Process Behavior Chart - ProcessBehavior

Process behavior chart is Donald Wheeler’s name for what most textbooks call a control chart — and the rename is the point. “Control” suggests the chart’s job is keeping a process inside specifications; Wheeler’s term says what the chart actually does: it characterizes how the process behaves, so you can tell routine variation from a genuine change. Same math (Shewhart’s), better name. This library takes its own name from Wheeler’s usage; the terminology appendix maps his vocabulary to the textbook terms.

The 60-second version¶

import numpy as np
import pandas as pd
from processbehavior import ProcessBehavior

rng = np.random.default_rng(12)
temps = rng.normal(150.0, 3.0, 30)
temps[20:] -= 8            # the process changed at batch 21

df = pd.DataFrame({'batch': range(1, 31), 'seal_temp': temps.round(1)})

study = ProcessBehavior(df).formulate(response='seal_temp', time='batch')
result = study.execute(chart='X', companion=True)

s = result.get_statistics('X')
print(f"Center line: {s['center']}")
print(f"Natural process limits: ({s['lpl']}, {s['upl']})")
print(f"Signals: {result.detect_signals(chart='X').count}")
result.plot()

Output:

Center line: 147.38
Natural process limits: (139.07, 155.69)
Signals: 26

Reading it in Wheeler’s terms¶

The limits are natural process limits — the voice of the process, not spec limits (the voice of the customer). They say what the process will do, not what you wish it would.
The planted change at batch 21 lights the chart up: the run of points above the center line before the drop and below it after both violate run rules, on top of the points beyond the limits. Routine variation doesn’t do that — that is the signal/noise distinction the chart exists to make.
Predictable (Wheeler) = “in control” (textbook): only routine variation. This process is not predictable — something changed, go find it.

Going further¶

Your First X/mR Chart — the guided version of exactly this example.
Wheeler terminology — the full vocabulary map.
Design-State Tour — what happens when your data has structure (factors, replication).