Quality Engineering and Statistical Process Control

Introduction

Quality engineering is the systematic design, measurement, and improvement of products and processes so that requirements are met consistently. In machine technology, quality is not limited to final inspection. It begins with understanding customer and regulatory requirements, continues through product and process design, and extends to production, delivery, service, and continual improvement.

A modern quality system combines engineering analysis with management discipline. Statistical process control (SPC) distinguishes natural process variation from unusual variation. Process capability evaluates whether a stable process can meet specifications. Measurement-system analysis determines whether data are trustworthy. Acceptance sampling provides a rational method for making decisions about batches when complete inspection is costly or destructive. Total quality management (TQM), ISO 9001, Six Sigma, and design of experiments (DOE) provide broader improvement frameworks.

Learning Objectives

After studying this chapter, the student should be able to:

1. Explain major quality philosophies and quality-management systems.
2. Apply the seven basic quality tools to manufacturing problems.
3. Select and interpret appropriate control charts.
4. Calculate and interpret process capability indices.
5. Explain acceptance sampling and measurement-system analysis.
6. Compare TQM, ISO 9001, Six Sigma, and DOE approaches.
7. Recommend defect-prevention and improvement actions using evidence.

1. Quality as a System Property

Quality has several related meanings. A product may conform to drawings yet fail in service if the design is inadequate. Conversely, a capable process may produce a product that does not satisfy customer needs because the specifications are wrong. Quality therefore involves both design quality—how well the design fulfills intended use—and conformance quality—how consistently production follows the design.

Important dimensions include performance, reliability, durability, serviceability, aesthetics, safety, and perceived value. In engineering, quality requirements should be translated into measurable characteristics such as diameter, hardness, surface roughness, tensile strength, leakage rate, or assembly force.

A useful distinction is:

• Common-cause variation: inherent random variation from the combined effect of many routine factors.
• Special-cause variation: variation attributable to a specific, identifiable change or abnormal condition.
• Defect: failure to meet a requirement.
• Defective unit: a unit containing one or more defects.

Reducing inspection alone does not create quality. Prevention is generally stronger than detection because defects are stopped near their source, reducing scrap, rework, warranty cost, and risk to users.

2. Quality Philosophies and Management Systems

2.1 Deming, Juran, and Crosby

W. Edwards Deming emphasized system thinking, reduction of variation, leadership responsibility, and the Plan–Do–Study–Act cycle. His view was that most quality problems arise from the system rather than from individual workers.

Joseph Juran described quality as fitness for use and organized quality management around the quality trilogy: quality planning, quality control, and quality improvement. Planning identifies customers and requirements; control compares actual performance with targets; improvement removes chronic causes of poor performance.

Philip Crosby defined quality as conformance to requirements and promoted prevention, the idea that “zero defects” is a performance standard rather than an assertion that mistakes are impossible. His approach stresses the cost of nonconformance.

These philosophies differ in emphasis but share several principles: leadership commitment, customer focus, prevention, process measurement, employee involvement, and continual improvement.

2.2 Total Quality Management

TQM is an organization-wide approach in which every function contributes to customer satisfaction and process improvement. Its principles include:

• customer-defined quality;
• process orientation;
• prevention rather than correction;
• evidence-based decisions;
• involvement of people;
• supplier partnership; and
• continual improvement.

TQM is a management philosophy, not merely a set of statistical tools. A control chart cannot compensate for unclear requirements, inadequate equipment, or leadership that rewards output while ignoring defects.

2.3 ISO 9001

ISO 9001 specifies requirements for a quality-management system. It is based on process thinking, risk-based thinking, documented information, leadership, customer focus, competence, performance evaluation, and improvement. Certification indicates that an organization’s management system has been audited against the standard; it does not guarantee that every product is defect-free.

A process approach identifies inputs, activities, outputs, resources, controls, risks, and measures. Typical controlled information includes drawings, work instructions, inspection plans, calibration records, training records, nonconformance reports, and corrective-action evidence. The Plan–Do–Check–Act cycle provides a practical structure for operating and improving the system.

2.4 Six Sigma

Six Sigma is a disciplined, data-based approach to reducing process variation and defects. The common improvement roadmap is DMAIC:

• Define: clarify the problem, customer requirements, scope, and goals.
• Measure: establish operational definitions and obtain reliable baseline data.
• Analyze: identify and verify important causes.
• Improve: develop, test, and implement solutions.
• Control: sustain gains through standardization and monitoring.

For new designs, DMADV or DFSS (Design for Six Sigma) may be used: Define, Measure, Analyze, Design, and Verify.

The normal distribution is often used as an approximation in capability analysis, but Six Sigma is not simply the pursuit of a numerical sigma value. The important discipline is linking critical-to-quality characteristics to verified causes and sustained controls.

3. The Seven Basic Quality Tools

3.1 Check Sheet

A check sheet is a structured form for collecting facts at the location where work occurs. Categories, time periods, machines, shifts, and defect types should be defined before data collection. Good check sheets use mutually understandable categories and record counts consistently.

3.2 Histogram

A histogram groups numerical observations into intervals and displays their frequency. It can reveal center, spread, skewness, multiple populations, or unusual observations. The shape must be interpreted with knowledge of sampling conditions; mixing shifts or machines may create a misleading multimodal pattern.

3.3 Pareto Chart

A Pareto chart ranks categories from largest to smallest, often with a cumulative-percentage line. It helps prioritize causes or defect types. The “vital few” principle is a practical guide, not a universal law: categories should be defined carefully, and a small count of severe safety defects may deserve priority over a large count of minor cosmetic defects.

3.4 Cause-and-Effect Diagram

The fishbone diagram organizes possible causes under headings such as Manpower, Machine, Method, Material, Measurement, and Environment. It structures investigation but does not prove causation. Suspected causes must be tested with data or controlled experiments.

3.5 Scatter Diagram

A scatter diagram plots paired observations to examine association between two variables, such as cutting speed and surface roughness. Association does not establish causation. A hidden variable, restricted data range, or time order may explain an apparent relationship.

3.6 Flowchart

A flowchart maps process sequence, decisions, feedback loops, queues, and inspection points. It is particularly useful for identifying handoffs, rework loops, and opportunities for mistake-proofing.

3.7 Control Chart

A control chart plots a process statistic against time with a center line and statistically derived control limits. It differs from a run chart because control limits are based on process variation and are used to distinguish common from special causes.

4. Statistical Process Control

SPC is a feedback method: collect data, compare the process with expected behavior, investigate signals, and improve the process. It should be applied to a defined process operated under reasonably consistent conditions.

4.1 Control Limits and Specification Limits

Specification limits come from design, customer, legal, or functional requirements. Control limits are calculated from process data and describe expected process behavior. A process may be stable but incapable of meeting specifications, or it may meet specifications temporarily while being statistically unstable. These are different questions.

For individual approximately normally distributed observations, estimated control limits are often written as:

Here μ is the process mean and σ is the process standard deviation. In practice, μ and σ are estimated from rational subgroups or moving ranges, and the limits are adjusted for the chart statistic. The ±3σ convention is a monitoring rule, not a guarantee that every future point lies within the limits.

4.2 Variables Charts

For subgroup means and ranges, the standard X̄–R chart uses:

X̄ is a subgroup mean, R is a subgroup range, X̄̄ is the average of subgroup means, R̄ is the average range, and A₂, D₃, and D₄ are constants determined by subgroup size. These formulas assume rational subgroups and approximately stable within-subgroup variation. For larger subgroups, an X̄–S chart uses standard deviations rather than ranges. For individual measurements, an Individuals–Moving Range chart is appropriate when rational subgroups of size greater than one are unavailable.

4.3 Attribute Charts

For counts of nonconforming units, use a p chart when subgroup sizes may vary and an np chart when subgroup size is constant. If p̄ is the average fraction nonconforming and n is subgroup size:

Limits below zero are set to zero. The binomial model assumes a fixed classification rule and independent units with a common underlying nonconforming probability, at least approximately.

For the number of defects per unit with constant inspection opportunity, a c chart is used:

For varying opportunity or area, a u chart is more suitable; its center line is defects per unit and its limits depend on the unit count inspected.

4.4 Interpreting Signals

Signals include a point beyond a control limit, a long run on one side of the center line, a trend, cyclic behavior, or an unusual clustering pattern. Rules should be selected before analysis and used consistently. When a signal occurs, preserve evidence, identify what changed, check measurement validity, contain affected product, and investigate the process cause. Do not automatically delete an observation merely because it is inconvenient; removal requires a documented, technically justified cause.

5. Process Capability Analysis

Capability analysis asks whether a stable process can satisfy specification limits. It should not be performed on an unstable process without first addressing special causes.

For a two-sided specification with upper specification limit USL and lower specification limit LSL:

Cp measures potential capability if the process is centered. Cpk measures capability considering centering. μ is the process mean and σ is the within-process standard deviation used for the study. The indices are meaningful only when the characteristic is measured on a suitable scale, the process is reasonably stable, and the distributional assumptions are acceptable or an appropriate nonnormal method is used.

Worked Example

A shaft diameter has specifications 49.90 to 50.10 mm. A stable process has μ = 50.04 mm and σ = 0.025 mm. Then:

The spread is potentially acceptable under some organizational criteria, but the process is poorly centered toward the upper limit. Adjusting the process mean downward, while preserving stability and verifying measurement accuracy, would improve Cpk. Capability targets such as 1.33 or 1.67 are organizational requirements, not universal laws; safety-critical applications may require more stringent criteria.

6. Measurement-System Analysis

A process cannot be improved reliably if its measurement system is inadequate. Measurement-system analysis (MSA) evaluates bias, repeatability, reproducibility, stability, linearity, and resolution.

• Bias: difference between the average measured value and a reference value.
• Repeatability: variation when the same operator measures the same item repeatedly with the same instrument.
• Reproducibility: variation between operators or appraisers.
• Stability: change in measurement performance over time.
• Linearity: change in bias across the measurement range.

A common crossed gauge repeatability and reproducibility study uses several parts, multiple operators, and repeated measurements. Variance components are estimated for repeatability and reproducibility, then compared with total observed variation or tolerance. The exact acceptance criteria depend on the industry and risk; commonly used percentage guidelines are screening rules, not substitutes for engineering judgment.

A useful resolution principle is that instrument discrimination should be sufficiently finer than the tolerance or process variation being evaluated. Calibration establishes traceability to a reference; it does not by itself prove that the measurement system is appropriate for the application. Operators need clear methods, stable fixtures, environmental control, and training.

7. Acceptance Sampling

Acceptance sampling examines a sample from a lot and decides whether to accept or reject the lot. It is appropriate when 100% inspection is impractical, expensive, destructive, or less reliable because of inspector fatigue. It does not improve the production process and should not replace process control.

A single-sampling plan is described by sample size n and acceptance number c. If x nonconforming units are found, accept the lot when x ≤ c; otherwise reject it. For a lot with nonconforming fraction p and sampling without replacement, the exact probability of observing x nonconforming items follows the hypergeometric distribution. When the lot is large relative to n, the binomial approximation is often used:

where C(n,x) is the number of combinations, n is sample size, and p is the lot fraction nonconforming. The probability of acceptance is:

A sampling plan has producer’s risk α, the probability of rejecting a lot at an acceptable quality level, and consumer’s risk β, the probability of accepting a lot at a rejectable quality level. These risks must be specified when selecting a plan. “Accept” means only that the sample result satisfies the plan; it does not prove that every unit is conforming.

8. Design of Experiments for Quality Improvement

DOE changes selected factors deliberately and observes responses, allowing effects and interactions to be estimated efficiently. Factors may include speed, feed, coolant condition, tool geometry, material, or heat-treatment temperature. Responses may include roughness, dimensional error, tool life, or defect rate.

A factorial model for two quantitative or coded factors can be written:

Here y is the response, x₁ and x₂ are coded factor levels, β terms are estimated effects, x₁x₂ represents interaction, and ε is random error. An interaction means the effect of one factor depends on the level of another. Replication estimates experimental error; randomization protects against time-related bias; blocking controls known nuisance variation. Center points can help detect curvature.

ANOVA partitions response variation into model and error components, but statistical significance should be accompanied by practical significance, engineering feasibility, residual checks, and confirmation runs. DOE is stronger than changing one factor at a time when interactions or curvature are important.

9. Integrated Case: Reducing Surface-Roughness Defects

The following is an illustrative case based on a hypothetical machining cell.

Background and Problem

A cell producing turned stainless-steel shafts receives increasing customer complaints about surface roughness. The requirement is Ra ≤ 1.60 μm. The monthly defect rate is 8%, causing rework and delayed shipments.

Analysis

The team first created a flowchart and check sheet. A Pareto analysis showed that roughness accounted for most customer complaints. A cause-and-effect diagram identified tool wear, feed rate, cutting speed, coolant concentration, material batch, vibration, and measurement technique as possible causes. An MSA study found that the roughness instrument and operator method were adequate for screening, but the fixture required standardization.

An Individuals–Moving Range chart showed a step increase after a tool-change procedure was altered. Investigation found that operators were replacing inserts by elapsed time rather than tool condition, and the new procedure did not specify tool seating or coolant verification.

Approach and Solution

The team used a small factorial DOE with feed rate, cutting speed, and coolant concentration, including replication and randomized run order. The analysis indicated that feed rate had the largest main effect and that feed rate interacted with tool condition. A confirmation run used a lower validated feed rate, a standardized insert-change method, and a coolant concentration check.

The process was then monitored with a control chart. After sufficient stable data were collected, capability was estimated. Before improvement, μ = 1.42 μm and σ = 0.18 μm for an upper specification of 1.60 μm. The one-sided capability index was:

After improvement, μ = 1.08 μm and σ = 0.10 μm:

The result indicates substantially improved margin, subject to continued stability and confirmation that the measurement system and sampling method remain valid.

Outcome and Learning Points

The defect rate fell, rework decreased, and the tool-change standard became part of the controlled work instruction. The case illustrates that a Pareto chart identifies priority, an MSA prevents false conclusions, SPC detects process change, DOE reveals factor relationships, and standardization sustains improvement. No single tool would have been sufficient.

10. Implementation, Governance, and Contemporary Practice

Effective quality engineering requires traceable data, defined ownership, competent personnel, and escalation rules. Nonconforming product must be identified and controlled so that it cannot be unintentionally used or shipped. Corrective action should address verified systemic causes rather than merely retraining individuals. Safety, environmental, and regulatory requirements take precedence over production targets.

Digital manufacturing systems increasingly collect machine, sensor, inspection, and maintenance data. Analytics and artificial intelligence can detect patterns, predict tool wear, and support anomaly detection. However, models require representative data, validated measurement systems, protection against data drift, cybersecurity, access control, and human oversight. An algorithmic alarm is not proof of a defect; false positives and false negatives must be evaluated. Data integrity, privacy where personnel data are involved, and explainability are governance responsibilities.

Best practice is to combine prevention and control: robust design, mistake-proofing, capable equipment, standardized work, supplier quality, calibrated measurement, SPC, periodic audits, and documented learning. Quality objectives should be connected to customer risk and process economics rather than pursued as isolated numerical targets.

Chapter Summary

Quality engineering integrates customer requirements, product and process design, measurement, statistical control, and continual improvement. Deming, Juran, and Crosby emphasize system responsibility, fitness for use, prevention, and conformance. TQM creates an organization-wide improvement culture, ISO 9001 structures a documented and auditable management system, and Six Sigma provides DMAIC and data-based reduction of variation.

The seven basic quality tools support data collection, prioritization, process understanding, and investigation. Control charts distinguish common-cause from special-cause variation, while capability indices evaluate whether a stable process can meet specifications. Cp describes potential spread capability; Cpk also considers centering. MSA is essential because unreliable measurements can invalidate every subsequent conclusion. Acceptance sampling makes a lot decision from a sample and explicitly involves producer’s and consumer’s risks; it does not improve the process. DOE efficiently investigates factor effects and interactions and is strengthened by randomization, replication, blocking, and confirmation.

The central professional principle is prevention: understand the process, verify the measurement system, use evidence to identify causes, implement technically sound controls, and sustain gains through standardized work and management review.

Key Terms

Quality: Degree to which characteristics fulfill requirements.

Conformance quality: Consistency of production with specified requirements.

Common cause: Routine source of process variation.

Special cause: Identifiable, non-routine source of variation.

SPC: Statistical monitoring and improvement of process behavior.

Control limit: Statistically calculated boundary for expected process variation.

Specification limit: Requirement boundary set by design, customer, or regulation.

Rational subgroup: Group of observations collected to represent short-term process variation.

Capability: Ability of a stable process to meet specifications.

Cp: Potential two-sided capability based on spread.

Cpk: Capability index accounting for process centering.

MSA: Evaluation of the adequacy of a measurement system.

Repeatability: Variation under repeated measurements by the same operator and instrument.

Reproducibility: Variation associated with different operators or appraisers.

Acceptance sampling: Lot decision based on examination of a sample.

Producer’s risk: Risk of rejecting an acceptable-quality lot.

Consumer’s risk: Risk of accepting a rejectable-quality lot.

TQM: Organization-wide philosophy of customer-focused continual improvement.

Six Sigma: Data-based approach for reducing variation and defects.

Design of experiments: Planned variation of factors to estimate effects and interactions.

Review Questions

Short Answer Questions

1. Distinguish common-cause variation from special-cause variation.
2. Why are specification limits different from control limits?
3. What information does Cpk provide that Cp does not?
4. Define repeatability and reproducibility in measurement-system analysis.
5. State the five stages of DMAIC.

Long Answer Questions

1. Explain how TQM, ISO 9001, and Six Sigma differ and how they can be used together.
2. Describe the seven basic quality tools and identify an appropriate use for each.
3. Explain the construction and interpretation of an X̄–R control-chart system.
4. Discuss the assumptions and limitations of capability analysis.
5. Explain the purpose, risks, and limitations of acceptance sampling.

Analytical / Application Questions

1. A stable process has USL = 25.10 mm, LSL = 24.90 mm, μ = 25.04 mm, and σ = 0.020 mm. Calculate Cp and Cpk and recommend whether centering or variation should be addressed first.
2. A machining process shows a sudden control-chart shift after a maintenance activity. Develop an investigation plan that includes containment, MSA verification, cause analysis, corrective action, and control.
3. Design a DOE strategy to reduce surface roughness using cutting speed, feed, and tool condition. Specify factors, responses, randomization, replication, and how you would confirm the result.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a control limit from a specification limit?

A. A control limit is set by the customer, while a specification limit is calculated from process data.

B. A control limit describes expected process behavior, while a specification limit states a requirement.

C. Both are calculated identically from the process mean and standard deviation.

D. A specification limit is used only for attribute data.

Correct Answer: B. A control limit describes expected process behavior, while a specification limit states a requirement.

Question 2

For the shaft example with Cp = 1.33 and Cpk = 0.80, what is the main issue?

A. The process is centered but excessively variable.

B. The measurement system must automatically be rejected.

C. The process is off-center even though its potential spread is better.

D. The specification limits are control limits.

Correct Answer: C. The process is off-center even though its potential spread is better.

Question 3

Which tool ranks defect categories from largest to smallest and often includes a cumulative percentage line?

A. Pareto chart

B. Scatter diagram

C. Flowchart

D. Histogram

Correct Answer: A. Pareto chart

Question 4

In an acceptance-sampling plan, a lot is accepted when the observed number of nonconforming units is:

A. Greater than the acceptance number

B. Equal to or less than the acceptance number

C. Equal to the sample size

D. Always zero regardless of the plan

Correct Answer: B. Equal to or less than the acceptance number

Question 5

Which DMAIC phase verifies important suspected causes using data?

A. Define

B. Measure

C. Analyze

D. Control

Correct Answer: C. Analyze

Question 6

A measurement study finds large variation when different operators measure the same part. This is primarily a problem of:

A. Repeatability

B. Reproducibility

C. Linearity only

D. Specification width

Correct Answer: B. Reproducibility

Question 7

For a stable process with USL = 10.10, LSL = 9.90, and σ = 0.020, what is Cp?

A. 0.67

B. 1.00

C. 1.33

D. 1.67

Correct Answer: C. 1.33

Question 8

What is the principal purpose of randomization in a design of experiments?

A. To guarantee a normal response distribution

B. To reduce bias from time-related or uncontrolled effects

C. To eliminate the need for replication

D. To make specification limits wider

Correct Answer: B. To reduce bias from time-related or uncontrolled effects

Question 9

A point beyond a control limit most directly indicates:

A. A guaranteed defective product

B. A likely special cause requiring investigation

C. That the specification limits are incorrect

D. That capability is automatically Cp = 1.33

Correct Answer: B. A likely special cause requiring investigation

Question 10

Which statement about ISO 9001 certification is correct?

A. It guarantees every product is defect-free.

B. It certifies that each operator has identical skill.

C. It indicates that the quality-management system has been audited against specified requirements.

D. It replaces the need for process control.

Correct Answer: C. It indicates that the quality-management system has been audited against specified requirements.