Robot and vacuum redesign for the Fastcorp DIVI vending machine

What was failing

This was before Atrick Design, while I was the product developer at a Connecticut contract manufacturer. The DIVI uses a robot arm with a vacuum tip to pick products and deliver them. The scope was to fix the robot and the vacuum pump while keeping the same controls and code.

The robot. The original arm was plastic and about 3 feet (about 0.9 m) from shoulder to vacuum tip. The machine bounced in shipping and the arm got pulled on during vends, and the plastic shoulder frame cracked around its bearings.

The vacuum pump. The original DC pump was a rotary vane pump with a plastic rotor and vanes. Most rotary vane pumps use short vanes that rub on the cylinder; these vanes were very tall, which made a lot of friction and heat, and a pump could run only about 10 vends in a row before it overheated and its thermal cutout shut it down. A bound-up pump could also burn out the switching transistor on the control board. The rotor was molded and then machined for the vane slots, and steps were left in some of the slots, likely from the plastic moving as its molding stresses relaxed. The vanes caught on the steps, bound up, then snapped free and slapped against the rotor. The result was a harsh rattle, and pumps that could no longer pick up products after about a year.

The motors. The robots in for repair had motors from many different makers, and a common failure was the winding insulation breaking down, shorting the motor and taking out the control board with it. Some motor brands were known to hold up, and one low-cost motor proved very reliable. Qualifying it took real effort, because the assumption was that a more expensive motor had to be better. The testing showed otherwise: in one batch of more expensive motors, about nine in ten failed within a month, and their gear bores were not concentric with the gear teeth.

The redesign

I redesigned the robot in sheet metal, with truss-style arms, and replaced the DC pump with an AC vacuum pump in its own vacuum box. The robot, vacuum and power were designed as modules, and the robot and vacuum box were used across three machine lines.

Duty cycle went from about 20% to 100%: the new pump ran at full vacuum for long periods without overheating. The machine got about 15 dB quieter, and the AC pump sounds like an ordinary vacuum cleaner, with none of the old pump's chatter. Assembly and break-in problems were eliminated.

I also designed the vacuum so it could be pulsed, running the pump at a lower average speed. To prove it, I used an Arduino to pulse the solid-state relay that runs the pump. In that testing on a DIVI, pulsing at about 40% made it about 35 dB quieter while still moving plenty of air, enough to talk at a normal volume beside the machine while it vended. The DIVI's control board could not pulse the pump, but a later machine built pulsing in, based on that testing.

The 250,000-vend requirement called for continuous-flex wire in the robot harness. That wire was stiff and hard to fit into the original shoulder cover, so I made the case for a new cover with the harness connectors on the side instead of the top. It made swapping a robot even easier and gave the wiring room to fit without rubbing on anything. Before the ETL testing, I stress tested the new harness past 64,000 cycles.

Tested before it shipped

The redesign passed a 250,000-vend life test and certification testing. The person I trained to assemble the robots wanted to see how long her robot would last and kept it running after the life test ended. She stopped at 350,000 vends, and it was still working fine.

I also built a robot test stand for companies integrating with the DIVI. It clamps down and carries a power supply, a vacuum box and a robot, so an integrator can do the whole integration without a full vending machine moved into their workspace. A colleague later used the same parts to build a four-station rig that ran robots through a short break-in test before they went into a machine.

Setting up each robot included aligning its two hose drive rollers, using a shaft with a vacuum hose on it, so the hose drove straight up and down instead of being forced at an angle. Then the set screw was snugged and the jam nut tightened. Building robots and machines from scratch has a steep learning curve, and I trained a lot of people in the assembly and its best practices.

The rest of the machine

The same project took on the smaller problems around the robot:

• Colder freezers. Fastcorp also sold the DIVI for ultra-cold ice cream, which has to be kept colder than the freezer's built-in thermostat allows. Separate temperature controllers let the freezers run colder.
• Circuit breakers in place of glass fuses, which customers struggled to find replacements for. More on the fuse fix →
• A vacuum hose nut that threads onto the hose, with two set screws pressing the hose into a cone seat to seal it and lock it in. It replaced a pinch clamp that occasionally caught. More on the hose nut →
• Touchscreens built into the door. Early touchscreens from a supply partner came as bulky boxes to bolt onto the machine. Instead of building an adapter and modifying the door around them, I took the screens apart and built them into a plate in the door, so they look like they belong. More on the touchscreen →
• A product display Fastcorp controlled. Bought-in displays kept being discontinued after about a year. Working with a component distributor, I built Fastcorp its own display, which can also play video.