Following the Thread: Drexel’s TopoKnit System Lays the Groundwork for Universal Design of Functional Fabrics

Before 3D printing, CNC routing, laser cutting and the tools of ubiquitous “making,” there was yarn and needle. For centuries, the earliest makers knitted things into being. Blankets, sweaters, gloves – all took shape by combining just a handful of basic stitches. Now a team of researchers at Drexel University is translating those loops and twists into a digital architecture of knitting — a key step in the process of incorporating new technologies into textiles.

 

While the promise of textile-embedded technology, or “functional fabrics,” has been on the horizon for decades, it has primarily been realized in the form of high-performance and technical military gear and high-end fashion concepts. In most of these garments, technology is an external addition, rather than an integrated feature, of the design.

 

One of the greatest barriers to full integration of technology into textiles and broader retail adoption of functional fabrics, according to David Breen, PhD, a professor in Drexel’s College of Computing & Informatics who has been computer modeling fabric since the 1990s, is that current software used for industrial design and production of textiles lacks the thread-level detail necessary for digital sampling and precision manufacturing of fabric devices.

 

“In order for these [technical] textiles to be widely deployed and reach their full industrial potential, computer-based modeling and simulation tools must be developed to support the design and optimization of knitted structures,” Breen’s group wrote in a recent paper in the journal Graphical Models.

 

The piece introduced TopoKnit, a suite of algorithms they developed as a tool for modeling the path of a yarn within a knitted textile. While it doesn’t solve the entire modeling challenge, the program does provide an essential element of the design process: documentation of how parts come together to make a finished piece — the equivalent of a blueprint in architecture.

TopoKnit translates stitch commands — like knit, purl and transfer — as they would appear in a knitting pattern, or the program of a digital knitting machine, into a map that shows where the yarn travels, loop by loop, and how it interacts with adjacent loops as the textile is formed. The resulting diagram, called a topology graph, allows designers to pinpoint where a piece of yarn is, with respect to the overall plane of the textile, at any given point within it.

 

Building up this baseline design information for knitting is coming at a time when more researchers are considering knits for making functional fabrics. Breen suggests this is partly because knitting supports more intricate yarn interactions than weaving, which is advantageous for creating electric circuits. In addition, knitting allows for more controllable design touch points, as well as the ability to generate 3D shapes without added manufacturing steps, such as cut and sew. 

 

“What’s interesting about knitting is that, at the stitch level, it has a completely programmable microstructure. Knitting is a kind of programming that maps stitch operations to specific physical structures.” Breen said. “Because you have the loops forming many different connections, knitting is more complex than weaving. Knitted fabrics have always been more difficult to model than woven fabrics. But it gives designers more entry points to manipulate various aspects of the material, which makes it very promising for building in new functionality.”

 

To put TopoKnit to the test, Breen’s doctoral student Levi Kapllani Maharaj worked with research partners and designers in Drexel’s Center for Functional Fabrics to generate a series of 100 patterns of 5x5 stitch configurations using the graphics interface on one of the Center’s digital knitting machines. The same stitch commands that went into the machine were also entered into TopoKnit to produce a topology graph. The team compared each graph to its corresponding graphic rendering to see if the stitch map matched the rendered model.

 

The graphs were an exact match in each case, showing that the TopoKnit system could be used to reliably produce manufacturing instructions for knit textiles via a sequence of steps that produces specific yarn topologies.  These topologies, descriptions of how the yarns contact and interconnect with each other, allow the system to flag stitch patterns in a design that would not be viable in production, which is an important step for prototyping.

 

 

While the accuracy of TopoKnit is important, Breen noted that it is still just the first of many steps toward a textile modeling program that can represent and simulate functional fabrics.

 

For any kind of high-level computer modeling, a topology graph is the foundation on which more noticeable characteristics, like shape, strength and movement, are built. But for textiles, that foundation was never established, because the process of churning out products was usually more urgent than redesigning them for high-level performance tools.

 

“It is ironic that fabric is one of the oldest human-created materials, but modeling it has proven to be extremely challenging and computationally expensive.” Breen said. “Steel beams are easier to study and model than knitted fabrics: because fabrics do things like stretch and twist, they require vast amounts of computational power when modeling them in the ways similar to that steel beam.”

 

Despite this engineering disconnect, research and development around functional fabrics has gained momentum mainly due to government projects focused on specific performance goals, like embedding communications or vital sign-monitoring technology into military uniforms. But to do this, the field has relied heavily on the expertise of individual designers and fabricators with deep experience in knitting.

 


The benefit of enhanced textile design platforms, Breen suggested, is that it would open up textile design to people with expertise in other areas, like electrical engineering or materials science. Topoknit is a technology-agnostic approach to developing tools for interoperability across machines and modeling platforms. And because it would allow designers to try new approaches with a better gauge on what will work – rather than spending time and resources on trial and error – better modeling should enable efforts more closely linked to consumer goods. 

 

“Some of these materials are really expensive. You can’t afford to do trial and error testing because of the limited supply of some of these advanced fibers and yarns,” he said. “The vision for computer-aided design is that you model, simulate and explore the design space. You do it all virtually, computationally, so you don’t have to go through the expensive process of making it and then seeing if it works. Or you could at least explore the design options and narrow down the tests you want to do with the finished piece.”

 

Building on the topology framework provided by TopoKnit, the next step for this research is to optimize the shape and behaviors of knitted textiles. Ensuring accuracy and modeling the shape will prime computer programs to meticulously reproduce the mechanical properties of textiles and ultimately direct knitting machines to produce textiles with specific performance capabilities.

 

“If we’re going to see the potential of functional fabrics fully realized, we need to get to the point where knitting is even easier than 3D printing – where you can put in all the desired parameters from size and shape to flexibility and thermal or electrical conductivity and press ‘go’ and a knitting machine will produce it. This work is putting us on the path toward that reality.”


 

In addition to Breen, and Kapllani Maharaj; Genevieve Dion, director of Drexel’s Center for Functional Fabrics; Chelsea Amanatides, PhD, a senior research engineer in the Center; and Vadim Shapiro, PhD, of the University of Wisconsin have contributed to the development of TopoKnit. The project was initially funded by the National Science Foundation.

 

Read the full paper here: https://www.sciencedirect.com/science/article/abs/pii/S1524070321000199