditionally been size-constrained via expensive cabinets with radiation protection as well as lengthy image acquisition times, issues that preclude routine use in the manufacturing environment. 

New NDT for composite parts production

What is now needed is NDT technology that can operate in a production environment, at production speed, without disrupting the processes around it. Fortunately, a raft of new and emerging technologies are being developed to close that gap.

Air-coupled ultrasound removes one of the core barriers to in-process UT integration: the need for a coupling medium. Unlike conventional systems that require immersion tanks or water jets, air-coupled systems require nothing beyond proximity to the component surface, making them significantly more amenable to integration on production lines. Laser ultrasonic (LUS) and laser-excited acoustics (LEA) technologies take the non-contact principle further, using pulsed lasers to generate and detect ultrasound with no physical contact at all. This makes these techniques potentially well-suited to environments where contact is impractical, such as freshly laid composite preforms, complex geometries or components operating at elevated temperatures. 

Robotic-automated UT is one of the more mature “emerging” technologies, combining multi-channel phased array UT with digital twin integration and Industry 4.0 connectivity. This enables automated inspection of complex composite structures and is now in use in multiple high-value aerospace applications.

Digital tomosynthesis (DT) is another significant new development, enabling in-process inspection of composites. Sitting between 2D radiography and full CT, it captures a series of low-dose X-ray projections from multiple angles and algorithmically reconstructs them into cross-sectional slices. This delivers meaningful 3D structural data in a fraction of the time required for traditional imaging methods, with very low radiation levels compared to conventional CT. The resulting reduced shielding requirements and mobility make deployment within existing production environments feasible for the first time, including in smaller workshops.

Adaptix’s NDT technology in action on a composite helicopter propeller blade.

U.K.-based Adaptix Ltd. (Oxford) is at the forefront of adapting this technology for industrial NDT, developing compact DT systems capable of imaging at the preform stage — where other volumetric methods cannot — to enable an earlier first point of inspection during manufacture. The company’s collaboration with Cranfield University (Bedford, U.K.), supported by the U.K.’s Aerospace Technology Institute (ATI), has also demonstrated the integration of DT with robotics to create a scalable 3D X-ray inspection system for large composite aerospace components. Research within the same program has also explored the use of AI for enhanced defect detection, with potential for substantially improving the prospects for near real-time automated inspection.

Why a new generation of composites NDT is critical

The timing of increased interest in these emerging technologies among manufacturers is not coincidental, with macro-environmental factors encouraging greater investment in NDT innovation. For example, production rate ambitions in aerospace make the end-of-line bottleneck increasingly untenable commercially. If fuselage panels or wing components are scrapped after days of AFP, the loss also includes machine time, energy costs and production assembly planning. For the wind energy sector, where blades and structural components are large and expensive to produce, reducing scrap rates carries both financial and sustainability costs that are difficult to separate. Composite materials are also often energy-intensive to produce and difficult to recover at end of life. Thus, in manufacturing environments where emissions are under scrutiny, reducing production waste is becoming a strategic priority.

Digital transformation is also playing a role. As manufacturing environments become more data-rich, the output of in-process inspection systems becomes more valuable, not just identifying issues at an earlier stage than was previously possible, but also feeding into digital twins and digital threads. These are key to improving process control systems that can also inform design iterations and populate traceability records required for regulatory compliance.

The continued push for NDT that enables in-process inspection of composite parts has real time, cost and profit implications for manufacturers. Companies who invest in this approach will not only gain an operational edge — such as lower scrap costs, increased productivity and stronger quality data — but as composite components become increasingly common and quality expectations rise, in-process inspection will likely evolve from market differentiator to baseline expectation. Thus, investing in NDT for in-process inspection is quickly transitioning from nice-to-have to necessity.

Read more about developments in next-generation, automated NDT for composites.

About the Author

Bryn Hughes

Professor Bryn C. Hughes, FREng is an accomplished scientist and engineer with a distinguished career in government R&D, focused on the U.K’s defense and security.

His work spans a broad range of electronic and mechanical engineering. He was responsible for the development of novel sensors for harsh maritime environments; worked with central government on the protection of the critical national infrastructure from electronic attack; and led the development of an advanced manufacturing facility within government. He has broad experience across defense S&T and across government.

A Fellow of the Institution of Engineering and Technology (IET), the Chartered Institute for IT (BCS) and the Royal Academy of Engineering, (FREng), Hughes continues to advise in both the public and private sector on the development of cutting-edge technologies. For Adaptix, he is chief scientist NDT.

Fabrum has developed lightweight composite liquid hydrogen (LH₂) tanks for zero-emissions aviation and demonstrated fast filling with <5 watts of heat leak. Source | Fabrum

The development of composite tanks to store liquid hydrogen (LH₂) continues, with multiple projects presented at JEC World 2026 but no less than 22 projects listed in a recent CW report on carbon fiber and composites in H₂ storage, including LeiWaCo, COCOLIH2T, H2ELIOS, OVERLEAF, PHOEBUS, fLHYing Tank, Lufo UpLift and many more.

In late 2025, Fabrum (Christchurch, New Zealand) successfully demonstrated rapid refilling of its composite tanks to store liquid hydrogen (LH₂) with <5 watts of heat leak, meeting industry requirements. The refueling was successfully completed at Fabrum’s dedicated LH₂ test facility at Christchurch Airport. This blog is based on my interview with Hugh Reynolds, co-founder and technical director at Fabrum, to better understand the company’s developments in composite LH₂ tanks and its outlook for carbon fiber and composites in these tanks.

From composites to superconductivity to LH₂ systems

Fabrum describes itself as a development and manufacturing company. “That gives us a wide range of applications that cross pollinate and allows us to learn, transpose and test new technologies,” says Reynolds.

Trained as a mechanical engineer, he spent years in metal and composites manufacturing before co-founding Fabrum to work on superconductivity applications. “The very cold liquids and materials these systems use require non-conducting containment, typically a vacuum flask like what is used in a thermos, except it must be nonmagnetic, sealed and withstand the pressures involved,” he explains.

The company spent 15 years working with composite equipment for superconductivity, including building one of the world’s first superconducting transformers in collaboration with the University of Canterbury (Christchurch, New Zealand). “We learned a lot about how porous composites can be if they’re not made properly, and even if they are made reasonably well,” notes Reynolds. “A lot of the techniques we’ve developed came out of that work.”

“And then, just prior to COVID, when people were starting to look at H₂ for alternative fuel and energy systems,” he continues, “we realized that our technology could also be used to make lightweight LH₂ tanks for aircraft. We also developed the cryogenic cooling systems for liquifying H₂.” Fabrum approached several companies with proposals that included supplying the whole fuel system, including a small H₂ liquefaction system, distribution system, onboard composite storage tank and LH₂ delivery to the fuel cell. “And in the process of working through that,” says Reynolds, “we realized that there was a lot of discussion about how you could do this but at that time, 6-7 years ago, almost nobody was actually doing it yet.”

Fabrum was then approached by an Australian mining company. “They wanted to decarbonize and had a very ambitious project to run megawatt-scale fuel cells on LH₂,” he explains. “That gave us a wonderful opportunity to make all of this work in a real-world industrial application.” The resulting storage system was a more traditional metal dewar construction and very large — 10,000 liters. “Part of the reason it was metal was to resist damage during service,” says Reynolds. “The tray is loaded with 250 tons of rock, and the possibility of damage to the outer skin of the tank is significant because it sits between the truck’s wheels under the tray. So, we made the outer skin from 10-millimeter-thick medium tensile steel.”

Challenge of first composite tanks for LH₂, triple skin for fast fill

Fabrum’s history is essential to understand where the company is today, he notes. “If you don't have a need for this technology, then you don't spend time developing it. The true commercial desire for H₂-powered aircraft is only about 5 years old. So, if you started down the path to develop a composite LH₂ tank 5 years ago, you've only got 5 years of experience. But we've been doing it for 20 years because we wanted exactly the same technology for our superconductors.”

It isn’t surprising, then, that Fabrum was the first to demonstrate a fully functioning composite LH₂ tank. “There were some traditional metal LH₂ tanks for mobility and perhaps some small-scale tests that showed a composite tank could hold LH₂, but we’re the only company or organization that I know of that's actually demonstrated a full composite tank with fuel delivery system that can actually operate on a continuous basis.”

Fabrum had to overcome major challenges, including thermal shock, H₂ leak tightness and vacuum leak tightness, explains Reynolds, “where you don't want the vacuum [in the space between the tanks] to decay. And then you have to be able to build such a system and make it affordable.”

“The one thing we didn't have already was the ability to do a fast fill,” he continues. After identifying the issues involved, Fabrum proceeded with a solution that resulted in its patented triple skin design. “We have a liquid containment vessel inside the pressure vessel made from a particular construction to handle the thermal shock during fast filling,” explains Reynolds. “But that containment vessel doesn't have to handle the pressure loads, and therefore you've decoupled liquid containment and thermal shock from pressure capability. That was key and also provides redundancy should any incident cause loss of  vacuum in the outer shell. You use vacuum there because it's the lightest weight, most effective insulation mechanism. But if that gets compromised, then you get a very high heat load onto your inner vessel, and the LH₂ is going to boil very rapidly.”

“The triple skin decouples that again and only 25% or less of the heat load gets into the LH₂, which means the rate we would have to vent in an emergency to prevent pressure buildup is substantially reduced. Our testing with LH₂ in real operating conditions shows that if you lost vacuum in flight, there's actually no issue with continuing to fly. That was a key milestone as we move toward certification of our systems for several drone and small aircraft.”

A custom-built Fabrum double-skin LH₂ composite tank (left) for AeroDelft in the Netherlands, sits alongside Fabrum’s existing double-skin (center) and triple-skin (right) LH₂ composite tanks. Source | Fabrum LinkedIn post

Composite LH₂ tanks without carbon fiber?

Fabrum has used carbon fiber, “but through our years of work with superconductive systems, we learned you've got to be very careful about system longevity,” he notes. “Carbon fiber’s high stiffness can cause matrix cracking during the cool down process for these cryogenic systems. So, we specifically design our laminates with other materials to avoid that problem.”

Fabrum’s tanks are fully composite with composite internal support structures, says Reynolds, “but use metal fittings and a special joining system that lets us connect them reliably to the composite. We’ve also developed some special manufacturing techniques, because if you can't build it economically, then it's not worth anything. These techniques provide very high vacuum tightness where we don't get leaks through the laminate. For example, carbon fiber is notorious for having voids down the fiber bundles. Because those bundles are so fine, getting resin into all of the thousands of tows is very challenging.” Instead, Fabrum uses glass fiber with epoxy resin in a particular way that has been proven to work over decades of development and refinement.

LH₂ tanks for aircraft must be composite and affordable

The fast fill operation demonstrated in late 2025 used a 180-liter tank sized to store 8-10 kilograms of LH₂. Fabrum has now designed two slightly larger wingtip tanks, each storing 10-20 kilograms, for a general aviation plane that seats up to six passengers. “This size of tank is what we're currently fitting to helicopters, vertical lift aircraft and autonomous aircraft,” says Reynolds. “We did a short study on the aircraft and found that if we use twin 20-liter tanks, by the time you account for reserve fuel allowance, we can get 1-1.5 hours of flight time.”

Compared to the systems it has developed for mining vehicles and ground applications, the tanks for aircraft are small. “But we’re using the same technology and learnings across these applications and already have the plans in place to convert the inner vessels of the mining system to composites, which will probably save about 1.5 tons of mass per storage vessel.”

For aircraft, notes Reynolds, composites are the only option precisely because of this weight savings. “We’ve already seen some European aerospace groups abandon their aluminum tank developments because the weight penalty is too high, and there is also a thermal penalty, while composites can be good insulators, depending on the materials and construction used.”

But are these composite LH₂ tanks really affordable? “They have to be,” says Reynolds. “We come from an industrial background and understand what our customers’ cost targets. We’re not trying to reinvent anything or use aerospace costing, and not using carbon fiber helps with that.”

Boil-off and dwell time

Key issues for all LH₂ tanks used for mobility are how to manage temperature and pressure during operation but also when vehicles sit idle. How do Fabrum’s composite LH₂ tanks compare in terms of boil-off and dwell time? “When they're being used and you're drawing the LH₂ fuel off, you don't have any issues because that boil-off energy is going out into the fuel cell or gas turbine and you can maintain pressure,” says Reynolds. “We tend to operate at design pressures of less than 12 bar. The issue for everybody is when you're not drawing fuel off but just sitting there. In that situation, you need our system’s very low heat leak.”

Fabrum has demonstrated a dwell or idle time of 20 hours for its composite tanks before a pressure is reached that requires venting of the boiled off H₂ gas. “But we expect that to reach 40-60 hours,” he says. “So much depends on the detailed design and what is actually required because tanks with a longer hold time could be undesirably heavy and/or costly.”

Another issue often discussed is the need to insulate all of the LH₂ fuel lines. “Regardless of whether you take liquid, gas or a mixture off the tank, you need to warm it up to the temperature required by the fuel cell,” says Reynolds. “So, the cryogenic H₂ has to go into a heat exchanger of some type and we always insulate the lines to that heat exchanger. For all of our commercial systems — including the mining applications — that insulation is fairly easy to achieve and while they're running, it stays above freezing point.”

Development timeline and certification

 

AMSL Aero’s Vertiia eVTOL (top) and Stralis Aircraft’s fuel cell propulsion retrofit for general aviation aircraft (bottom) are using LH₂ and Fabrum’s composite tanks to offer significantly longer range for zero emissions flight. Source | AMSL Aero, Stralis Aircraft

Fabrum expects to be flying three different aircraft with a composite LH₂ tank within the next 12 months. It’s working with AMSL Aero (Sydney, Australia) and its Vertiia H₂-electric vertical takeoff and landing (eVTOL) aircraft and with Stralis Aircraft (Brisbane, Australia), supplying the LH₂ fuel system for its H₂-electric propulsion being certified as a retrofit for the Beech Bonanza. “We have a couple of other partners, including a helicopter company, that are looking at converting to LH₂ from their compressed H₂ gas systems,” notes Reynolds.

The next step is small commuter aircraft with up to 19 seats, which requires a significant amount of work for certification, but much less than aircraft with 70-100 seats. “That’s why ZeroAvia and a lot of other companies developing H₂ propulsion and aircraft are working under 19 seats,” says Reynolds. “Actually, developing the LH₂ tank is only a small part of the work required to achieve certification. The majority is all the other parts of the systems, including the fuel cells, and you have to have multiple redundancies on everything. We've designed our tanks with the certification process in mind, and we’re also supplying the whole system. We build the tanks, the heat exchangers that warm the LH₂ and the system that supplies pressure- and temperature-controlled warm gas to the fuel cell or combustion engine. The customer is typically the integrator and looks after everything from the warm gas onwards.”

Source | Alliance for Zero-Emission Aviation

Reynolds believes the first step of certifying smaller aircraft can be achieved by 2030, which is in line with the “Roadmap for the deployment of hybrid, electric and hydrogen flights in Europe” published in April 2026 by the Alliance for Zero-Emission Aviation (AZEA), which has more than 200 members including Airbus, Aciturri, Aernnova, Daher Aerospace, GKN Aerospace, IATA, Leonardo, MTU, Rolls-Royce, Safran, major airlines and airports, industry associations, regulatory bodies and more. That roadmap is targeting entry into service by commercial airliners for up to 100 passengers by 2040 and 20,000 hybrid, electric and H₂-powered aircraft by 2050.

He notes Fabrum has already started looking at certification, “because there's no point going down a technology path that can't be certified. We have a team that's already certified Jet A1 systems for Airbus, and we're taking that knowledge and that approach into the commercialization of our LH₂ fuel systems today.”

Tanks for space, thermoplastics, licensing for high volume production

The other industry that has a long history in using LH₂ tanks is space launch vehicles and they often do use carbon fiber. “The thermal cycling is nowhere near the same,” says Reynolds. “We’ve done work around these applications and even with reusable rockets, you might see 100 or 200 cycles. But that’s at the other end of the spectrum from an aircraft operating every day for years. The mission profile and specification for the equipment is just totally different.”

What about using the toughness of thermoplastic composites to fight microcracking when using carbon fiber at cryogenic temperatures? “I see a lot of challenges with using thermoplastic polymers, and while I think those are solvable, the question is whether they are able to be certified for commercial passenger aircraft. What you can do in a lab is very different than what is required to certify an aircraft fuel system for flight and put it into commercial service. So, at the moment, we’re using thermoset composite technology — not because it can't be done another way, but because the technical challenges involved and compliance pathway for those alternatives are substantial. We don’t believe carbon fiber is the right material for the primary LH₂ containment.”

According to the AZEA roadmap, the demand for lightweight, affordable composite LH₂ tanks could increase significantly over the next 10-15 years. Will Fabrum need to change its approach to deal with higher production volumes? “We recognize that we're not going to supply and ship from New Zealand at a large scale,” says Reynolds. “In our work with Tier 1 companies, we've discussed licensing to set up contract manufacturers where needed. Our systems are not dependent on expensive or exotic materials but instead use clever design and processes that are also practical and affordable. There are some things we do that I don't believe anyone else in the world does, but we can transfer that technology to licensed partners, who also won’t need to be aligned with a major carbon fiber supplier.”

Fabrum composites beyond LH₂

Fabrum does a lot work that uses composites but is unrelated to LH₂. “We're not a company that produces a set group of products,” says Reynolds, “but mainly have technologies that we apply to a wide range of areas.” He notes a customer in Europe that builds magnetic gearboxes for deep sea submersibles and other applications. Featuring embedded ferrites, the gearboxes hermetically seal the drive system and are impervious and leak tight at kilometers of ocean depth.

Fabrum has developed a wide range of superconducting technology using composites that handles thermal shock cycling, impulse pressure and vacuum with low heat loss and thin-walled construction. Demonstrated in generators, motors and transformers, its composite architecture is well-suited for aviation, space and industrial applications. Source | Fabrum

“We also build leak tight helium systems for companies working with electric motors, including production of motors for MagniX (Everett, Wash., U.S.).” Fabrum also works with superconducting systems, where their non-magnetic composite containment is used for devices that use a powerful magnetic field to confine plasma for nuclear fusion. “They all need high field magnetics and use a lot of composites for their support structures,” says Reynolds.

“We've also built superconducting transformers,” he continues. “We’ll supply those for a superconducting train project we’re working on as well as levitation cryostats. I'd say we're the only commercial company worldwide still in operation that builds superconducting equipment out of composites.”

Notably, Airbus has included superconductive systems as part of its technology roadmap for its ZEROe H2-powered aircraft. “Superconductivity is the next revolution,” says Reynolds, “ which will gain adoption once superconducting tape can be made affordably. When that happens, there will be a need for all these composite systems we make.”

These do use a bit of carbon fiber to dissipate static charge, but they mostly comprise glass fiber, says Reynolds. Here also, thermosets are preferred, he says, “because thermoplastics have a much higher coefficient of expansion, so, they're not as dimensionally stable, which is critical for the precision required in these applications.” The company will use a range of materials, including aramid fibers, titanium and other metals, and has developed broad in-house manufacturing and machining capabilities.

Derisking LH₂ tanks for the industry

If Fabrum has demonstrated composite LH₂ tanks that can be filled, emptied, refilled as well as sit idle and perform in flight, why are so many groups in Europe still developing their own LH₂ composite tanks? “Those projects support local industries,” says Reynolds. “We understand that and now have a team in Europe. It’s a challenge, for sure. But I think we also offer the ability to help derisk these systems. We’ve got an end-to-end solution in place and that's valuable because, especially in LH₂, if you don't do the right thing at each stage, you hamstring the following stage.”

He gives the mining trucks as an example, where Fabrum went from no one doing this to a system using up to 600 kilograms of LH₂ to power a 1.2-megawatt fuel cell with complex requirements on how it was dispensed. “We went with metallic tanks at that time, but as I said, we’re now ready to replace those with composites.”

Regarding how Airbus and the aviation industry is going to meet 9G crash load requirements, Reynolds readily acknowledges that carbon fiber will be needed. “Our approach has been to focus on the essential part that had to be proved, which is handling the LH₂ fuel,” he adds. “There is knowledge in the industry that can solve the other issues with structural integration, etc. But getting the LH₂ fuel handling to work as needed, reliably and with an affordable, lightweight system — that was key. And we’ve now proven that with a composite tank for small aircraft but we also know how to transfer half a ton of LH₂ in the shortest time possible with the least losses. We’ll continue to scale these systems for efficient refueling and zero emissions propulsion and also continue to advance certification.”

The DEMEX module mounted on a CEAD robotic extruder at the Institut für Leichtbau (Institute for Lightweight Engineering), University of the Armed Forces Munich, actively printing a thermoplastic part with LED-assisted interface heating. Source | LEAM Technologies

Large-format additive manufacturing (LFAM) has recently earned a genuine commercial foothold in composite tooling, structural prototyping and fixture production. The ability to deposit large-scale thermoplastic structures in tens of hours without massive tooling investment and with near-total geometric freedom is an enticing proposition that aerospace, energy and maritime manufacturers have found difficult to ignore. However, the technology has been limited by interlayer bonding strength that the industry has struggled to address.

The mechanism behind this limitation has reached nemesis status for many AM engineers, even though it’s part of the fabric of the manufacturing technique. Extrusion-based AM deposits molten material bead by bead, layer by layer. For the part to hold its shape, each deposited layer must cool and solidify quickly enough to bear the weight of everything above it. But layer bonding of one bead to the next is a thermally driven molecular diffusion process. Polymer chains need heat and time to migrate across the interface and physically entangle. The same rapid cooling that keeps the part standing actively prevents those chains from completing that migration. 

The result is a material that is mechanically strong in the X and Y axes, along the extrusion directions, but significantly weaker across the Z-axis, through those interfaces. In high-performance, semi-crystalline thermoplastics and carbon fiber-reinforced composites, where crystallization kinetics are especially demanding, this anisotropy can be severe enough to disqualify LFAM from structural end-use consideration entirely.

Bonding quality impact on cost, scale

The industry’s response to this problem has, until recently, focused on the material rather than the process. Specialty AM-grade polymers are formulated to slow crystallization, extending the bonding window and giving polymer chains more time to diffuse across each interface. This works up to a point, but it comes at a significant cost. 

In industrial-scale applications, AM-grade thermoplastics can reach €8 per kilogram versus as little as €0.50 per kilogram for equivalent standard injection molding grades. Beyond cost, even specialty materials leave operators constrained by layer time strategies, essentially limiting print speed based on how long a given geometry needs to cool. 

On large parts with long layer perimeters, the substrate can drop far below its optimal bonding temperature before the nozzle completes a full circuit and returns. As parts grow larger, the problem worsens. A mold for an aircraft structure with a 5-meter perimeter on each layer, for example, faces inherently longer cooling intervals than a small bracket, meaning the very applications that would benefit most from LFAM’s scale capability suffer most from its thermal limitation.

This is the gap that LEAM Technologies GmbH (Munich, Germany), a spin-off from the Chair of Carbon Composites at the Technical University of Munich (TUM), set out to close. Founded in 2023 by Patrick Consul, Benno Boeckl and Ting Wang, and building directly on doctoral research into LFAM thermoplastics processing, the company’s thesis was that the layer bonding problem was fundamentally a thermal control problem. If the thermal environment at the precise moment and location of deposition could be managed accurately, the trade-off between bonding quality and dimensional stability could be decoupled rather than merely managed.

Light as a process variable

The print head viewed from below shows the ring of focused LED emitters and individual lens assemblies arranged concentrically around the central nozzle, which together generate the localized melt pool at the deposition interface. Source | LEAM Technologies

The solution LEAM Technologies developed, called directed energy material extrusion (DEMEX), is an add-on module that mounts near the nozzle on an existing large-format extruder and travels with the print head. Its core action is focused, high-power LED radiation directed onto the substrate surface in a spot approximately 20 millimeters in diameter, positioned just ahead of the nozzle. This creates a localized melt pool roughly 0.5 millimeter deep, making it thin enough that the bulk of the material below remains cool, crystallized and structurally stable, but deep enough that the incoming bead meets a fully molten surface rather than a cold, semi-crystalline one.

“It is the difference between welding two polymers cohesively or trying to glue a hot layer to a cold one,” explains Patrick Consul, CEO and co-founder of LEAM Technologies. “In the former, both surfaces are in a plastic state and fusion across the interface is genuine; in the latter, adhesion occurs, but structural continuity does not. DEMEX engineers the welding/fusion condition at the extrusion interface, precisely and continuously, throughout the build.”

The choice of LEDs as the energy source reflects specific engineering priorities. Unlike infrared lamp systems, LEDs offer instantaneous power modulation which provides the dynamic response needed to track the rapidly changing thermal conditions of an active print, where the substrate temperature at any given point depends on part geometry, ambient conditions, elapsed layer time and deposition speed simultaneously. Unlike lasers, LEDs emit within the visible spectrum, which eliminates the need for costly laser safety enclosures and aligns the installation’s safety requirements with those of wire arc welding, a process most manufacturing environments already accommodate. 

“The total system cost is often lower than that of a typical laser safety cell,” Consul notes. “This makes integration not only easier but also significantly more economical.” The system uses three active emitters in a triangular optical arrangement, providing 360° coverage around the nozzle without requiring any orientation adjustment during computer-aided manufacturing path planning. It integrates through open communication standards and reads positional data directly from the machine controller without overwriting the program, making retrofit onto existing LFAM hardware straightforward.

Precision over the melt pool

What distinguishes DEMEX from a preheating attachment is the measurement architecture built around the emitters. Infrared cameras with fields of view wider than the generated melt pool monitor two zones at once: the melt pool being created by the heaters, and the incoming substrate surface just ahead of it. This simultaneous, dual-zone measurement gives the system real-time data on both bonding temperature and substrate stability, which are two competing demands, and allows independent setpoints to be applied to each. The system can raise the interface to bonding temperature while also using pressurized air to cool regions trending toward instability, managing both conditions in parallel rather than trading one against the other.

The LED emitter ring mounts directly below the extrusion mechanism, integrating thermal control hardware into the print head without adding significant bulk to the existing machine architecture. Source | LEAM Technologies

In practical terms, for a carbon fiber-filled thermoplastic, like Airtech (Huntington Beach, Calif., U.S.) Dahltram I-350CF polyetherimide (PEI) with 20% carbon fiber, the heater surface temperature targets approximately 360°C, close to the material’s recommended melt temperature, while the substrate setpoint sits near 250°C, around the glass transition temperature, where the material is stable enough to carry subsequent layers. 

All temperature measurements are logged against the tool center point (TCP) coordinates of the print head throughout the build, producing a spatially resolved thermal point cloud for every completed part. Each point in that cloud records what contact and substrate temperatures existed at that location during printing. Out-of-specification regions can be identified in real time, without cutting into the part or waiting for postprocess inspection.

This adaptive thermal control also enables meaningful productivity gains. “Because substrate temperature is a more robust process metric than elapsed time, it is insensitive to wall thickness variation, geometric complexity and part size. The system can push print speed closer to the actual thermal stability limit,” says Consul. “Using adaptive feed rate control alone, print times are approximately 10% shorter than conventional constant layer time approaches on equivalent geometries.” 

Combining active heating, adaptive feed rate and selective cooling reduces total print time by approximately 50% compared to constant layer time strategies, demonstrated on polyethylene terephthalate glycol-modified (PETG) and polycarbonate (PC) materials without measurable degradation in mechanical performance.

Measured outcomes

The mechanical data from early adopters establishes the magnitude of what DEMEX delivers at the material level. At the Institute of Lightweight Engineering at the University of the Armed Forces in Munich, researchers working with polyamide 6 (PA6) filled with 40% carbon fiber by weight (a grade widely specified in automotive structural components) recorded approximately 30% improvement in interlayer tensile strength compared to samples printed at the thermal stability limit without active heating. 

The untreated reference specimens exhibited an average tensile strength of 42.7 MPa, an elongation at break of 0.73% and an elastic modulus of 5,895 MPa. After a combination of active heating and controlled cooling, elongation at break increased to 1.04% (+42.5%) relative to the untreated reference and elastic modulus remains nearly unchanged at 5,783 MPa. “The most important thing about the elastic modulus is that the scatter or variability is reduced,” notes Consul. “The active heating and selective cooling combination achieved both the highest elongation at break and the highest stiffness across the layer interface of any test configuration.”

While this result appears contradictory, both properties were being suppressed by the same root cause: a poorly bonded interface failing in a brittle adhesive mode. Improving that interface with DEMEX doesn’t push stiffness and ductility against each other — it removes the defect that was limiting both, allowing the interface to resist deformation longer before failing, and then fail cohesively rather than by sudden delamination.

Thermoplastic deposited directly onto a previously printed surface during overprinting tests. Source | LEAM Technologies

In another test at the Netherlands Aerospace Centre (NLR, Amsterdam, Netherlands), investigations with a slow-crystallizing grade of polyaryletherketone (PAEK) from Victrex (Thornton Cleveleys, U.K.), often used in demanding aerospace structural applications, produced results that directly illustrate the severity of the problem DEMEX solves, as well as the extent of its resolution. 

“Without DEMEX, NLR could not complete hollow test specimens in unreinforced PAEK at all; internal thermal stresses at the layer interfaces caused delamination and cracking during the print itself, before the part was ever loaded,” explains Consul. “With DEMEX, setting the substrate target at 314°C yielded interlayer tensile strength of approximately 92 MPa against 96 MPa measured within the extrusion plane — near-isotropic performance, in a material that had previously been unprintable in this form.” 

DEMEX system at the Netherlands Aerospace Centre (NLR) prints thermoplastic material, with the LEAM Technologies LED heating glow visible at the deposition zone. Source | NLR

In fiber-reinforced PAEK, consistent, near-full Z-direction strength was achieved at substrate temperatures around 300°C, representing close to a 100% improvement over unheated reference samples. Critically, the failure mode in heated specimens shifted from brittle adhesive separation at the layer interface to cohesive ductile failure, where material from both adjacent layers tore together rather than simply peeling apart. 

That transition signals that the interface is no longer the weakest structural element, which is precisely the condition required for structural part qualification. At the system scale, the NLR demonstrated DEMEX capability by manufacturing a large-scale, aerospace-grade mold with a layer perimeter five times longer than their previous maximum — the first mold toolpath length was 2 meters versus LEAM Technologies’ ≈10-meter toolpath in its larger section — with full print stability throughout. Additionally, contact temperatures in some areas were increased from the original version, and despite the longer layer length, the final part showed no signs of delamination.

Structural ambitions

The implications for where this technology can go are potentially vast. “Composite tooling, autoclave molds, infusion tools and trim fixtures all already represent an established commercial market for LFAM, and it is exactly the application class where long layer perimeters make thermal management most difficult,” highlights Consul. “DEMEX resolves this at scale. The ability to process standard injection molding grades, rather than specialty AM formulations, opens a cost pathway that could bring LFAM into direct competition with compression molding for structural end-use components.” 

“The spatially logged thermal dataset generated during every DEMEX-equipped print represents a further strategic asset,” he continues. “As aerospace and defense qualification frameworks begin to develop standards specific to additive processes, the kind of continuous, location-specific process evidence DEMEX produces is exactly the form that process qualification audits will demand.”

This simultaneous, dual-zone measurement gives the system real-time data on both bonding temperature and substrate stability, which are two competing demands, and allows independent setpoints to be applied to each.

The overprinting work at the NLR where thermoplastic material is deposited directly onto carbon fiber composite panel substrates and the carbon fiber/PAEK validation results together point toward continuous and long fiber co-processing as a logical next step. DEMEX has already been exercised at the most demanding end of the short fiber spectrum and has shown it can form a structural bond at the interface between an extruded thermoplastic and an existing composite surface. Integrating continuous fiber systems into that process is the next frontier, and the existing work is the foundation it will build from.

DEMEX will be further tested with the Manufacturing Technology Centre (MTC, Conventry, U.K.), an independent research and technology organization that bridges the gap between academia and industry by developing and proving innovative manufacturing processes. The MTC plans to integrate DEMEX into its LFAM cell in 2026, alongside data logging and in-process monitoring upgrades. 

The capability to overprint stiffening features like ribs, flanges and reinforcing elements directly onto thermoplastic composite panels opens a further manufacturing route that bridges LFAM with the structural composite world in ways no conventional process can replicate. “We’re hoping that this year, the first of those applications will actually transition into serial production,” says Consul. “Most likely in the energy and maritime sectors, two markets we identified as the nearest-term candidates for serial production transition in 2026.”

A large, printed curved hull component from recycled polypropylene at IMPACD, a manufacturer of large-format parts for professional maritime use. Source | LEAM Technologies

The adoption in marine is already underway. IMPACD Boats (Woudsend, Netherlands), is pioneering the use of DEMEX technology to produce structural maritime components and complete boats for professional applications. By using recycled polypropylene, IMPACD improves sustainability for its large-format prints that are ready for immediate use on the water. Though not using composites, IMPACD has proven the  business case, and is part of a Dutch ecosystem — including CEAD, Dutch Boat Factory and other companies — that is advancing 3D printing in a range of recreational, commercial and defense vessels. “We’re not running a pilot,” says Marieke de Boer, CEO and co-founder of IMPACD Boats. “We’re creating applications that didn’t exist before.”

For additional margin, shorter pi joint segments were installed under the upper wing skins as well, where the wing dihedral transitions to a level dog-bone-shaped laminate that bridges across the aircraft cockpit. Source (All Images) | DarkAero Inc.

The DarkAero 1 is a prototype high-speed, long-range, experimental kit aircraft designed and manufactured by DarkAero Inc. (Madison, Wis., U.S.). Development includes an extensive load test campaign to verify the all-composite airframe’s structural integrity prior to high-speed flight testing. Flight test campaign planning and ground tests identified that the center wing box butt joints and flap control system mount rigidity would require modification prior to flight. A new solution was needed.

The DarkAero 1 center wing box now relies on pi joints to stabilize the center wing skins while under compression. The challenge of this structural retrofit created an engineering playground riddled with rigorous design constraints from initial conception to in-house static load testing of the full wing assembly.

Note from Ginger Gardiner, CW executive editor: This technical article was written by Ryan Stube, chief engineer for DarkAero, and is published mostly as he submitted it. However, I have also inserted short segments from my initial discussion with him that, to me, also tell the human side of the engineering story, which I find fascinating. Also, watch the video near the end, which features Stube, and is why I reached out to DarkAero in the first place.

Who is DarkAero?

The DarkAero 1 prototype aircraft prior to beginning a thorough ground test campaign.

Ginger Gardiner (GG): It started with the idea of building the fastest, longest-range composite aircraft you can build in a garage. The founders — three brothers with degrees in aerospace, mechanical and electrical engineering — wanted to fly from the Midwest to either coast as a weekend trip. That evolved into the DarkAero 1, a prototype side-by-side, two-seat aircraft aimed to cruise at 275 miles per hour with a 1,700-mile range.

Stube explains that one of the brothers was building an experimental aircraft from a plans-built kit, which required a lot of fabrication knowledge and thousands of hours of work starting from raw materials. A more appealing alternative is known as a quick build kit where builders bond subassemblies together straight from the factory. Thus, DarkAero started out as an experimental kit aircraft company, with the plan to supply parts and subassemblies for composite aircraft to individual builders.

However, through developing the DarkAero 1, the company’s skills, capabilities and team grew, says Stube. “We now teach aerospace composites manufacturing and moldmaking courses. That led to working with students on a variety of projects, including planes, boats and cars. We also provide services where we help with design, build tooling or even fabricate entire airframes. This contract work has led to continual growth, and the company is always looking for driven individuals to join our team.”

Wet wing design, butt joint assembly

Ryan Stube (RS): Although not required by governing regulations for experimental category aircraft, DarkAero’s internal standards are more reflective of a certified aircraft program. Development of the DarkAero 1 has been working through an extensive load test campaign to ensure adequate structural integrity of the all-composite airframe prior to flight testing. In parallel, the flight test campaign and respective flight envelope are being formalized.

GG: The speed and range requirements for the DarkAero 1 drove the wing to be somewhat of a new approach, “because we focus on manufacturing just as much as design, targeting an empty weight of 750 pounds — roughly half of what a comparable size Lancair would weigh,” says Stube. The plane’s fast cruise also places a lot of aerodynamic loads on the structure.

RS: The prototype wing with 23-foot, 5-inch wingspan is an all-bonded, carbon fiber-reinforced composite assembly that weighs in at just over 100 pounds. The original stressed skin design relied primarily on the skins themselves and then secondarily on a hollow grid internal structure to prevent skin buckling. However, instead of having a bunch of individually molded ribs — which would require a lot of labor to produce — we use honeycomb-cored panels made in-house from 4 × 8-foot sheets that are CNC routed into the 2D shapes we need for both the wing and fuselage. We then bond these together using controlled surface preparation, epoxy paste adhesive and a proprietary assembly process to form a grid of simple butt joints.

The DarkAero 1 center wing section showing curved region (red lines) where airfoil cross-section blends into the fuselage body (top left). Red arrows show the primary in-plane loading of the lower wing skin under negative g-loads and the respective secondary out-of-plane loading condition (bottom left). This secondary condition comes from the lower wing skin having eccentrically loaded curved panel geometry near the wing root which places the center shear web joints in tension. FEA, for illustration only, shows an aft view of the pilot side wing root under the same loads (right). The red arrow indicates where the lower wing skin deforms away from the rest of the structure.

The wing is also a 77-gallon fuel tank spanning from the wing roots to tips. The lower wing skin is a single part — from wing tip to wing tip — while the upper wing skins are two separate parts with a dog-bone-shaped central member transferring loads between them in the fuselage/center wing box. The lower wing skin also forms the cockpit floor and skin of the fuselage below the center wing box. This means the lower wing skin geometry seamlessly transitions between an airfoil profile and the fuselage profile near the wing root. The hollow grid structure of bonded honeycomb cored panel butt joints to the skins continued through the center section of the wing box via five honeycomb shear webs. With this structure, the wing passed initial positive and negative-g static proof load testing.

Necessary redesign in center wing box butt joints

RS: However, as mentioned above, in formalizing the flight test envelope, it became apparent that we needed to increase the wing’s ultimate lift load strength to resist vertical gusts, which presents an equal probability of abruptly increased lift in both the positive and negative directions. Increasing the proof load value meant the center region of the wing would experience higher secondary bending stresses primarily due to the inherent curvature of the lower wing skin. These stresses induce out-of-plane loading during negative-g lift conditions while the lower wing skin laminate is in compression. Having an aerodynamically clean transition between the wing and fuselage helps reduce drag, but it creates a difficult structural problem when using an all-bonded, stressed skin design with a locally curved load path.

GG: There is curvature in this lower wing skin panel without any inherent stability. “It’s basically an unsupported panel,” notes Stube. “So, our problem became how much do I need to support this eccentrically loaded panel? It’s curved, and that problem is always hard to analyze and hard to define, because even though the material itself could handle the tensile and compressive loads that it will see in flight, it’s eccentrically loaded and wants to buckle.” So, internal structural resistance to out-of-plane loading was still needed to prevent that buckling from occurring.

RS: Through the structural deformation modes observed during previous static wing load tests, we recognized that the first failure mechanism of the wing at higher loads was the out-of-plane tensile strength of the butt joints between the center shear webs and the lower wing skin where the fuselage body curvature blends with the wing airfoil. Outboard of this central region, the upper and lower wing skins are nearly parallel to each other and do not have inherent buckling tendencies or instabilities. Therefore, modification to pass additional proof load testing was focused on the center region of the wing and a new joint method between the lower wing skin and the center shear webs.

Pi joints as a solution

GG: Butt joints were originally chosen because it’s an easier manufacturing technique when you know you have an appropriate margin of safety, says Stube. But they also knew there are stronger joints than a simple butt joint.

 

Cross-sections of butt joint (top) and pi joint (bottom) element test specimens prior to pull testing in a universal test machine. The honeycomb panels and base laminates replicate the existing materials in the original DarkAero 1 prototype wing structure. 

RS: Pi joints more evenly transfer the out-of-plane loading conditions into double lap-shear joints. This is ideal, as paste adhesives for assembly are much stronger in shear than peel or tensile loading. With the requirement of maintaining the honeycomb shear web architecture to preserve existing structural analysis efforts, pi joints tailored to the honeycomb panel dimensions showed a potentially promising solution to the problem.

The DarkAero 1 had not previously used pi joints anywhere in the airframe due to the more labor-intensive manufacturing methods required to produce the joint geometry. The aircraft primarily relies on vacuum-assisted resin transfer molding (VARTM) and unsupported post-cure of complete composite assemblies. With the budget-constrained resources available when the original prototype airframe was built, complicated composite parts were reduced to a minimum, and the aircraft was designed with low-cost, high-quality manufacturing in mind. Developments in room temperature storage, out-of-autoclave (OOA) epoxy prepregs now allow for more complicated prototype composite part fabrication while still maintaining relatively low cost.

One of the first steps toward implementing pi joints in the DarkAero 1 wing was to perform a brief feasibility study, laying the foundation for the entire structural retrofit. Pi joint manufacturing techniques, joint strength characterization, detail-level test specimen laminate changes, existing wing structure deconstruction, new structure installation and assembly post-cure were all worked through prior to seriously pursuing the pi joint structural retrofit. With a rough but feasible path forward in mind, initial structural characterization tests started with simple element-level joint pull tests in a universal test machine.

As-manufactured element laminates for load testing

GG: Although DarkAero hadn’t made pi joints for this exact use case, it had used the intended materials and manufacturing techniques for other applications. “At the end of the day, we needed to make sure that we could not only build them but also create an assembly that worked for this redesign and then we had to test that assembly,” says Stube. “That was why we took the physical approach of ‘build it and break it.’ Then, we just followed the basic building block pyramid to move up the scale in specimen testing.”

RS: Although DarkAero has finite element analysis (FEA) and computer simulation capabilities, because the wing redesign would still require physical test specimens to be manufactured and tested, it was more efficient to rely more heavily on real-word data from a limited physical test program. Preliminary test samples were manufactured using variables specific to the DarkAero 1 wing. The pi joint base laminate would be secondary bonded with an assembly paste adhesive to the existing prototype lower wing skin — a resin-infused, post-cured plain weave spread tow laminate. To accurately simulate the substrate’s future surface energy properties during eventual pi joint installation, composite laminate sections for element-level testing were manufactured using the same fabric, infusion epoxy and post-cure as the actual lower wing skin.

With composites, the final assembly strength properties are not only determined by the physical materials being used, but also the exact fabrication and joining processes as well. Preliminary plans relied on vacuum bagging to introduce clamping pressure while the assembly adhesive cured during pi joint installation into the existing center wing section geometry. With the original center wing box structure removed, the lower wing skin mold was used to serve as the support fixture for maintaining the left and right wing dihedral and incidence angles.

Although intricate clamping fixtures could have been built to interface with the mold structure, it was more efficient and reliable to use the pressure differential of a vacuum bag to apply uniform clamping pressure during assembly. Even though element-level test samples are typically small and, in this case, could have been bonded together using simple hand clamps, the intended vacuum bagging method was still used to produce the samples. This means the initial database testing would more accurately simulate the uniform bondline thickness and environmental-related strength properties of what the wing structure would likely have due to curing under vacuum. Initial element-level testing showed promising out-of-plane strength increases compared to estimated new joint strength requirements and provided enough assurance to further pursue pi joints.

Advancing pi joint design, multiple performance improvements

RS: Instead of heavily relying on new computer simulations to instruct joint build parameters, the original wing design model and new approximate calculations were used to loosely guide the joint design toward the new strength requirements. Additional element testing included characterizing the sensitivity of the pi joint configuration to foreseeable potential manufacturing deficiencies as well as laminate design refinement. Variables such as honeycomb panel shear web bondline thickness and clamping pressure or joint dimensional tolerance and fit-up were quickly tested to create approximate acceptance criteria for the prototype wing components. Laminate design parameters including the unidirectional (UD) noodle cross-sectional area and ply orientation sequencing were varied to briefly refine the joint performance.

At the conclusion of the element-level testing, the out-of-plane first failure strength of the pi joint between the honeycomb panel and wing skin increased by more than 300% compared to the original butt joint ultimate strength. This was not the only performance improvement, however.

The butt joint test specimen failure mode was a subtle first indication of failure, with very limited strength and stiffness decrease and within about 85% of ultimate failure, which occurred via cohesive failure at the lower substrate or failure in the lower laminate. In contrast, the pi joint test specimens experienced a first failure through delamination near or through the UD noodles. This first failure showed a brief decrease in reaction load due to the resulting increased elasticity and then continued to withstand lower levels of out-of-plane loading while additional failure occurred. Ultimate failure of the pi joints had higher variance than their first failure, but still measured more than 250% higher than the butt joint ultimate strength. Nearly all pi joint ultimate failures were via delamination between the upright and pi joint base laminate plies instead of disbond failures at the bondline or surface of the simulated lower wing skin.

 

Multi-pi joint structures prior to installation in the center wing. These primary structural elements are manufactured using out-of-autoclave, room temperature storage prepreg. These had to be made in three separate sections to enable proper installation due to the physical space constraints in the center wing area.  

With the joint strength, materials and assembly methods characterized, development efforts transitioned to brief, detail-level tests of multi-pi joints. These are a closely spaced (2 inches apart center-to-center) series of parallel pi joints comprising the five center shear webs in the DarkAero 1 wing. With a further understanding of more central wing box assembly-specific load characteristics, the team manufactured, inspected and installed larger sections of multi-pi joints in the center wing.

Dimensional and manufacturing process constraints initially seemed to reduce the probability of successfully increasing the existing wing strength to near zero, but calculated development work and a building block approach allowed for design updates to be implemented with an adequate amount of confidence in the new design along the way. Full component static proof load testing of the rebuilt DarkAero 1 wing eventually verified the redesigned structure for a larger high-speed gust encounter operating envelope in both positive and negative lift load conditions. The stiffness increase of the entire wing assembly — how much the wingtips deflect under applied load — was exciting to record in real time. Although not a previous concern, the wing tip deflection per g was decreased by more than 25%.

GG: According to Stube, the finalized pi joint assembly did add a couple more pounds to the wing structure, but the increase it achieved in structural margin for gust loads was well worth it. The whole team was really excited that the pi joint redesign worked. “It seemed like this was an engineering problem that might not have a clear answer; there were so many conflicting requirements and tight constraints. We went from a butt joint to a pi joint assembly in a way that was also proving out the manufacturing, including adhesive application, bondline thickness and porosity, cure and just access for installation. We solved all of this AND it worked the first time.” And yet, the team is already developing the next iteration for when the DarkAero 1 actually enters production.

The modified DarkAero 1 center wing box structure dry fit could be performed outside of the wing structure. The curvature of the lower wing skin can be seen toward the ends of the lower pi joints. 

RS: The exact multi-pi joint design used for the prototype wing is not directly reflective of intended future production configurations, as the retrofit drove non-ideal design trade-offs for efficient fabrication. During production, where the structure can be manufactured with all assembly steps in mind from the start, the exact manufacturing techniques and joint geometry can be better optimized for reliability in higher volumes. For example, the pi joint geometry could be directly molded and co-cured with the lower wing skin laminate. Hybrid pi joint and shear web-like features could also be combined to eliminate secondary bonding steps, and the center wing box structure could be further blended into the neighboring geometry. Although the pi joints did show higher residual strength after initial damage occurred, further increase in the joint’s damage tolerance could be achieved through laminate stitching or tufting, z-pinning or even pi joint-shaped 3D woven fabrics.

Whatever the engineering or manufacturing application may be, the DarkAero team tries not to get too reliant on any one specific process or material and continually refers back to the project requirements to drive design and manufacturing solutions. Sometimes, building physical hardware to get actual test data can be quicker and more accurate than computational analysis and this center spar modification was an interesting engineering problem to solve in this way.

GG: DarkAero is trying to provide innovative composite solutions in a more affordable way. “Most companies have used much more expensive methods to make pi joints that also required a long internal approval process,” says Stube. “We simply need to move faster with a tighter budget.”

So, stay tuned — CW will publish a future article on DarkAero’s novel approach for producing composite pi joints.

About the Author

 

Ryan Stube

Ryan Stube is the chief engineer at DarkAero Inc., returning to the company after being the first intern in 2019. He is currently leading the development program of the DarkAero 1 focusing on bringing the clean sheet designed prototype aircraft safely into flight testing. His previous aerospace experience includes leading Merlin rocket engine refurbishment at SpaceX, with specialization in thermal protection systems, turbopumps and engine testing. ryan.stube@darkaero.com

The Carbon Fiber Design project, led by national lab and university researchers, is working toward developing industry-ready, lower-cost carbon fibers with suitable compression performance for wind applications and beyond. This micrograph shows the project’s three lobe diameter (versus typically circular) carbon fiber filaments, which are optimized for performance and cost targets. Source | Oak Ridge National Laboratory (ORNL)

While carbon fiber composites have been adopted for a variety of applications, many research and industry efforts continue to work toward optimizing carbon fiber production further to meet cost, rate and/or performance needs for specific industries.

One of these efforts is the Carbon Fiber Design project. Since 2020, this team of researchers — led by Oak Ridge National Laboratory (ORNL, Oak Ridge, Tenn., U.S.) with Sandia National Laboratories (Sandia, Albuquerque, N.M., U.S.) and Montana State University (MSU, Bozeman, Mont., U.S.), and funded by the U.S. Department of Energy (DOE) — has been working toward the development of cost-efficient carbon fiber that is performance-optimized initially for wind industry applications. As the project progresses, the researchers seek industry partnerships to scale the current results toward commercial products and applications in the wind market and beyond. 

“We started with the hypothesis that a more optimal carbon fiber can be produced for cost-driven applications, with wind energy being the initial focus,” explains Dr. Brandon Ennis, materials and design lead, Aerodynamic Technology & Energy Systems department at Sandia. “Cost can be prohibitive for using carbon fiber in many wind turbine models, but despite this wind blades are the highest-volume application for commercial-grade carbon fiber.”

ORNL researchers (Bob Norris pictured in back, Fue Xiong in front) use current lab-scale wet spinning line to produce PAN tows for manufacture into carbon fibers. Source | ORNL

ORNL researchers had already spent more than a decade working on projects related to reducing the production costs — and ultimately the market price — of carbon fiber to enable wider adoption. This has included evaluation of carbon fiber made using lower-cost, readily-available textile-grade polyacrylonitrile (PAN) precursor, as well as work on optimizing carbon fiber conversion processes including use of advanced plasma oxidation (APO) in partnership with 4M Carbon Fiber (Knoxville, Tenn., U.S.).

Through these investigations, “we found that we could produce fibers around 50% cheaper [compared to current commercial fibers],” Ennis says, due to a combination of cheaper precursor and increased production throughput resulting in reduced production costs.

However, in terms of performance, “the compression strength was 20-30% lower [than the standard],” he says. For the goal application of wind blade spar caps in particular, this posed a problem. 

Bob Norris, distinguished R&D staff member at ORNL, explains that in wind turbine blades, “carbon fiber is used as a structural spar cap element, which functions like an I-beam or a spar in an aircraft wing, where you have the caps on either end, and one side is in tension and the other is in compression. In this application, you really need to make the compression more like the tension to get the full value of carbon fiber.”

The Carbon Fiber Design project, then, set out to leverage and expand the previous work ORNL and its partners had already done toward producing lower-cost fibers, with a focus on increasing compression strength.

This came down to optimizing the shape and size of individual carbon fiber filaments, as discussed in depth in the project’s most recent publication, “Initial assessment of alternative carbon fiber geometries for design of cost-effective compressive performance: Size effect studies.”

Carbon fiber cross-sectional geometry: From circular to multi-lobed

“Commercial carbon fibers [used in the composites industry] are pretty much universally circular in cross-section,” Ennis explains. “It’s a simple shape, and it represents an industry focus on tensile strength. Because of the symmetry, circular fibers have more inherent processing consistency as well as favorable surface quality that has a bigger impact on tensile strength,” which is essential for many aerospace applications that carbon fiber composites are used for.

That’s where ORNL’s previous textile-grade PAN-based carbon fiber work comes in. Fibers produced for use in textiles, which are manufactured using different spinning methods than those used to make industrial carbon fiber, generally have noncircular cross-sections. In working with textile-grade fibers, researchers began to see the correlation between the fiber shape — which was, in this case, “a sort of kidney bean shape,” Ennis says — and the final product’s compressive strength.

Local diffusion thicknesses of a circular fiber versus a three-lobe fiber, each with the same cross-sectional area of 40 microns². Source | Sandia National Laboratories (Sandia)

“The reason we’ve hypothesized that noncircular fibers could be better in terms of compressive strength is that the fiber itself has a higher bending resistance compared to a circular fiber with the same cross-sectional area,” he says. “In a filament, there is a geometric property called the ‘area moment of inertia,’” which essentially makes the fiber more resistant to microbuckling, that can in turn lead to compressive failure in a composite. The noncircular shape presents more resistance to microbuckling. “The hypothesis is that you have higher area moments of inertia, which can delay the onset of failure and potentially improve fiber alignment in the composite, resulting in a higher compressive strength.”

In the case of the textile-precursor fibers, the asymmetric kidney bean-shape fiber diameter “showed some benefits, but wasn’t optimized. Additionally, asymmetrical shapes tend to introduce uncertainty, causing higher variation in the resulting composite properties,” Ennis adds.

Next, researchers conducted an analytical study to assess the performance of fibers manufactured with different numbers of “lobes” of various shapes seen in each filament’s cross-section. The conclusion was that three- and six-lobed fibers show the most promise in terms of cost and compressive strength, especially for applications requiring high fiber volume fraction.

Using these findings, researchers most recently looked at two aspects in parallel: The effects of the size of the filament on compressive strength, and the best process for manufacturing three- or six-lobed fibers.

Sandia contributed modeling work, ORNL ran manufacturing trials and MSU performed materials testing, with the overall aim of consistently producing fibers with the desired lobed shapes. “It starts with designing holes in the spinneret die to produce the desired carbon fiber shape based on our modeling studies. ORNL has iterated in manufacturing trials to control for geometry and mechanical performance. The three-lobe fiber trial studies have even produced composites with a fiber volume fraction of up to 69%, which is higher than any of the circular fibers we’ve tested. It’s really promising,” Ennis says.

Results: Higher throughput, lower production costs, increasing compressive strength

The project’s manufacturing trials, material testing, simulations and cost models have led the researchers to pursue three-lobed fibers so far of up to 11 microns in effective diameter (most fibers are in the 5-8 micron range). Effective diameter, used for the measurement of non-circular shapes, is calculated as the equivalent diameter for a fiber’s cross-sectional area, Ennis explains.

Performance. “In the manufacturing trials at ORNL, we have observed that the three-lobe shape gives us the most robust geometry, which has been able to achieve fiber volume fractions close to 70% for the infused samples. If our observed correlation of compressive strength and fiber size holds as we increase in size, we expect to see slightly more significant increases in compressive strength for the geometries currently under development which have fiber areas that are two-and-a-half times the circular fiber and a bending resistance that’s around nine times higher,” Ennis says.

A spinneret custom-designed by ORNL to wet spin three-lobed PAN filaments. Source | ORNL

Faster production. In addition to achieving the desired fiber performance, the three-lobed fibers also have been estimated to increase the throughput on the carbon fiber line by two to three times that of a typical circular cross-section fiber.

“The rate-controlling step in converting precursor to carbon fiber is the oxidation furnace, and this is controlled by a specific metric of fiber dimension called the diffusion thickness,” Ennis explains. “In a circular fiber, the diffusion thickness is simply the radius. For a non-circular fiber, it’s the distance to a plane of symmetry. For a three-lobed fiber, the planes of symmetry bisect each lobe, and the benefit is that you can significantly reduce the diffusion thickness for the same cross-sectional area, and that increases the production line speed and throughput.”

He adds, “How fibers progress through the conversion line is constrained by the diffusion thickness, but what’s interesting is that circular fibers just geometrically have the highest diffusion thickness per area. If you were optimizing for this constraint, you would never choose a circular fiber. It’s the worst geometry for that one metric.”

As the size/area of the fibers is increased, the diffusion thickness matches up to that of a circular fiber — “in that case, you have roughly the same line speed, but because of the larger fiber area, the production throughput can be doubled for a conversion line.”

Lower-cost production. Compared to a baseline commercial-grade carbon fiber, “the project approach can reduce production costs by up to 50%, depending on the tow count,” Ennis says. In addition, the increase in throughput reduces energy consumption and therefore costs — “on the order of 40-50% reduction in energy cost during the conversion stage.”

Next steps: Continuing trials, scaling up with industry partnerships

What’s the next step for the Carbon Fiber Design project? Currently, Sandia is continuing numerical modeling studies, with ongoing efforts to look at other mechanical properties like transverse performance. Meanwhile, ORNL is optimizing and scaling up its wet spinning process — which involves partnerships with carbon fiber manufacturers interested in the technology and evaluation of company-specific PAN chemistry within future trials.

“There’s been hesitancy that we’ve seen in the past from the carbon fiber industry regarding changes to fiber or production characteristics, but we’re seeing a trend where there’s been increasing interest and hope to see this continue,” Ennis says. The project’s advisory board today includes carbon fiber manufacturers alongside wind blade manufacturers and others.

ORNL is working on both scaling up and automating the amount of these specialized carbon fibers it can make, in order to contribute larger test samples, as well as investing in new dies and equipment for producing larger fiber sizes to evaluate for even further performance increases.

“The goal is certainly commercialization,” Ennis emphasizes. A separate DOE I-Corps program aims at development of a formal commercialization strategy. ORNL’s Norris adds that part of this work is bringing other applications to the table to drive the market pull for this approach. Potential areas include automotive leaf springs or bumper structures, pultruded infrastructure I-beams, drill risers and tension legs for offshore oil and gas, interior aerospace floor or storage structures and more.

Dr. Bill Carter, VP of advanced production automation technology at Boeing Engineering & Technology, delivered the keynote address at SAMPE 2026. Source | CW

If there was one phrase you couldn’t escape while walking the aisles, sitting in technical sessions or eavesdropping on hallway conversations at SAMPE 2026, it was this: high rate. From the opening general session, and through 3 days of panels, paper presentations and exhibition floor demonstrations, the composites community made clear that the central challenge of this moment is no longer whether advanced materials belong in aerospace and defense platforms — that question was settled long ago — but whether the industry can produce them fast enough, consistently enough and at sufficient scale to meet what is shaping up to be an extraordinary decade of demand.

SAMPE has always positioned itself as a place where academia, industry and government intersect, and this year’s edition leaned hard into that identity. Co-hosted by the SAMPE Seattle and Carolinas chapters, the program spanned more than 100 technical sessions, two panels and hands-on tutorials, with content stretching from fundamental research through scale-up to full production.

From the outset, the tone of the week was set by the keynote delivered by Dr. Bill Carter, VP of advanced production automation technology at Boeing Engineering & Technology, formerly of HRL and DARPA (where he ran portfolios in hypersonics, space manufacturing and scalable nanomaterials).

Carter opened with an image of a structure full of stringers, frames and floor beams — which turned out to be a ~1400 BC Egyptian funerary boat sealed with tensioned reeds. His point: The materials and process problems we wrestle with today — gap fill, tolerance stack-up, joining dissimilar materials, organizing humans around complex builds — are millennia old. What’s new is the urgency. From a defense perspective, in the midst of major conflict, existing inventory disappears in weeks, leaving manufacturing capacity as the decisive national capability.

Carter walked through his new, deliberately systems-focused organization spanning design support, digital production, robotics, process automation, 3D measurement, NDT automation and scientific imaging — built on the premise that production risk has to be understood end-to-end. Boeing’s approach involves a “hub-and-spoke” data architecture, autonomous error-correcting robotics and quantum computing for materials modeling, Carter explained. And all of these technologies are geared toward encouraging curiosity, taking risks and using modeling and the documented experience of others to streamline the advancement of technology.

Carter closed with a fear that keeps him up at night: a future major conflict where the U.S. industrial base can’t surge fast enough. According to Carter, working to streamline production processes to avoid this scenario not only help to safeguard us from such a vulnerable position, but also afford us the opportunity to fundamentally rethink aerospace production.

What does this mean for where the composites community sits in mid-2026? The technical focus of the industry has shifted to some degree. Fiber, resin and process innovation continue to be important — but the most momentum in the industry surrounds contextualized factory data, AI-assisted inspection, autonomous robotics, digital twins extended into the supply base and qualification pathways that support new technology maturation without holding rate hostage.

The aerospace and defense pull is unmistakable. High-rate commercial production ramps, autonomous and uncrewed platforms, hypersonics, space launch and a defense industrial base under acute pressure to demonstrate surge capacity have become the backdrop of nearly every high-rate composites conversation.

The cultural question of whether the industry can recover what Carter called a certain “audacity” seen in early spaceflight — a willingness to assume some level of risk-taking and failure as part of the process of innovation —  while honoring the safety discipline that aerospace requires is being forced to the forefront of how we get things done. It’s a reminder that events like SAMPE are a vital gathering place to put the right people in the same room long enough to make the next round of progress possible.