Some materials are easier for robotic handling systems to deal with than others. Metal and plastic parts have well understood physical characteristics, but relatively little work has been done to determine the idiosyncrasies of the many fabrics that are used to make clothing. For this reason, the design of a robotic system to manufacture clothing represents a difficult challenge and requires a good deal of innovative thinking.
Professor Paul Taylor has loads of experience in innovative thinking. He’s been involved in the design of robotic handling systems for the garment industry for many years, first at Hull University and now in the Department of Mechanical, Materials and Manufacturing Engineering at the University of Newcastle-upon-Tyne.
One of his projects was to develop a line for the robotic manufacture of men’s and ladies’ briefs. The fabric used is a knitted cotton based material, supplied in single and double knit. So before embarking on the development of the system, Professor Taylor and his research team had to think carefully about the properties of the materials, such as static cling, friction, flexibility and deformability.
From the start a modular system was decided on that would take elements of clothing along an automated manufacturing line. The first hurdle was to separate a single component part of the underwear from a stack of similar material and place it onto the automated track, in a way similar to the manual assembly process.
To perform the separation of one piece of material from another, Dr Douglas Kemp, an early team member, developed a destacking technique. The robotic destacker comprises a number of different sensors, a gripper and an air delivery system. A tactile sensor on the unit determines the top and the edge of the stack. The destacker then moves to a predetermined position adjacent to the material stack and an air jet is turned on. Compressor air blows down through the material, causing the top fabric to vibrate and lift from the rest of the stack, and a knife finger is inserted. A photoelectric cell detects its presence. Once identified, grippers hold the material and a robot arm transfers it to the conveyor belt.
The system employs error detection and correction. A high signal indicates that no material is present; a normal signal shows a piece of material is there. No signal indicates more than one piece of material which the system rejects. By using an infrared detector, the system can also determine the amount of movement of the top piece of material, and the air jet can be adjusted to accommodate different material types.
Once on the conveyor belt the fabric is carried to the next station, where its orientation is checked – even with a reliable stack placement and destacking technique, single piles of material are still going to change their position and orientation on the conveyor belt.
‘You either need to correct the problem, or to put cameras on the robots to cope with it’, says Taylor. ‘With black backgrounds and white underwear, cameras are a possible solution, but with floral print material and a dirty industrial background, the camera solution can be more of a problem.’
So alternative methods were examined. In this case, a linear flat vibrating surface was used to impel the straight edge of the fabric towards a fixed barrier. The vibrating plate is tilted so that the fabric moves under gravity parallel to the barrier once contact has been made. An optical sensor halts the process once the fabric reaches the desired position. The technique gives an accuracy of +/- 1mm.
Once correctly oriented, the three pieces (front, gusset and another front piece) are picked up by a robot gripper and transported to a sewing machine where they are sewn together.
Linear conveyor feeders with velocity and position feedback are used for most of the sewing functions. Jigs attached to these feeders are loaded at one end by a jig loading robot. The first sewing operation involves three component parts: the sewing of two gusset pieces with a backpiece sandwiched in between. These pieces are separated by narrow barriers to allow for the inversion of the seam. Once this operation is complete, the two gusset pieces are folded around the seam to conceal the stitching.
After a similar operation to add the front panel of the briefs, the sub-assembly is transferred to another jig by another gripper robot. Here, leg tape or elastic is placed around the semi-circular arches that form the leg openings. While sawing around a semi-circular structure is easy for a human operator, it produces some challenges to automate. ‘We had the idea that since we had a material that could bend, we would straighten it,’ says Professor Taylor. The idea of straightening the garment was developed so that the two semi-circular leg openings could be presented to a pair of sewing machines in a linear fashion, where the two bindings could be placed around the legs.
To ensure accuracy, retro-reflective, optical sensors were used to locate the edge of the underwear to within 1.5mm prior to sewing.
The final assembly stage entailed finding a way to put sideseams and a waistband around the underwear. It was achieved with the help of a foldable two-part gripper. Between them, they offer one edge of the garment up to an overlocking machine that allows it to join the first side seam. The gripper opens the garment out to put the waistband on. Then it closes and the other sideseam is sewn on before the completed garment is removed.