OT: My first python package - pytunelogix

[destination unknown]

I posted this elsewhere, but this is my first attempt at a python package:
A PID tuning suite for Logix
https://github.com/Destination2Unknown/pytunelogix

Anyone with Python experience feel free to recommend any improvements...
Contributions are welcome

Notes:

First foray into the world of Python

Python to PLC communication is based on pylogix (which is excellent)

This is still very much a Work in Progress

Also works with Logix Emulator (V32+)

It's based on a First Order Plus Dead Time system model and uses general tuning rules, which may not be suitable for all systems

The Method:

Step 1 -> Record a step response using the logger (or save logix trend as csv)
Step 2 -> Open the step response CSV in the tuner
Step 3 -> Test the tune using the Python Simulator
Step 4 -> Test and fine tune the ControlLogix PID with the process simulator

Logger:




Tuner:




Simulator:




Logix Simulator:

 

[g.mccormick]

This is awesome! Thank you for the contribution!

Question: Does the PID gains calculation change as a function of the pid form (dependent or independent) that is selected? If so, would it be beneficial to offer a check box and calculate the gains based on that?

 

[destination unknown]

This is awesome! Thank you for the contribution!

Question: Does the PID gains calculation change as a function of the pid form (dependent or independent) that is selected? If so, would it be beneficial to offer a check box and calculate the gains based on that?

The independent and dependent PID are equatable:

Independent -> Kp*E + Ki*Σe + Kd*Δe
Dependent -> Kc(1 + 1/Ti*Σe + Td*Δe)

To convert:

Kc=Kp
Ti=Kp/Ki
Td=Kp*Kd

and conversely

Kp=Kc
Ki=Kc/Ti
Kd=Kc*Kd

The tune is calculated for independent gains and optionally these are converted to dependent form.
The form of the PID doesn't impact the tune, same same but different

 

[drbitboy]

@destination unknown: what environment to you use for Python? I usually run WSL/PuTTY/Xwindows, which is solid enough; it's a bit of a dogs's breakfast to setup the first time, but still preferable for me as I am more comfortable in the Linux command-line world.

I know there are native Win10 Python environments, and I assume they can handle TK and plotting; can you recommend one?

 

[destination unknown]

I use VSCode, it's a good all rounder

 

[drbitboy]

I use VSCode, its a good all rounder

And which Python installer (Python.org, Anaconda)?

 

[destination unknown]

3.10 from Pyhton.org

 

[Peter Nachtwey]

@destination unknown: what environment to you use for Python? I usually run WSL/PuTTY/Xwindows, which is solid enough; it's a bit of a dogs's breakfast to setup the first time, but still preferable for me as I am more comfortable in the Linux command-line world.

I know there are native Win10 Python environments, and I assume they can handle TK and plotting; can you recommend one?

vscode will run on Linux and Mac. I have it installed on all my computers. It even runs on my ARM based Raspberry PI but it is a little slow. A lot of software people at work use vscode.
vscode will work well enough for most people and applications. vscode does take some setting up. The configuration files are .json files so they are easy to modify.

I used pyinstaller from
https://pypi.org/project/pyinstaller/

I have a python app that uses Tkinter and Matplotlib that I have used pyinstaller on. pyinstaller also requires some fussing with to get working optimally.

 

[destination unknown]

pyinstaller is excellent and as stated works well with Tkinter and Matplotlib but getting Windows to run python files doesn't take that much effort anyway

 

[drbitboy]

vscode will run on Linux and Mac....

And it even has a VIM plug-in, so I don't have to edit with the vile, ridiculous, injurious Microsoft TextBox (or equivalent in Linux and Mac)!

 

[Peter Nachtwey]

Off topic but somewhat related.

I use Jupyter Lab for learning and trying out new things. Jupyter lab makes it easy to follow along with examples on the internet. If you use python I highly recommend installing jupyter lab. BTW, forum member rwatson was the one that goaded me in to installing jupyter lab and jupyter hub.

At work we have Jupyter Hub installed. Jupyter Hub is like Jupyter Lab but unlike Jupyter Lab, Jupyter Hub runs on a remove server. In our case the remote server is at Delta and we can VPN in to access. The cool thing about Jupyter Hub is that I can share my work with others at the business and even if I am using a Raspberry PI, it still runs at the speed of our server.

I still use IDLE a lot. Mostly because I can have many windows open at a time for cutting and pasting between windows.

For serious work/debugging VSCode is good but many of the python gurus on YouTube use pycharm or sublime. VSCode has many pluggins. drbitboy mentioned VIM but pylint is good too. And then there is Kite that will try to finish the line for you.
VSCode works with many languages.

 

[destination unknown]

The independent and dependent PID are equatable:

Independent -> Kp*E + Ki*Σe + Kd*Δe
Dependent -> Kc(1 + 1/Ti*Σe + Td*Δe)

To convert:

Kc=Kp
Ti=Kp/Ki
Td=[COLOR="red"]Kp*Kd[/COLOR]

and conversely

Kp=Kc
Ki=Kc/Ti
Kd=Kc*[COLOR="Red"]Kd[/COLOR]

The tune is calculated for independent gains and optionally these are converted to dependent form.
The form of the PID doesn't impact the tune, same same but different

TYPO
The correct conversions:

Kc=Kp
Ti=Kp/Ki
Td=Kd/Kp

and 

Kp=Kc
Ki=Kc/Ti
Kd=Kc*Td