Students: Jesper Tjauw-A-Hing, Remt Bertz, Erwin Damman, Ka-Chun Tsang (Fall 2017)
Introduction
Hazeu Orchids asked us to develop an automated solution to package their plants in boxes. Their current workflow for packaging involves people physically lowering two trays of plants, next to each other, in a box.
They already have an automatic taping machine they use to seal the boxes. Also, pre-folding of the box can be considered out of scope, since off the shelf solutions for folding boxes are readily available.
Possible solutions
A good solution for this problem could be to develop a method to fix the plants and fold the box around them. However, this would require a radically different kind of box, and a relatively large machine for which we don’t have the resources at the moment.
The SMR project guidelines required us to use an available robot arm from our Robot Lab and some kind of computer vision involvement. Our first ideas revolved around imitating the current workflow, i.e. lowering the plants in boxes using a Universal Robot combined with a tall slender end of arm tool. This design direction proved to be difficult because of the lack of free room within the boxes when the trays of plants are inside it. It proved to be difficult to design a stabile tool that was able to retract from the box after placement of the plants.
Simply dropping the plants in the boxes could of course damage the flowers and would not be an acceptable solution.
Our solution
Our final concept involves using the free space in the bottom of a semi-folded box to accommodate a fall-retardation device which could appear and disappear underneath the trays of plants.
A retracting platform was designed to be able to create enough area to balance the trays on top. This platform lowers the plants gently in the box using a 650 mm travel pneumatic cylinder. To reliably retract the table, an inductive sensor is placed near one of the vertical guide rails to detect when the platform approaches its stowing space in the roller conveyor underneath.
The retracting platform is made to fit the hole in the bottom completely. This requires a precise placement of the semi-folded box. A set of side rails and a set of 3 pneumatically actuated clamps, coupled with PLC control and a photosensor near the front of the setup, fix the box in place.
A Universal Robot is used to pick up two individual trays from fixed locations and place them one by one on the retracting platform. This is a static program for a simple horizontal translation. Communication between PLC and UR is done via Modbus TCP. A very simple but effective gripper has been designed and built.
To meet the vision requirement a Cognex camera is used for quality control, to confirm that two trays were placed on top (before lowering them) and have successfully disappeared into the box (after lowering). At this moment, if a problem is found, the system just waits until the problem is magically solved before continuing. Supplemental error handling can be added later if needed.
To move the box onto the sliding slope, the roller conveyor is tilted pneumatically. Finally the two remaining open flaps in the bottom of the box (the sides) are closed by a folding device at the end of a slope.
A HMI layout is developed to provide both automatic mode which displays the state the machine is in, and a manual mode to actuate all the cylinders by hand. Since automatically controlled pneumatic cylinders are potential hazards, fail-safe system consisting of 2 emergency stop buttons is implemented. When either of these buttons is pushed, all pressure is released from the system.
Conclusion
The developed system is fully capable of automatically placing two trays of plants in a pre- (semi-) folded box. The required elements (robot arm and computer vision) were used in this design. One full cycle (filling a box and setting up for the next one) takes about 30 seconds.
The UR robot is, in terms of speed of the system, the bottleneck. This is largely caused by the weight of a full load of flowers (2x 5kg), which is within the UR10’s specs, however testing proved this was more than it could handle. Two separate placing motions are required and the extra round-trip takes a long time. The rest of the system has nothing to do but wait, so it’s time wasted. Consequently, a more powerful robot would speed up the entire process by more than 50%.
Furthermore, due to the static nature of the pickup and placement points for the trays, a dynamic and sophisticated tool such as a UR10 is not strictly needed to perform the function of placing the trays. This translation could be realized much quicker (and cheaper) by a simple pushing motion combined with some guide rails instead of a cool but relatively slow, €35.000 UR10.
