Paint Flow Control Project, Python and Fuzzy Logic Simulation, Chapter 3.14

First: our project is still alive but I don't have funding yet to build the new control system. My cohort co-developer at Vivid Inc, Thomas Nguyen, has traveled to Ontario to check out the paint-gun control robot controller and other equipment that UEI and Vivid hope to acquire for the system at Vivid. This robot was used to control the pointing of the paint gun. The Ontario equipment checked out Ok except for missing the pulse to current converter/transmitter for Micromotion Coriolis flowmeter. The equipment will be shipped up to Vivid in Santa Clara. Also Thomas was recruited to visit a painting facility in China, "out in the boondocks" west of Shanghai. They are painting some keypads for the UEI remote control devices. Thomas suggested changes to significantly speed up the painting process. He's back from China and filled me in on the latest news. I did a search for replacement for the flowmeter pulse to current converter/transmitter but mainly ...

Currently, Getting Started With Python: Awaiting the funding for the Paint Flow Control Project, I've been working on a simulation program for the Fuzzy Logic control algorithm. I had been using an evaluation copy of National Instruments LabWindows/CVI program but the extension ran out and, since our project is still unfunded I thought I would develop a similar program using Python Dive Into Python) and there's plenty of videos and other Python info online, so I dove right in.

(https://www.python.org/). I had used Python for a simple Ethernet communication task while at LLNL and found it easy to develop in and liked its simple and clean syntax, plus it has some interesting features such as the Lists and Tuples. However, I hadn't become really competent in developing a Python program, so figured the Fuzzy Logic simulation work would be a good way to get more Python savvy. Python includes Object Oriented (OO) capability and I had wanted to become more familiar with OO. I have a copy of Mark Pilgrim's book "Dive Into Python" (Dive In) and so I decided to Dive In.

For my simulation code, I would need a GUI that would allow me to enter a Setpoint for the paint flow and other values such as the system response and others, plus show the computed flow values and display a graph of the computed flow. Python includes a GUI design capability with Tkinter (Tkinter) that I figured would be adequate for the labels, buttons etc that I would need. And I was aware of the graphing program matplotlib (matplotlib) that is available for use with Python, so decided to try to incorporate a graph into my GUI using matplotlib.

I had signed up with Lynda (Lynda) in order to learn Python, Java and other stuff, so began watching their Python courses. Fortunately their Python course includes an introduction to Object Oriented coding for Python, plus they have a Tkinter course to get me started. I found the Lynda courses to be very useful getting started with Python, giving me some confidence that I could develop my Fuzzy Logic simulation code in Python. After reviewing the info in "Dive Into Python" I then viewed the Lynda Python course on Python 3 (Python-3) and followed it up with their Tkinter course (Tkinter). I had decided to use Python 3, figuring that it was the Latest and Greatest. However, I found out later that Python 2 would probably be adequate for my tasks.

I downloaded version 3.4.3 from the Python website and while viewing the Lynda courses, I began developing my Python code from my old LabWindows/CVI code, including the "Widgets" for the labels and entry values. I was able to quickly get my code running, doing the Fuzzy Logic simulation and was ready to include a graph of the simulation results. Including a matplotlib graph was a big stumbling block even though the matplotlib website has a lot of examples for graphs and chart.

matplotlib installation
But I hadn't even installed matplotlib on my PC. I tried installing matplotlib but since I was/am such a Newbie, I ran into difficulties. Checking the matplotlib website, they suggest using the free downloads from either Canopy (Canopy) or Anaconda (Anaconda) to install all the necessary programs to run matplotlib.So I first installed Canopy but after using it for a couple of days, I realized that Canopy uses Python version 2.7.X and since I had planned to use Python 3, I then decided to install Anaconda which uses version 3. Most likely, version 2.7.X would be perfectly adequate for my tasks but being ignorant of the differences between vers 2.7.X and 3.4.3, I decided to drop Canopy and install Anaconda which uses version 3.4.3. Then I began trying to use Anaconda but ran into some coding difficulties and assuming that it was an Anaconda problem, I decided to stop using Anaconda and find some other way to install matplotlib. I don't know what caused my problems with Anaconda, but I simply assumed that it "had to be an Anaconda problem" and couldn't possibly be my coding inadequacies :-). Anyhow,  I searched around and found an "edu" website that a matplotlib installation (Worchester Polytechnic Institute) and decided to give it a try. Although either the Canopy or Anaconda installation would probably have been adequate, the Worchester installation has been a godsend and I was finally able to make headway in my code development.

matplotlib Code Development
Although there are plenty of examples on the matplotlib website, those examples used pyplot and I didn't want to use pyplot as I believe it produced charts on separate windows, separate from the main window of the GUI. had already started my code development following Youtube videos by Sentdex (Sentdex). Sentdex's first videos used pyplot but later ones seemed more suited to what I wanted to do and I began massaging my code to be closer to his examples (Sentdex matplotlib example). His subsequent example was even closer to my needs, but he is using multiple pages and displaying the graph on a secondary page, whereas I wanted to have my graph in a frame on the main page of my GUI (). However, Sentdex provides some very useful information including the use of OO coding in his examples. Sentdex's video on a Live graph was very useful to me, showing me the way to "animate" my graph (Sentdex Animation Graph).

Sentdex's Graphing Video at https://www.youtube.com/watch?v=JQ7QP5rPvjU

However, he is still using multiple pages for his GUI, and also it seems to me that his coding is unnecessarily complex, but that's just my humble opinion. Anyhow, I was able to simplify Sendex's code and put a Live graph in a frame on my main GUI page as shown below. But I am grateful for Sendex's videos, finding them very useful.

My Python Fuzzy Logic Simulation
I have to admit that I was overwhelmed with all the info from Sentdex's videos and also I made several goofs in my code, particularly trying to use a variable before I had defined it. And I was confused about how to make a Global variable in Python. I wanted to make some Globals, although globals are frowned upon in OO programming, but I couldn't think of another way to initialize some variables. In Sentdex's Animation video, mentioned above, he is graphing data that he reads from a disk file. For my code, I wanted to save the simulation data somehow in memory rather than writing and reading disk files and finally hit on the idea of global lists for the x and y simulation data. That seems to work very well and saves on wear and tear on my hard disk.

Anyhow, I don't want to include the whole simulation code but here's the first few lines showing the import of matplotlib, tkinter, etc, the globals for the xList and yList values and the animate graph code. The code turned out to be fairly simple and works quite well for my simulations.

Beginning of My Python Simulation code, showing the animate graphing function.

Below is a window from the code output, showing simulation results for the Flow Response of 50% (explained in previous posts), Control Multiplier of 0.5, E Dot Gain of 0.5 and various setpoint values from 0.5, 1.0 and 1.4.

GUI with graph from my Python Simulation code.

I updated the graph to include the setpoint, green trace, the control output signal, red trace, and the simulated flow values.This simulation run is for a slower system response of 25% (see explanation in earlier posts). I stepped the setpoint values to 0.5, 1.0 and 1.4. So ideally the simulated flow (blue trace) should equal the setpoint (green trace) if the Fuzzy Logic algorithm is optimum. Note that especially in the early part of the graph that the simulated flow settles with a large error but it seems to get better after the first couple of steps. The chart is a bit confusing, showing the three traces, but is very useful in understanding what is really going on in the Fuzzy Logic simulation.

GUI with graph showing the setpoint (green), output control signal (red) and simulated flow (blue).

So my Python program will allow me to simulate different Fuzzy Logic Look-Up-Tables (LUTs) for different system response settings, and hopefully will help to develop the best LUT table for the Paint Flow Control System. This is my first try at a complete Python program with a GUI and perhaps I will be able to develop a supervisory program in Python to communicate with the Digi International BL2600 Single-Board-Computer that I plan to use to do the actual paint flow control.