pyshewhart - Statistical Process Control Charts

Statistical process control charts (also known as "Shewhart charts" after Walter A. Shewhart) are widely used in manufacturing and industry as a quality-control tool.

The aim of this project is to provide an easy-to-use Python module for generating the following types of control charts:

• X̅ and R
• X̅ and S
• Cumulative Sum (CUSUM)
• P-attribute

In addition to checking the data against the standard 3σ control limits, it can also be evaluated against the well-known "Western Electric rule set" to detect additional statistical anomalies:

---

Usage

As a Python library:

An X̅ and R (mean and range) chart can be generated as simply as:

import random
import pyshewhart

measurements = [random.random() for _ in range(200)]
pyshewhart.XbarR(measurements)

---

If you have time information, you can also provide that. Supported time types include:

• datetime
• timedelta
• various string formats (as supported by dateutil.parser.parse)
• Unix timestamps
• elapsed seconds (int or float)

seconds = range(200)
pyshewhart.XbarR(seconds, measurements)

Notice that now the plot shows both sample numbers and elapsed seconds.

---

Console Script:

Given a CSV file field_strengths_nT.csv containing timeseries and measurement data:

2022-10-12, 59194.401
2022-10-13, 59176.972
2022-10-14, 59179.485
2022-10-15, 59189.235
      ...
      ...
2023-10-08, 59196.725
2023-10-09, 59194.435
2023-10-10, 59187.456
2023-10-11, 59196.135

we can generate an X̅ and S (mean and standard deviation) plot with the following shell command:

$ pyshewhart field_strengths_nT.csv xbar-s 7 --units "nT" --title "Field Strength"

---

Examples

A few other examples can be found in the examples/ directory.

Several of these are recreated from a well-known textbook titled Managing Quality: Integrating the Supply Chain by S. Thomas Foster, including this X̅ and R chart:

Also a P-attribute chart:

The following shows the capabilities of a CUSUM chart to detect small amounts of drift.

First we see normal data:

And here we see the chart detecting a small amount of simulated drift: