On 23 October 2026, the Fuling District government in Chongqing published details of a pilot line for 70 MPa Type IV hydrogen vessels wound dry from domestic carbon fibre. The reported figures are specific: roughly 20 percent less carbon fibre than the wet process it replaces, a measured burst pressure above 157.5 MPa, a maximum volume above 400 L, and a winding speed of 6 m/s. If you buy thermoplastic prepreg tape for hydrogen tanks, the vessel is not the interesting part of that disclosure. The interesting part is the words “dry winding”, because they move nearly the whole quality burden of the process upstream, onto the tape itself.
What the disclosure reported, and what it left open
The published facts, attributed to the Fuling District disclosure of 23 October 2026: the line uses prepreg to control impregnation instead of the traditional wet resin bath, the carbon fibre is domestically sourced, the process is described as fully self-controlled, and the liner-to-valve-seat joint is described as a molecular-level seal. The named application areas are hydrogen fuel cell vehicles, stationary energy storage and unmanned aircraft.
What the disclosure does not contain is equally worth noting. There is no statement of serial capacity, no cycle-life data, no batch-to-batch consistency data, and no qualification statement for any vessel part number. A pilot line demonstrates that a process can be run. It does not demonstrate that the same process runs identically on the three hundredth day of the year, which is the property a pressure vessel programme actually depends on.
That gap is not a criticism of the project. It is a description of the stage it is at, and it tells suppliers something useful: the tape specification being written right now, during pilot and trial work, is the one that will govern supply later. Getting written into it early is cheaper than displacing an incumbent afterwards.
Why dry winding turns a resin problem into a tape problem
In wet filament winding, dry roving is pulled through a resin bath and onto the mandrel. Resin content is a process output, set by bath viscosity, die geometry and squeeze-off, and it can be nudged while the machine runs. Excess resin is not a disaster, because it flows: it fills the gap between adjacent courses, wets out the fibres under tension, and levels small thickness errors. The mandrel is fed a compliant, self-correcting material.
Dry winding removes that self-correction. The tape arrives with resin already in it, at a resin content fixed at the impregnation stage. Nothing downstream can add resin or take it away. Fibre volume fraction is decided before the tape reaches the creel, and if a course does not sit where the winding programme expected, there is no flowable excess to fill the difference.
Three failure modes follow directly, and all three are tape properties rather than machine properties:
- Width becomes a gap and overlap number. A course that runs marginally narrow leaves a resin-rich channel between itself and its neighbour, and that channel is where voids and delamination start. A course that runs marginally wide builds local thickness into a ridge that the following pass cannot flatten. On a flat panel neither matters much; on a domed end, both do.
- Thickness errors accumulate instead of averaging out. If the nominal ply is 0.125 mm and the tape delivers 0.132 mm, twenty courses end up roughly 0.14 mm thicker than the model. That is a small number until you remember that the local radius of curvature of a dome changes with the stack, and local stress follows the geometry.
- Resin content drift changes the design case. A hoop-wound pressure vessel is designed around hoop stress, which depends on fibre volume fraction. If resin content wanders between batches, the same winding programme produces a different fibre volume fraction, and the vessel you test is not the vessel you calculated.
There is a fourth, less discussed issue: unwind behaviour. Tape has to leave the spool without blocking, without shedding fuzz at the slit edge, and at a tension the controller can hold steady. Fuzz is not cosmetic; a shed bundle entering the nip becomes a distorted course. Slit quality is a genuine specification item, and it is usually the first thing buyers forget to ask about.
Three specifications to write into the purchase order
If you are sourcing tape for a dry-wound vessel, these are the questions that separate a workable supply from an expensive experiment.
First, ask for tolerance as a process value rather than a nominal. Not “width 12.7 mm” but what the width does across a full roll, from the first metre to the last, and what the slit edge looks like. A supplier who can only quote a nominal is telling you the number was measured once.
Second, ask for resin content and areal weight per batch, with the test method named, on the certificate that ships with the goods. Then ask for two consecutive batches rather than one. One batch of data is a sample; two batches start to describe a process. The values that matter here are the same ones that matter in any thermoplastic prepreg: areal weight, resin content, and their spread.
Third, ask about handling and moisture, not storage temperature. Thermoplastic tape does not need the deep-frozen logistics of thermoset prepreg, and treating it that way adds cost for no benefit. What it does need is dry handling before consolidation, because surface moisture still generates voids at the interface. This matters more for some resin systems than others. PA6 picks up moisture more readily than PA12 or PPS, and that difference shows up in the process window rather than in the datasheet.
A fourth question is worth asking quietly: can the fibre lot and resin lot be traced from the roll back to the batch? In a qualification programme this is not paperwork, it is the only way to explain a test result six months later.
Where this goes over the next two to three years
What follows is our assessment, not a statement of fact, and it should be read as one.
The Type IV vessel is the volume segment in hydrogen because it stores gas at moderate pressure with a polymer liner rather than a metal one, which keeps weight and cost down for trucks, buses and stationary storage. Its dominant cost is carbon fibre. A process that reports roughly 20 percent less fibre consumption and a winding speed of 6 m/s attacks the two largest cost levers at once, which is why we expect dry winding to take share quickly in the simplest geometries — cylindrical bodies with domed ends, in high volumes and repeating part numbers.
We do not expect it to take over everything. Wet winding remains more forgiving on complex domes, on short runs and on part numbers where the tooling amortisation never arrives. The split we would expect is by volume and geometry rather than by year: high-volume, simple-shape vessels go dry, low-volume and awkward ones stay wet for longer.
The second trend we expect is a shift in what “qualified” means. At pilot stage, the question a buyer asks is whether a supplier can make the grade at all. Once a vessel programme is running, the question becomes whether the supplier can hold the grade across batches, because a winding programme that has been validated against one resin content cannot absorb a drifting one. Switching costs after that point are high, which is why the pilot stage is the cheap window to enter and the expensive one to skip.
The third is resin selection. PA6 costs less and processes at lower temperature, but its moisture behaviour and its interface with the liner need attention. PA12 is the common choice where cold performance and moisture tolerance matter more than raw cost. PPS enters where thermal and chemical demands rise. Our expectation is that PA6 grows fastest where cost decides the programme, and PA12 where lifecycle and cold-side behaviour decide it. That is a judgement about direction, not a forecast with a date attached.
The bottleneck that will actually bite, in our view, is not carbon fibre supply. Fibre has many sources. Consistent, well-slit narrow tape with stable resin content and controlled unwind behaviour has far fewer, and slitting capacity is the least glamorous constraint in the whole chain.
Where our product families sit
We make thermoplastic UD tape and prepreg in CF/PA6, CF/PA12, CF/PPS, CF/PAEK (PEEK), CF/PC and CF/TPU, along with the glass-fibre equivalents including GF/PA6 and GF/PC. We also produce carbon fibre fabric in plain, twill and satin weaves, carbon fibre laminates at 1.0, 1.5, 2.0 and 3.0 mm that can be machined, aramid fabrics, carbon fibre tow, chopped fibre and carbon nanotube masterbatch.
The published specification range is tow sizes from 3K to 24K, fabric areal weight from 50 to 600 g/m², and fabric width up to 1270 mm. For tape, tolerance and resin content are confirmed per grade rather than quoted as a single house number, because the useful answer depends on the resin system you are running.
To be clear about what that is and is not: qualification for pressure service is a customer-driven, part-specific process, and we do not claim it. What we supply is industrial-grade material with grade-specific TDS, SDS and COA, backed by documented in-process checks on areal weight, weave consistency and resin content — the properties that decide whether a tape runs cleanly through a winding programme. If you are building a trial for a wound vessel, send the resin system, the reinforcement, the areal weight and resin content you need, the tape width, and the process you intend to run, and we will answer against your process instead of against a catalogue page.
This article separates published facts, attributed to their source and date, from our own assessment, which is marked as such. Winding parameters, tolerances and handling conditions are confirmed per grade and are not interchangeable between resin families.