Skip to content
Omnikalc

CUSUM chart design calculator

Enter the process standard deviation, the shift you want to catch and the decision interval to get the reference value and the average run length in control and after the shift.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Results

Reference value K

1.000

Decision interval H
10.000
In-control average run length (two-sided CUSUM)
469
Average run length after the shift
10.3
Average run length of a Shewhart chart with 3 sigma limits after the same shift
43.9

The CUSUM chart calculator gives the reference value K, the decision interval H and the average run lengths of a cumulative sum chart, in control and after a shift, and compares them with a Shewhart chart.

How the calculation works

  • Enter the process standard deviation.
  • Enter the shift to detect in standard deviations and the decision interval h.
  • The calculator gives K, H, the in-control average run length, the average run length after the shift and the value for a Shewhart chart.

Example

For a process with a standard deviation of 2, a shift of 1 sigma to detect and h = 5, the reference value is K = 0.5 × 2 = 1.000 and the decision interval is H = 10.000. The in-control average run length is about 469 samples, and the shift is caught after about 10.3 samples, against 43.9 for a 3 sigma Shewhart chart.

Why it matters

The calculator helps quality engineers, process analysts and students to design a CUSUM chart that finds small shifts of a mean quickly, in manufacturing, laboratories or monitoring of any measurement.

Good to know

Choose the shift that matters in practice, not the smallest you can imagine, since a design for a very small shift needs a long time to signal anyway. Standardize the data with the target and the standard deviation if you work in units of sigma. Two sums are kept, one for upward shifts and one for downward shifts, and the in-control run length of the two-sided chart is about half of the one-sided value. When the signal comes, the start of the shift is estimated as the last time the sum was zero. The approximation assumes independent normal data, and autocorrelated data give more false alarms than predicted.

The control chart limits calculator sets Shewhart limits, the EWMA chart calculator offers another way to catch small shifts, the p-chart calculator handles proportions and the process capability calculator compares a process with its limits.

Frequently asked questions

What is a CUSUM chart?

A cumulative sum chart adds up the deviations of the measurements from a target, minus an allowance K. The sum is reset to zero when it goes negative, and a signal is given when it exceeds the decision interval H. Because it accumulates small deviations, it detects a small steady shift much faster than a Shewhart chart.

How are K and H chosen?

K is usually half the shift you want to detect, so K = δ/2 in standard deviations, and H is 4 to 5 standard deviations. A larger H gives fewer false alarms but a slower detection. The calculator uses Siegmund’s approximation for the average run lengths, which is accurate to a few percent for usual designs.

How does it compare with a Shewhart chart?

For a shift of one standard deviation, a Shewhart chart with 3 sigma limits needs about 44 samples to signal, while this CUSUM needs about 10 with a similar false alarm rate. For large shifts of 3 sigma or more, the Shewhart chart is as quick, so the two are often used together.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.