Paint Flow Control Project, Chapter 3.25, The controllers worked in China!

Thomas Nguyen took two of the control systems to a UEI (UEI) painting facility in China and got them to work, and work well! Thomas is my cohort at Vivid Inc. (VividInc) in Santa Clara. Thomas and I have been working on various paint flow control projects since Thomas and I and my friend Gary Johnson (GWJ) developed a paint flow controller in 1991 for K&H Finishing in San Jose.

Since I last reported on our progress, we made a few design changes in the systems, the most obvious being the enclosure for the controllers. Also I revised the controller software considerably to include a Manual mode upon startup of the controller.

We had selected an AW Gear Meters flow meter (AW Gear Meters) and their MX9000 display unit for measuring the paint flow. My first enclosure design did not accommodate the MX9000 and Thomas wanted a revision to the enclosures to include mounting for the MX9000.

So I had Protocase (Protocase) fab some new enclosures that provided mounting for the MX9000.
Protocase provides a very usable design software and I designed the enclosure below. The MX9000 fits in the cutout on the left side of the enclosure and the 4D Systems 4DCAPE-70T Touch-Screen LCD (LCD) fits in the cutout on the right side. Protocase added some welds in the corners, front to back, and fabbed a very sturdy enclosure. They ground down the welds and after painting, the welds don't show.

Below is a photo of the enclosure with the AW Gear Meters MX9000 and the 4DSystems 4DCAPE-70T in place, Thomas Nguyen taking the photo and reflected in the LCD screen. :-)

The Enclosure with an AW Gear Meters MX9000 Flow Meter Display and a 4DSystems 4DCAPE-70T LCD Touch Screen.

Noemi Fortes had installed the LCD's, Beaglebone Black (BBB) micros, power supplies etc in two of the original enclosures and then when Protocase shipped me the larger enclosures, she installed the MX9000's and the remaining LCD's, the BBB's etc in the larger enclosures. Wanting the larger enclosures with the MX9000, Thomas and I moved the parts from the first enclosures to the larger enclosures and Thomas took those two controllers to the UEI painting facility in China while Noemi and I worked on the two other new enclosures.

At Thomas' request, I modified the software quite a bit in order to provide two operating modes: Manual (open-loop control) and Automatic (closed-loop control). Upon startup, the controller is in Manual mode just to get the paint flowing through the system. Then the operator can switch to Auto mode. Below is a photo of the screen in Auto mode with the system in desktop testing. For desktop testing, I connect the output analog control signal back to the input and include a software low-pass filter of the measured signal to simulate the actual paint flow control. On the LCD screen, the green trace is the setpoint, the red trace is the output signal and the blue trace is the "measured" (simulated) paint flow.

The control software uses Fuzzy Logic as I had described in previous posts but is somewhat enhanced using a computed "System Gain" value. The computed System Gain is determined during the initial Manual mode by dividing the paint flow value by the control signal value. Then when the controller is switched to Auto mode, the System Gain value is used to set the starting control output, thereby getting a control output close to the ultimate value that would then be fine-tuned with the Fuzzy Logic algorithm. It's sort of cheating, but a very effective way to kick-start the controller to the best control value on switching to Auto mode. Hmmm, I wonder if it can be patented.

Screenshot of the control software in a test configuration.

Here is a photo of the controller in actual operation at the Vivid Inc facility. The photo is a bit fuzzy (no pun intended) but the controller is operating in Auto mode with a set-point of 800 mL/min and flow reading of 793 mL/min, The MX9000 is reading a flow of 864 mL/min, about 9% higher than the reading on the controller's LCD, probably a calibration error.

Paint Flow Controller in operation at the Vivid Inc facility, Santa Clara, CA.

The controllers work very well in the UEI painting facility but of course there were a couple of problems. The main problem was with the flow meters getting gummed up with chunks of dried or gummy paint. In initial testing at the Vivid Santa Clara facility, we had problems with the paint getting gummy and clumps of the paint drying in the tubing between testing. Thomas frequently had to open up the AW flow meter and clean it out. The AW flow meters are used in automobile painting in U.S. manufacturing plants, so should be good for our use. I figured that the paint in the China facility would be pure and of consistent viscosity, so Thomas wouldn't have any problems getting the controllers operating perfectly but it turned out that because the tubing in the China facility had many right-angle turns that the paint flow would be slowed and the paint would become coagulated, resulting in a clogged-up flow meter. To remedy the problem at the China facility, Thomas insisted that the tubing be re-plumbed, making smooth turns to keep the paint flowing smoothly so as to eliminate the clogging and gumming-up problems. We will see how well that will be accomplished.

Happily, the controllers work well. The Fuzzy Logic software with my System Gain kick-starting feature seems to be ideal. The Chinese operators are able to work with the LCD touch-screen and do necessary adjustments to the MX9000 flow meter display units for their particular paint flow values. And the Beaglebone Black micro with Python plus tkinter and Matplotlib software seems to be perfect for the control system software.

We are expecting that UEI and Vivid Inc. will be ordering more of the controllers in the next year, eleven or twelve, since they are finding that the controllers work well and are expected to greatly reduce defects and be a real money-saver,