Control Charts in Statistical Process Control

Every process varies: no two parts are exactly the same. The question is whether the variation is the normal, predictable noise of a stable process, or a sign that something has changed: a worn tool, a new batch of material, a different operator or a drifting sensor. Control charts, developed by Walter Shewhart at Bell Laboratories in the 1920s, answer that question in real time, so problems are caught before many bad parts are made.

Common causes and special causes

• Common cause variation is inherent in the process: many small, random influences such as minor temperature fluctuations, material variation and machine vibration. Reducing it requires changing the process itself, usually a management or engineering decision.
• Special (assignable) cause variation comes from specific events: a broken tool, a wrong setting, a contaminated batch. It should be found and removed by the people running the process.

Reacting to common cause noise as if it were a special cause (tampering) actually increases variation, while ignoring special causes lets defects pile up. Control charts help tell the two apart.

Anatomy of a control chart

Download diagram (WebP)

Types of control charts

Variables charts (measurements)

• X-bar and R chart: small subgroups (typically n = 2 to 10, often 5) are taken at regular intervals. The X-bar chart tracks subgroup means (process centring); the R chart tracks subgroup ranges (process spread).
• X-bar and S chart: uses standard deviation instead of range, better for larger subgroups (n above about 10).
• Individuals and moving range (I-MR): for one measurement at a time, such as batch processes or slow production.
• CUSUM and EWMA charts: sensitive to small, gradual shifts.

Attributes charts (counts)

• p chart: proportion defective in samples of varying size.
• np chart: number defective in samples of constant size.
• c chart: number of defects per unit of constant size, for example scratches per panel.
• u chart: defects per unit when the inspection area varies.

Control limit formulas for X-bar and R charts

With k subgroups of size n, the grand mean X̿ and average range R̄:

$$UCL_{\bar{X}} = \bar{\bar{X}} + A_2\bar{R} \qquad LCL_{\bar{X}} = \bar{\bar{X}} - A_2\bar{R}$$ $$UCL_R = D_4\bar{R} \qquad LCL_R = D_3\bar{R}$$

For n = 5, the constants are A₂ = 0.577, D₃ = 0 and D₄ = 2.114 (from standard SPC tables).

Rules for detecting out of control signals

Besides points outside the limits, the Western Electric and Nelson rules look for non-random patterns, for example:

1. One point beyond 3σ.
2. Two of three consecutive points beyond 2σ on the same side.
3. Four of five consecutive points beyond 1σ on the same side.
4. Eight consecutive points on one side of the centre line (a shift).
5. Six points in a row steadily increasing or decreasing (a trend), such as tool wear.

More rules catch smaller shifts sooner but also raise false alarms, so organisations choose a sensible set.

Implementing control charts

1. Choose critical characteristics linked to customer requirements.
2. Make sure the measurement system is capable (gauge R&R), since measurement error appears as process variation.
3. Decide subgroup size and frequency to capture variation between, not within, subgroups.
4. Collect initial data, calculate trial limits, investigate and remove special causes, and recalculate.
5. Train operators to plot points and follow a reaction plan when a signal appears.
6. Review charts regularly and recalculate limits after deliberate process changes.

Choosing the right chart

The choice depends on the data type and how samples are taken. If the characteristic is measured on a continuous scale, such as length, weight or temperature, and parts can be grouped into rational subgroups, use X-bar and R or X-bar and S charts. If only one measurement is available at a time, use an individuals and moving range chart. If parts are simply judged good or bad, use p or np charts, and if defects are counted on each unit, use c or u charts. When small, slow drifts matter, such as tool wear or catalyst ageing, consider EWMA or CUSUM charts.

Control charts in modern manufacturing

The value of a chart lies in the reaction to it, so every chart needs a named owner and a clear response when a signal appears.

Today, SPC software collects data directly from gauges, CMMs and machine sensors, plots charts in real time on shop floor screens and alerts engineers automatically. Control charts are also the core tool of the Control phase of Six Sigma DMAIC, and they help eliminate defects, one of the seven wastes of lean manufacturing. The same ideas apply to process plants, where variables such as temperature, pressure and flow from process instruments are monitored with control charts for consistent quality.