PadhaikoAll subjects

Statistical Quality Control

Probability and Statistics (ENSH 304)

Open in the Padhaiko app

5.1 Quality control and its importance in engineering

Statistical Quality Control (SQC) is the application of statistical methods—such as control charts, process capability analysis, and acceptance sampling—to monitor, control, and improve manufacturing and service processes so that products consistently meet engineering specifications.

5.1.1 Chance vs. Assignable Causes of Variation

Every engineering process exhibits variation. Shewhart classified process variation into two distinct categories:

AspectChance (Common) CausesAssignable (Special) Causes
OriginInherent, unavoidable random micro-variations built into the machine, material, and environmentSpecific, identifiable disruptions (tool wear, miscalibration, defective raw-material batch, operator error)
MagnitudeMany small, independent fluctuationsLarge shifts, drifts, or spikes relative to baseline noise
Process StateProcess is in statistical control (stable and predictable over time)Process is out of statistical control (unpredictable)
Corrective ActionCannot be removed by operator tweaking; requires fundamental engineering/system redesignMust be investigated, diagnosed, and eliminated immediately

5.1.2 Importance of Quality Control in Engineering

Definition

Process Capability Indices (Cp and Cpk)

For a process in statistical control with mean , standard deviation , Upper Specification Limit (), and Lower Specification Limit ():


5.2 Control charts for variables (-chart, -chart, and -chart)

A control chart plots a quality statistic over time against three horizontal reference lines:

where is the plotted sample statistic (, , or ). If a point falls outside (or exhibits non-random patterns such as consecutive points on one side of or a monotonic trend), an assignable cause is suspected.

(Note: - and -charts are variables control charts for continuous measurements, whereas the -chart is an attributes control chart for the fraction of nonconforming units.)

5.2.1 -Chart and -Chart (Variables Control Charts)

Suppose rational subgroups of size each are collected. For subgroup (), compute its sample mean and range . Let:

Definition

Control Limits for X-bar and R Charts

Using the standard Shewhart subgroup constants () for sample size :

  1. -Chart (Monitors Process Dispersion/Variability):

(Always inspect the -chart before the -chart, because is used to set the width of the -chart limits.)

  1. -Chart (Monitors Process Centering/Mean):

(When population is known, ; otherwise estimate .)

Standard Shewhart Control Chart Constants
Subgroup Size ()
Example 5.1

5.1

Ten rational subgroups of size piston-ring diameters (in ) yield the following subgroup means and ranges :

Subgroup

Given , , , and for :

  1. Construct the control limits for the -chart and -chart.
  2. State whether the process is in statistical control, and estimate from .

Compute grand mean and average range :

Compute -chart and -chart control limits:

  • -chart:
  • -chart:

Evaluate process state and estimate :

  • All ranges satisfy , so process variability is in control.
  • On the -chart, subgroup has : the process mean is out of control at subgroup (indicating an assignable cause).
  • Process standard deviation estimate: .
Answer:
  1. -chart: , , ; -chart: , , . 2. -chart is in control, but -chart is out of control at subgroup (); .

5.2.2 -Chart (Control Chart for Fraction Defective)

When quality is recorded as pass/fail (conforming vs. nonconforming), the -chart tracks the sample proportion defective across samples of size (based on the Binomial distribution with ).

Definition

Control Limits for the p-Chart

Let be the number of defective items in sample of constant size (). The average fraction defective and control limits are:

(If , is set to because a proportion cannot be negative.)

Example 5.2

5.2

Ten consecutive inspection lots of printed circuit boards each had the following counts of defective boards:

Construct the -chart ( limits) and determine whether any lot is out of control.

Compute average fraction defective :

Compute standard error and control limits:

Inspect sample proportions : Lot has . All other lots range from to , well inside . Thus lot is out of control.

Answer:

, , . Lot () is out of control.


5.3 Six sigma concepts

Six Sigma () is a rigorous, data-driven quality and process-improvement methodology aimed at near-zero defects by systematically measuring and reducing process variability.

5.3.1 Meaning of Quality and DPMO

In a centered process, the nearest specification limit sits from the mean ( if centered, or with a typical long-term mean drift). Under the Six Sigma standard, the specification limits sit from the target mean; even if the long-term process mean drifts by toward either limit, the distance to the nearest specification limit remains , giving a defect rate of:

(i.e., defect-free performance).

Definition

Core Six Sigma Metrics

Sigma Quality Levels (With Standard 1.5-Sigma Long-Term Shift)
Sigma LevelProcess Yield ()Defect Rate ()Typical Engineering Interpretation
Uncompetitive, high scrap and rework
Traditional baseline manufacturing
Industry average for well-managed plants
High-reliability engineering operations
World-class Six Sigma performance

5.3.2 DMAIC and DMADV Methodologies

Six Sigma projects execute through structured five-phase frameworks:

  1. DMAIC (used to improve an existing process):
    • D — Define: Identify customer requirements (CTQs), project scope, and problem statement.
    • M — Measure: Collect baseline process data, validate measurement systems, and compute current .
    • A — Analyze: Identify root causes of variation and defects using Pareto charts, fishbone diagrams, hypothesis tests, and regression.
    • I — Improve: Design and test process changes (e.g., Design of Experiments) to eliminate root causes and center the process.
    • C — Control: Sustain gains using standard operating procedures and statistical control charts ().
  2. DMADV (Design for Six Sigma — used to create a new product or process from scratch): Define, Measure, Analyze, Design, Verify.

Organizational Belt Structure: Champions (executive sponsors), Master Black Belts (technical mentors), Black Belts (full-time project leaders), Green Belts (part-time team practitioners), and Yellow Belts (process support members).

Example 5.3

5.3

  1. (DPMO) An electronics line inspects power-supply modules, each having critical solder-joint opportunities, and records total defects. Compute and , and state the approximate sigma level.
  2. (Process Capability) A shaft-turning process is in statistical control with and against specifications and . Compute and and interpret the result.

Part 1 — Compute and :

From the sigma table, corresponds to approximately a process ( at ).

Part 2 — Compute and :

Because , the process has potential to improve by recentering from to , though it still falls slightly short of the benchmark.

Answer:
  1. , (); 2. , (off-center toward and below ).

Practice Exercises — Statistical Quality Control & Six Sigma

Question 1

Distinguish between chance (common) causes and assignable (special) causes of variation. Explain why the -chart is always examined before the -chart, and outline the five phases of DMAIC in Six Sigma.

Chance vs. Assignable Causes & Chart Order:

  • Chance causes are inherent random fluctuations in a stable process; assignable causes are large, removable disruptions (tool wear, bad material batch) that push a process out of control.
  • The -chart is checked before the -chart because sets the width of the -chart control limits (); if process spread () is out of control, the limits are invalid.

DMAIC Phases:

  1. Define: Scope the problem and customer requirements.
  2. Measure: Establish baseline data and defect metrics ().
  3. Analyze: Isolate root causes of defects/variation.
  4. Improve: Implement and verify process fixes.
  5. Control: Lock in improvements with control charts and standard procedures.
Answer:

Chance causes are inherent random noise whereas assignable causes are removable process faults. The -chart is evaluated first because determines the -chart's control limits. DMAIC stands for Define, Measure, Analyze, Improve, and Control.

Question 2

Twenty subgroups of size machined pins give and . Given , , , and for , compute the control limits for the -chart and -chart, and estimate the process standard deviation .

Compute and :

Compute control limits and :

  • -chart: , , .
  • -chart: , , .
  • Estimated : .
Answer:

-chart: ; -chart: ; .

Question 3

Across daily samples of injection-molded housings each, a total of defective housings were observed. Determine the center line and control limits for a -chart. If day had defective housings, is the process in control on day ?

Compute and :

Compute limits and check day :

On day , , so the process is out of control on day .

Answer:

, , ; day () is out of control.

Question 4

A CNC milling process operates in statistical control with mean and standard deviation . The engineering tolerance is (, ). Calculate and , and state whether the process meets a requirement.

Compute :

Compute one-sided capability indices and :

Answer:

, : does not meet because the mean is off-center by (recentering to would raise to ).

Question 5
  1. Explain why a Six Sigma () process corresponds to rather than .
  2. A manufacturer inspects control boards, each with defect opportunities, and finds total defects. Compute and .

Why yields : A centered normal distribution at has only . However, real processes experience a typical long-term mean shift of up to , leaving between the shifted mean and the nearest specification limit; the one-sided tail area beyond is , or .

Compute and :

Answer:
  1. A long-term mean shift reduces the distance to the nearest specification limit to , giving ; 2. , .
← Correlation and Regression