add docs on ziegler-nicols tuning

This commit is contained in:
Andrew de Quincey 2021-04-30 22:53:43 +01:00
parent a7fafeed65
commit 25794fa012
3 changed files with 84 additions and 1 deletions

BIN
docs/kiln-tuner-example.png Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 32 KiB

83
docs/ziegler_tuning.md Normal file
View File

@ -0,0 +1,83 @@
# PID Tuning Using Ziegler-Nicols
This uses the Ziegler Nicols method to estimate values for the Kp/Ki/Kd PID control values.
The method implemented here is taken from ["ZieglerNichols Tuning Method"](https://www.ias.ac.in/article/fulltext/reso/025/10/1385-1397) by Vishakha Vijay Patel
One issue with Ziegler Nicols is that is a **heuristic**: it generally works quite well, but it might not be the optimal values. Further fiddling may be necessary.
## Process Overview
1. First of all, you will record a temperature profile for your kiln.
2. Next, we use those figures to estimate the parameters.
## Step 1: Record Temperature Profie
This must be done without any interference from the real PID control loop. To do so, run:
```
python kiln-tuner.py ziegler.csv
```
The above will drive your kiln to 400 and record the temperature profile to the file `zn.csv`. The file will look something like this:
```
time,temperature
4.025461912,45.5407078
6.035358906,45.5407078
8.045399904,45.5407078
10.05544925,45.59087846
...
```
## Step 2: Compute the PID parameters
Once you have your zn.csv profile, run the following:
```
python zieglernicols.py zn.csv
```
The values will be output to stdout, for example:
```
Kp: 3.853985144980333 1/Ki: 87.78173053095107 Kd: 325.9599328488931
```
(Note that the Ki value is already inverted ready for use in config.py)
------
## Sanity checking the results
If you run
```
python zieglernicols.py zn.csv --showplot
```
It will display a plot of the parameters. It should look simular to this ![kiln-tuner-example.png](kiln-tuner-example.png)
(Note: you will need python's `pyplot` installed for this to work.)
The smooth linear part of the chart is very important. If it is too short, try increasing the target temperature (see later).
Note the red diagonal line: this **must** follow the smooth part of your chart closely.
## My diagonal line isn't right
You might need to adjust the line parameters to make it fit your data properly. You can do this as follows:
```
python zieglernicols.py zn.csv --tangentdivisor 8
```
`tangentdivisor` modifies which parts of the profile is used to calculate the line.
It is a floating point number >= 2; If necessary, try varying it till you get a better fit.
## Changing the target temperature
By default it is 400. You can change this as follows:
```
python kiln-tuner.py zn.csv --targettemp 500
```
(where the target temperature has been changed to 500)

View File

@ -90,7 +90,7 @@ def calculate(filename, tangentdivisor, showplot):
Kd = Kp * Td
# outut to the user
print(Kp, 1 / Ki, Kd)
print(f"Kp: {Kp} 1/Ki: {1/ Ki}, Kd: {Kd}")
if showplot:
plot(xdata, ydata,