Carnegie Mellon Textiles Lab Creates Automated Knitting Processes for 3D Meshes

You might be surprised to find out how often the worlds of knitting and 3D printing collide, with innovation and technological savvy that offers a wide range of new benefits for all involved. From a circular 3D printed knitting machine to projects translated into 3D printing to a Disney Research industrial compiler, there has been surprising innovation in knitting over the past few years. Now, Carnegie Mellon researchers are getting into the action with an automatic process for creating 3D meshes.

The goal of the team at the Carnegie Mellon Textiles Lab was to work with conventional modeling programs in creating a computational approach for making 3D meshes. To surpass current limits imposed by traditional knitting machines and the need for manual design, the researchers focused on developing technology that could produce helix-free mesh that is quad-dominant with uniform edges. The new technique also forges a knitting path with a specialized tracing procedure, along with scheduling instructions, and is outlined in a recently published paper, ‘Automatic Machine Knitting of 3D Meshes.’

The authors explain that with their new method, greater accessibility will be available to everyone interested in machine knitting—rather than just experts. They state that their knitting machines will actually be as easy to use as 3D printers, featuring:

• A guided process that results in a knitting machine graph calculating input surface.
• A tracing algorithm converting graphs for knitting.
• A scheduling algorithm that assigns knitting needles appropriately.

“For the purpose of this work, it is sufficient to think of knitting machines as devices for constructing generalized cylinders,” state the researchers in their paper.

“By adding loops of yarn to cycles in various ways, knitting machines can create generalized cylinders of different shapes. For instance, a cylinder can be widened by creating new loops, narrowed by stacking loops before knitting through them, or bent by knitting new yarn around only a portion of the circumference. The only constraint on these operations is that any new loop must be introduced adjacent to the last formed loop otherwise a long tail of yarn will be left to trail across the cylinder.”

Their method uses a manifold 3D triangle mesh. It consists of at least two boundaries, along with a monotonic knitting time. The researchers also required that such input have constant boundaries to produce mesh through remeshing, tracing, and scheduling. For this particular project, the team used a Shima Seiki SWG091N2 15-gauge V-bed knitting machine, with Tamm Petit acrylic yarn. Designs were knitted in half-gauge at 30 percent speed. Loop width and height were measured at 3.66mm and 1.73mm throughout the research.

Plush versions of the kitten, a teddy bear, a horse, the Utah teapot
and the Stanford bunny.

With this process, the team was able to make automatic knitting patterns for basic objects such as plush toys—along with allowing enthusiasts on all levels to enjoy adding their own creative touches. Fashion designers have the same latitude and can also use 3D scans for fabricating items like custom gloves.

Garments and accessories fabricated using the automated system.

“Knitting machines are as robust and repeatable as 3D printers, but – until now – they have not enjoyed the same popularity. We believe this deficiency stems from the lack of easy-to-use software and consumer-level hardware,” conclude the researchers.

“By automatically producing machine knitting instructions from 3D models, our system makes these machines as easy to control as 3D printers. We see this as a crucial step in moving industrial-style machine knitting from an industrial technique to a widely-available fabrication technology.”

During remeshing, a graph with row and column edges is constructed iteratively by: (1) finding a next cycle (purple) one row-height away from the current cycle (yellow), (2) adding nodes to the next cycle and marking nodes to keep (purple) or discard (green) based on the time function, (3) linking the nodes to the active cycle, (4) discarding the marked nodes and trimming the mesh, and (5-7) updating the active cycle and repeating. The level set of tactive is shown as a dotted white line. If the next cycle occurs after tactive, it is entirely accepted (2-3) otherwise segments that lie under the contour are accepted if longer than 2lr (5-6).

Authors included Vidya Narayanan, Lea Albaugh, Jessica Hodgins, Stelian Coros, and Jim McCann. Find out more about the Carnegie Mellon textiles lab study here.

Discuss this article and other 3D printing topics at 3DPrintBoard.com or share your thoughts below.