SPC in Manufacturing: Control Charts for Fabrication

Statistical process control (SPC) in manufacturing, including fabrication shops, uses control charts to separate normal process variation from signals that something has changed. A drift in cut length, hole size or weld size is then caught before parts are scrapped. SPC is one of the core tools of Six Sigma, and it complements the process controls in the ISO 3834 welding quality system.

This guide covers which chart to use, how to build an X-bar and R chart with a worked example, how to read signals and how to calculate capability. It ends with where SPC pays off in a fabrication shop, from dimensions to hardness checks (see the guide to hardness testing methods), and the mistakes that make charts useless.

Quick answerSPC in fabrication monitors measurements from the shop floor on control charts. Plot subgroup means and ranges against limits calculated from the process itself, not from the drawing tolerance. A point beyond a control limit, or a run of points on one side of the centre line, signals a special cause to investigate. Capability indices Cp and Cpk then compare the stable process spread with the tolerance.

Figure 1: A control chart with centre line and limits beside the measured steel part it describes.

Key takeaways

Control limits come from the process; specification limits come from the drawing or code. They are different things.

Check the range chart first, then the means chart, because unstable spread makes the means chart unreliable.
A stable process is a precondition for capability indices such as Cp and Cpk.
Use variable charts for measurements and attribute charts for counts of defects or defectives.

Only act on signals; adjusting a stable process in response to every point makes variation worse.

What Is SPC in Fabrication?

Statistical process control is a method monitoring process data on control charts to detect variation needing action before defects are made.

In a fabrication shop the measured characteristics are usually dimensional or process readings: the length of cut parts, hole diameters, fit-up gaps, weld leg size, distortion, dry film thickness of paint, preheat temperature and hardness readings. SPC samples these at planned intervals, plots them in time order and compares them with limits that describe what the process normally does. The shop then reacts to real changes and leaves normal variation alone.

The method was developed by Walter Shewhart in the 1920s and remains the basis of modern quality practice. It is a form of prevention: the cost of a signal investigated on the bench is far lower than the cost of scrap, rework or a rejected delivery. The wider improvement framework that uses SPC is Six Sigma, covered in the conclusion.

What Are Common and Special Causes of Variation?

Every process varies. Common-cause variation is the steady, random scatter built into the process, such as small differences in material, machine and operator technique. Special-cause variation comes from a specific, identifiable change, such as a worn blade, a new heat of steel, a changed jig or a different operator. Control charts exist to tell the two apart.

Table 1: Common-cause and special-cause variation in a fabrication shop
Feature	Common cause	Special cause
Pattern on chart	Random scatter inside the limits	Points beyond limits, runs, trends or shifts
Typical example	Normal cut-length scatter of a stable saw	A worn blade moving the mean cut length
Who acts	Management, through process redesign	The operator or engineer, by finding the cause
Right response	Improve the process; do not adjust per reading	Investigate and remove the cause

The distinction prevents two costly errors: reacting to noise, which adds variation, and ignoring a real change, which lets defects accumulate.

Which Control Chart Should You Use?

The chart depends on the type of data and the subgroup size. Measurements such as lengths and thicknesses use variable charts. Counts of defective items or defects use attribute charts. The decision flow below covers the common cases.

Figure 2: Choosing a control chart. Variable data give more information per reading than attribute data and need smaller samples.

Table 2: Control chart types and typical fabrication uses
Chart	Data	Fabrication example
X-bar and R	Measurements, subgroups of 2 to 9	Cut length, hole diameter measured on 4 parts per batch
Individuals and moving range	Single measurements, one per period	Preheat temperature of each joint, furnace soak temperature
p chart	Fraction of defective items, variable sample size	Fraction of welds rejected by radiographic testing per week
np chart	Number of defective items, constant sample size	Number of rejected parts in each lot of 50
c chart	Count of defects in a constant area or unit	Surface defects per painted panel
u chart	Defects per unit, variable unit size	Weld imperfections per metre of weld

How Do You Build an X-bar and R Chart?

An X-bar and R chart plots the mean and the range of each small subgroup of measurements, with limits calculated from the data. The steps are:

Choose the characteristic to chart and confirm that the measurement system is accurate and repeatable.

Define a rational subgroup, usually 3 to 5 consecutive parts, and a sampling interval.
Collect 20 to 25 subgroups while the process runs normally.
Calculate the mean and range of each subgroup, then the grand mean and the mean range.

Calculate the centre lines and control limits using the constants for the subgroup size.
Plot the ranges first, then the means, and look for points outside limits and for patterns.
Investigate and remove any special causes, recalculate the limits from the stable data, and use the chart for ongoing monitoring.

Figure 3: Samples measured at planned intervals on the shop floor supply the data for control charts.

A subgroup should be a rational one: parts made close together in time under the same conditions, so that variation inside the subgroup is common-cause only and any change shows between subgroups. Mixing two machines or two shifts in one subgroup hides exactly the differences the chart is meant to reveal. The measurement system matters as much as the process; gauge variation that is large relative to process variation makes every chart noisy, which is why a gauge study is done first. For hardness or other test data used on charts, see the hardness conversion calculator for how readings are converted between scales.

What Does a Worked X-bar and R Example Look Like?

A plasma cutting table produces parts with a length tolerance of ±1.5 mm. Four parts are measured each hour and the deviation from nominal length is recorded. Ten subgroups of four give the following means and ranges.

Table 3: Cut length deviation, subgroups of four (mm)
Subgroup	Readings	Mean	Range
1	0.4, -0.2, 0.1, 0.3	0.150	0.6
2	-0.1, 0.5, 0.2, -0.3	0.075	0.8
3	0.2, 0.0, -0.4, 0.3	0.025	0.7
4	0.6, 0.1, -0.2, 0.2	0.175	0.8
5	-0.3, 0.4, 0.1, 0.0	0.050	0.7
6	0.1, -0.5, 0.3, 0.2	0.025	0.8
7	0.5, 0.2, -0.1, 0.4	0.250	0.6
8	0.0, -0.2, 0.3, -0.4	-0.075	0.7
9	0.3, 0.6, 0.0, 0.1	0.250	0.6
10	-0.2, 0.1, 0.4, -0.1	0.050	0.6

Step 1: grand mean and mean range X-double-bar = 0.975 / 10 = 0.0975 mm R-bar = 6.9 / 10 = 0.69 mm Step 2: control limits for n = 4 (A2 = 0.729, D3 = 0, D4 = 2.282) UCL(X-bar) = 0.0975 + 0.729 x 0.69 = 0.601 mm LCL(X-bar) = 0.0975 – 0.729 x 0.69 = -0.406 mm UCL(R) = 2.282 x 0.69 = 1.575 mm LCL(R) = 0 Step 3: check stability All ranges (0.6 to 0.8) are below 1.575. All means (-0.075 to 0.250) are inside -0.406 to 0.601. The process is in statistical control. Step 4: capability against the +/-1.5 mm tolerance (d2 = 2.059) sigma estimate = R-bar / d2 = 0.69 / 2.059 = 0.335 mm Cp = (1.5 – (-1.5)) / (6 x 0.335) = 1.49 Cpk = min(1.5 – 0.0975, 0.0975 + 1.5) / (3 x 0.335) = 1.40

The ranges and means show a stable process, so the limits can be used to monitor future subgroups. Cp is 1.49 and Cpk is 1.40, so the process spread fits the tolerance with a little room to spare and the mean is close to the middle of the tolerance. The figure below extends the chart with eight further subgroups to show how a shift appears.

Figure 4: X-bar chart extended to 18 subgroups. Eight consecutive means above the centre line signal a shift even though no point is beyond a control limit.

How Do You Read Control Chart Signals?

A point outside the control limits is the clearest signal, but patterns inside the limits matter too. The Western Electric rules are a widely used set; some rule sets use slightly different run lengths, so a shop should fix its own and apply them consistently.

Table 4: Common control chart signals
Signal	What it usually indicates
One point beyond a 3-sigma control limit	A sudden change, such as a tool break or a setup error
Two of three points beyond 2 sigma on the same side	A developing shift in the mean
Four of five points beyond 1 sigma on the same side	A smaller sustained shift
Eight points in a row on one side of the centre line	A shift in the mean, as in the extended example (some rule sets use nine)
Six points steadily rising or falling	A trend such as tool wear or temperature drift

When a signal appears, stop and ask what changed: material heat, tool, fixture, operator, gauge, environment or method. Record the cause and the action on the chart. If the cause is removed, the chart should return to its normal pattern. Control limits are recalculated only after a real, permanent process change, not to make the chart look better.

What Are Cp and Cpk?

Cp and Cpk are capability indices that compare the width of a stable process with the tolerance. Cp measures the potential capability if the process were centred, and Cpk accounts for how far the mean sits from the nearer specification limit.

Capability indices Cp = (USL – LSL) / (6 sigma) Cpk = min(USL – mean, mean – LSL) / (3 sigma) sigma is the within-subgroup estimate, R-bar / d2, for a stable process.

Table 5: Commonly used interpretation of Cpk
Cpk	Typical reading
Below 1.00	Process not capable; defects expected
1.00 to 1.33	Marginal; needs tight control
1.33 or above	Capable; a common minimum target
1.67 or above	Required by some customers for critical characteristics

These thresholds are customer and industry conventions, not universal rules, so use the value stated in the contract or customer standard. Capability indices assume a stable process and roughly normal data. If the chart shows special causes, fix them before quoting capability, and treat results from fewer than about 25 subgroups as provisional. A related index, Ppk, uses the overall standard deviation and describes long-term performance.

Where Does SPC Apply in a Fabrication Shop?

SPC pays off where a characteristic is measured regularly, matters to the customer and can be adjusted. It is less useful for one-off parts, where there is no process history to chart.

Table 6: Candidate SPC applications in fabrication
Characteristic	Chart	Why it is worth charting
Cut length or hole position	X-bar and R	Detects tool wear and table drift before parts fail to fit
Preheat and interpass temperature	Individuals and MR	Shows whether welding conditions stay inside procedure limits, linked to the preheat temperature calculator
Weld cooling time between 800 and 500 °C	Individuals and MR	Tracks cooling conditions that affect hardness and toughness, see the weld cooling time t8/5 calculator
Dry film thickness of paint	X-bar and R	Controls coating thickness against minimum and maximum limits
Radiographic or ultrasonic rejection rate	p chart	Shows whether weld quality is stable between lots and welders
Weld imperfections per metre	u chart	Compares visual and NDT findings across jobs
Hardness of heat-treated parts	X-bar and R	Detects furnace or quench drift, see the guide to quenching steel
Grain size or microstructure counts	Individuals and MR	Monitors incoming material, see the ASTM grain size number calculator

The same data feed supplier and quality-system reviews. Charts of incoming material properties support vendor evaluation, and results from third-party inspectors can be charted the same way, as discussed in the guide to third-party inspection in welding. For test data, the Charpy impact testing guide describes how impact energies are generated and why they scatter. Related guides are in the Materials Testing category and the Manufacturing Metallurgy category.

What Are Common SPC Mistakes?

Requirements vary. The chart rules, targets and sampling plans that apply depend on the customer, the contract and the standards in force. Use this guide as a starting point and adapt it to the project.

Confusing control limits with specification limits. A part inside the tolerance can still come from an out-of-control process, and an in-control process can still make out-of-tolerance parts.
Setting control limits from the tolerance. Limits must be calculated from the process data.
Skipping the measurement system check. If gauge error is large, the chart cannot show real changes.

Mixing process streams in a subgroup. Different machines, shifts or heats in one subgroup inflate the range and widen the limits.
Tampering. Adjusting the process after every reading increases variation.
Quoting capability on an unstable process. Cp and Cpk are meaningful only when the process is in control.
Recalculating limits too often. This hides the very shifts the chart exists to find.

Charting without acting. A chart that nobody investigates is a record, not a control.

Quick Reference: Control Chart Constants and Formulae

Table 7: Constants for X-bar and R charts by subgroup size n
n	A2	D3	D4	d2
2	1.880	0	3.267	1.128
3	1.023	0	2.574	1.693
4	0.729	0	2.282	2.059
5	0.577	0	2.114	2.326
6	0.483	0	2.004	2.534
7	0.419	0.076	1.924	2.704

Table 8: Control limit formulae
Chart	Centre line	Control limits
X-bar	Grand mean	Grand mean ± A2 x R-bar
R	R-bar	LCL = D3 x R-bar; UCL = D4 x R-bar
Individuals	Mean of readings	Mean ± 2.66 x mean moving range
Moving range	Mean moving range	UCL = 3.267 x mean moving range
p chart	p-bar	p-bar ± 3 sqrt(p-bar (1 – p-bar) / n)
c chart	c-bar	c-bar ± 3 sqrt(c-bar)

Negative lower limits on attribute charts are set to zero. The constants above are the standard published values; check them against the standard or reference manual your customer specifies.

Key Terms

Statistical process control (SPC): The use of control charts and statistics to monitor and improve a process.
Control chart: A time-ordered plot of process data with a centre line and control limits.
Control limits: Limits calculated from process data, normally three standard deviations either side of the centre line.
Special-cause variation: Variation from a specific, identifiable change in the process.
Rational subgroup: A small group of measurements taken under the same conditions so that variation within it is common-cause only.
Process capability (Cp, Cpk): Indices comparing the spread and centring of a stable process with the tolerance.
Specification limits: The upper and lower limits set by the drawing, code or customer.
Tampering: Adjusting a stable process in response to normal variation, which increases variation.

Frequently Asked Questions

What is SPC in manufacturing?

Statistical process control is a method that monitors process data on control charts to separate normal variation from signals that need action. It lets a shop correct a drifting process before it produces defects, rather than relying on final inspection alone.

What is the difference between control limits and specification limits?

Control limits are calculated from the process data and describe what the process normally does. Specification limits come from the drawing, code or customer and describe what is acceptable. A process can be in control and still make parts outside tolerance.

How many subgroups are needed to set up a control chart?

Guidance commonly given is 20 to 25 subgroups of three to five readings from a stable process. Fewer subgroups give less reliable limits and should be treated as provisional until more data accumulate.

What do Cp and Cpk mean?

Cp compares the tolerance width with the process spread of six standard deviations. Cpk also accounts for how far the process mean is from the nearer specification limit. Both assume a stable process, and 1.33 is a commonly used minimum Cpk target.

When should control limits be recalculated?

Recalculate limits only after a real, permanent change in the process, such as new equipment or a revised method, or after removing special causes from the initial data. Do not recalculate to make a chart look in control.

Can SPC be used for welding?

Yes, for measurable characteristics such as preheat temperature, heat input, weld size and cooling time, and for attribute data such as rejection rate per lot. It does not replace the qualification of procedures and welders required by the code.

What is the difference between a p chart and a c chart?

A p chart tracks the fraction of items that are defective, such as welds rejected per lot. A c chart tracks the count of defects in a constant unit, such as surface defects per panel. Use u charts when the unit size varies.

Conclusion

SPC in fabrication turns routine measurements into an early warning system. Choose the chart that matches the data, build it from a rational subgroup and a checked gauge, calculate limits from the process and respond only to real signals. Once the chart is stable, Cp and Cpk show how the process spread compares with the tolerance. The rules and targets that apply depend on the customer, contract and standards in force. To place SPC inside a complete improvement method, with define, measure, analyse, improve and control phases, the next step is the Six Sigma guide.

About This Guide

This guide was prepared by the MetallurgyZone technical team from the sources listed under Standards and References. Verify requirements against the standard edition, customer specification and contract applicable to your application.

Standards and References

ISO 7870-2, Control charts: Shewhart control charts, International Organization for Standardization (iso.org).

ISO 22514 series, Statistical methods in process management: Capability and performance, ISO.
ISO 13053, Quantitative methods in process improvement: Six Sigma, ISO.
ASTM E2587, Standard Practice for Use of Control Charts in Statistical Process Control, ASTM International (astm.org).

Statistical Process Control (SPC) Reference Manual, Automotive Industry Action Group (aiag.org).
D. C. Montgomery, Introduction to Statistical Quality Control, Wiley.
W. A. Shewhart, Economic Control of Quality of Manufactured Product, 1931.