The Silkscreen Project and more Fuzzy Logic

This is me with a painting by my wife Heide in the background.

Thomas Nguyen of Vivid Inc. (http://www.vividinc.com/) contacted me in January 2012 about a new control system project for a silkscreen system. I had worked with Thomas on paint flow control projects since 1991: at K&H Finishing in San Jose, at Enclosures Engineering in Fremont, CA and with Control Coat LLC. Originally I had used Fuzzy Logic on the system for K&H Finishing but we ended up with a PID algorithm to control the paint flow. At Enclosures Engineering, owner George Halldin insisted on a Fuzzy Logic controller and had in mind to sell the controller to painting companies around the world. K&H Finishing and Enclosures Engineering were in the business of painting computer enclosures, TV remote controllers and the like, with very expensive copper infused conductive paint for RF shielding to minimize RFI problems with the devices. Since the conductive paint is quite expensive, the paint flow controllers were needed to accurately mete out the required amount of paint and nothing more. I had made my first cut at a Fuzzy Logic control algorithm for the paint flow controllers and the silkscreen system appeared to be another excellent opportunity to employ Fuzzy Logic as the system dynamics of the silkscreen process were completely unknown, similar to our initial work with the paint flow system for K&H Finishing back in 1991. The SilkScreen Project was similar in that conductive ink was to be used to screen onto plastic panels that would be used for game or toy and computer keyboards and the conductive ink is expensive, as is the conductive ink for RF shielding purposes. I was happy to have another project that I could use Fuzzy Logic and hopefully improve my control algorithm.

Franklin (or Frank) Eventoff of Sensitronics LLC in Washington state (http://www.sensitronics.com/) was the client who wanted to build systems to control the ink flow to his silkscreen systems. He planned to be able to silkscreen up to 1 million panels with the conductive ink for keyboards.

Frank's initial plan was to use four separate pressure pots like shown below, each with a solenoid valve to allow the tooth-paste like ink to be metered to the silkscreening mechanism.

Pressure Pot.

In each pressure pot would be a small cup holding some of the conductive ink. Air pressure introduced through a coupling would force the ink out an exit tube to the silkscreening system. After silkscreening a plastic panel and allowing an hour for the ink to dry, the resistance of the panel would be measured using an instrument from Mistral (http://www.engsolinc.com/stfeatures.html) and the ink conductivity would be adjusted to meet the panel resistance requirements.

Frank had suggesting using Opto22 equipment for the valve control, but I figured that I would have to connect a control system to the Opto22 devices and decided that I could probably use a microcomputer with analog measurement capability and relay driver outputs thereby eliminating the Opto22 devices. Consequently I opted to select a microcomputer module to do the whole silkscreen system control. Because of my previous work with Z-World microcomputers and the Dynamic C software, I looked for Z-World modules and found that Z-World had been bought out by Digi Inc. So I contacted Digi and began selecting a microcomputer for controlling the solenoid valves, a Digi BL2600, a Relay Driver board and a development system with Digi's Dynamic C compiler. The Digi BL2600 includes an Ethernet interface besides relay driver and analog circuits. Digi Inc had purchased Z-World a few years ago and sells the old Z-World modules and the Dynamic C software in addition to developing new microcomputers and other modules: (http://www.digi.com/products/wireless-wired-embedded-solutions/single-board-computers/bl2600). Thomas' employer, Vivid Inc., purchased the Digi equipment for the project and I retrieved it from Thomas and promptly got busy working on the software for the Digi BL2600 microcomputer using National Instruments LabWindows/CVI (http://www.ni.com/lwcvi/).

Near the end of March 2012, Frank contacted Thomas and said he wanted to speed up the system development. However, I was doubtful that I could hurry-up the development since we hadn't ordered/received any of the microcomputer components or the solenoid valves and Frank's design didn't seem to be nailed down yet. Also I was completely ignorant about the panel resistance measurements with the Mistral measurement system. Consequently I had not yet started on the Fuzzy Logic algorithm and software development/design. The following week Frank decided to change from solenoid valves to motor driven valves and then Thomas convinced Frank to revise the design of the ink delivery system from the pressure pot and valve scheme to a quad pump system as the valves would be a power hog, overtaxing a controller. Also I suspect that the ink cups in the pressure pots would occasionally run dry and it would be cumbersome and time consuming to open the pressure pots to replenish the ink cups. Then Thomas convinced Frank that we wouldn't need the recycling pump. Consequently, the new design was to employ three pumps to pump inks of differing conductivity. I made a simple block diagram of the system using only three pumps and at this time with the Mistral panel resistance measuring instrument:

Simple Block Diagram Including the Mistral Resistance Measuring Instrument.

A panel would be silkscreened and then dried, taking about an hour per silkscreened panel. Then only after drying for about an hour, the resistance of the panel could be tested with the Mistral Panel Resistance Measuring Instrument and adjustments made to the ink pump speeds to correct for under-resistance or over-resistance of the panel.

I had used LabWindows (LW/CVI) on projects at the Lawrence Livermore Lab, the Smart Sampler system that I previously posted and also an "unmentionable" project. Too bad that I can't post about the unmentionable project as it was quite involved, including an experimental device, a Field Programmable Gate Array (FPGA), a microcomputer and a PC. I programmed the FPGA in VHDL, the microcomputer in QNX and the PC in LW/CVI. But as I say, the project is unmentionable. Anyhow, I liked LW/CVI very much and found it to be easy to develop in and a very capable development system. However, I didn't have a licensed copy of LW/CVI, but had an evaluation copy and I planned to purchase a LW/CVI license if and when the SilkScreen Project was funded. Also I began working in earnest on the Fuzzy Logic algorithm for controlling the ink conductivity by adjusting the pump motor speeds.

However, early in April 2012 Frank wanted to hold off on ordering the pump motors and consequently the system development and the project seemed to drag on for months, not going anywhere. I figured the project was dead but then in December 2012 Frank wanted to have a meeting with Thomas. Frank was ready to get going on the project and wanted to purchase some motors. Because of the tooth-paste consistency of the ink, Thomas insisted on a high power motor. I had never used any Bodine motors but knew of them from my work at LLNL that they were reliable and they also provided very capable controllers for their motors. So we ended up selecting some high torque 55 RPM 1/5 HP Bodine 34WB Brushless DC Parallel Shaft Gearmotors model 1087 for Frank to order along with Bodine 3911 Controllers that allowed us to vary the motor speed from 0 to 55 (http://www.bodine-electric.com/Feeds/Products/ProductModel.asp?Context=16&Name=34B-WX%20Series%20Parallel%20Shaft%20BLDC%20Gearmotor&Model=1087&Sort=11689).

RPM while providing enough power to squeeze the tooth-paste-like ink to the silkscreen machine.

The Bodine 1087 High-Torque Gearmotor.

At the end of April 2013, Thomas had a single Bodine 1987 motor and an ink pump running to test with the tooth-paste-like ink. Thomas had convinced Frank that the recycling ink pump was unnecessary and so the system was settling on three ink pumps that would be adjusted to get the desired ink conductivity and resistance of the silkscreened panel. Then Frank ordered two more of the Bodine motors and controllers. The motors and controllers were sent to Thomas. Thomas assembled the motors on a heavy aluminum backplane and the controllers in a large fiberglass enclosures and got them to me for my software development work.

A schematic of the microcomputer system is shown below. Note that it doesn't include the Mistral Instrument as I was unsure how to tie it in and even unsure if we would need the Mistral Instrument.

Schematic of Silkscreen Controller, June 2013

I was still unsure about the Mistral Instrument and late in August 2013 Frank sent us a 2" x 3" sample of a silkscreened panel and the printed circuit board contacts that they planned to use with the Mistral. So I rigged up a crude hookup to the 2" x 3" sample using spring-type paper clips and devised a method to measure the resistance with the BL2600 microcomputer. My simple circuit used the voltage output from the BL2600's Digital-To-Analog Converter and a fixed series resistor, then measured the voltage across the 2" x 3" sample. With some simple algebra and knowing the DAC's output voltage and the value of the fixed series resistor, I could compute the resistance across the 2" x 3" sample, finding that the value agreed well with Frank's measurement of the sample. So it turned out that we would not have to use the Mistral Panel Resistance Measuring Instrument, saving some money and actually simplifying the system with my resistance measuring scheme. I put the code for this measurement in my developed LabWindows/CVI code that I developed using National Instruments  (http://www.ni.com/lwcvi/).

Meanwhile I was continuing to work on the Fuzzy Logic algorithm for the SilkScreen Project and will report on the algorithm in a future post. I was developing my LW/CVI GUI and code and simulating the operation of the system. As I said, Frank wanted to use four pumps and motors to mete out the ink to the SS system. The conductivity of the three inks would be as follows: one with low conductivity, one with the expected required conductivity and one with a higher conductivity. In my control algorithm, I decided to fix the motor speed of pump for the required conductivity and allow the user to be able to set that motor speed on the GUI. My LW/CVI control program would adjust the speed of the other two motors using my Fuzzy Logic algorithm. My LW/CVI code included the capability to simulate the system's operation and the operation of the Fuzzy Logic algorithm. A screenshot of the GUI is shown below:

Screenshot of Silkscreen Program GUI, Showing Simulation of SS System Control.

The software and GUI are set up for developmental and experimental purposes and consequently there are many selections and factors to select and adjust. As a result, there are several things in the GUI screenshot to explain:

1. The Connect button enables the Ethernet connection from the BL2600 microcomputer to the PC,
2. After the Ethernet connection is made, the green Connected "light" is illuminated,
3. The Max values of the motor speeds in % can be entered along with the initial motor speeds, and the fixed pump speed for Ink #2 is set,
4. Frank wanted the motors to be switched on and off. I set the off time at 1 sec and the Power On Time can be set with the slider,
5. Below the Connected light is a a text box for the User to enter the Panel Resistance Setpoint,
6. Some factors can be set on the right side of the GUI,
7. The Ink Pumps can be started with the Start Ink Pumps button at the upper right-of-center, 
8. Data acquisition can be enabled with the Data Acq On/Off,
9. For simulation, the Simulation On/Off button can be pressed,
10. After the Setpoint has been entered, Automatic Control can be enabled with the Enable Control button,
11. Simulation was enabled and the graph shows the individual pump motors being cycled on and off and the Fuzzy Logic control algorithm in action to control the pump speeds when the Setpoint is changed from a high resistance to a low resistance.

So finally the system was ready to test with the silkscreening apparatus in September 2013, an experimental test. However, so much time had passed since the project beginnings in January of 2012 and Frank's company had found that a manual system to control the ink conductivity would work just fine. So Frank canceled the project. Too bad, because Thomas and I had put so much time and effort into the project but it was another "all for naught" project.

So why was the project canceled? I made a Gantt chart to show the tasks and the timing, below:

Silkscreen Project Approximate Timing.

The Gantt chart shows that my software development was a major factor although there was a nine month delay before Frank gave us the go-ahead to purchase the motors and pumps and to begin assembling the system. Another time consuming project was the Panel Resistance Measurement method that actually took only a week after I received the 2" x 3" sample and the probes to develop the measurement scheme. However, cancellation of the project probably came down to the fact that an automated system for meting out the ink was really not needed as Frank's team was able to produce a manually operated control system where the operators could adjust the ink conductivity after panels were silkscreened and the panel resistance measured. So anyhow, the project was dead but at least it was an interesting learning experience, business-wise and engineering/software-wise, giving me an opportunity to increase my knowledge of Fuzzy Logic control and improve my FL algorithm, that I'll report on in a future post.