Quality Management and Statistical Quality Control

Introduction and context

Quality management is the coordinated set of activities used to ensure that products, services, and processes satisfy stated, implied, and statutory requirements. In mechanical engineering, quality is not limited to final inspection. It begins with understanding customer requirements, continues through product and process design, and extends to supplier control, production, maintenance, delivery, and field performance.

A modern quality system therefore treats quality as a property designed into a process rather than something inspected into a product. This perspective is important because inspection can detect defects after resources have already been consumed, whereas process control can prevent their occurrence.

Learning objectives

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

1. Explain major dimensions and definitions of quality.
2. Classify and calculate the cost of quality.
3. Describe Total Quality Management (TQM), ISO 9001, and continuous improvement.
4. Apply the seven basic quality tools and perform root-cause analysis.
5. Construct and interpret statistical process-control charts.
6. Calculate and interpret process capability indices.
7. Explain acceptance sampling and design of experiments.
8. Apply Six Sigma and the DMAIC framework.
9. Evaluate quality decisions in technical, ethical, and organizational contexts.

1. The meaning and dimensions of quality

The traditional manufacturing view defines quality as conformance to specifications. This remains essential, but it is incomplete. A product may conform dimensionally and still fail because it is unreliable, difficult to maintain, unsafe, or unsuitable for the user.

A useful multidimensional model includes:

Quality requirements may be explicit, such as a dimensional tolerance, or implicit, such as reasonable noise, safe operation, and legal compliance. Engineers must translate the voice of the customer into measurable critical-to-quality characteristics (CTQs).

Quality should also be distinguished from grade. Grade describes a category or set of intended features; quality describes how well the delivered item satisfies its requirements. A low-grade product can have high quality if it consistently meets its intended requirements.

2. Cost of quality

The cost of quality (COQ) is the total cost associated with achieving quality and with failing to achieve it. It is commonly divided into prevention, appraisal, internal failure, and external failure costs.

Here, C_P is prevention cost, C_A is appraisal cost, C_IF is internal-failure cost, and C_EF is external-failure cost. The categories should be defined consistently within an organization; otherwise comparisons are misleading.

Prevention costs include training, quality planning, process capability studies, mistake-proofing, supplier development, and preventive maintenance. Appraisal costs include inspection, testing, calibration, audits, and measurement-system evaluation. Internal-failure costs arise before delivery and include scrap, rework, retesting, downtime, and corrective action. External-failure costs arise after delivery and include warranty claims, returns, recalls, field repair, liability, and loss of reputation.

Worked example: cost of quality

A plant reports the following annual costs: prevention $80,000, appraisal $120,000, scrap and rework $210,000, and warranty claims $90,000. Therefore:

The failure component is:

Thus, 60% of the recorded COQ is failure cost. A rational improvement program might increase prevention expenditure if the expected reduction in failure cost exceeds the additional prevention cost. COQ is not merely an accounting report; it is a basis for prioritizing improvement.

3. Total Quality Management and quality systems

Total Quality Management is an organization-wide management philosophy based on customer focus, process thinking, employee involvement, fact-based decisions, leadership commitment, and continual improvement. TQM does not mean that every employee independently performs inspection. It means that responsibility for quality is distributed throughout the value-creation system.

Important principles include:

• Customer focus: Requirements and user outcomes guide design and improvement.
• Leadership: Managers establish direction, resources, and a culture in which quality can be reported without fear.
• Engagement of people: Operators, technicians, engineers, suppliers, and customers contribute knowledge.
• Process approach: Inputs, activities, controls, outputs, and feedback are analyzed as a system.
• Evidence-based decision-making: Data are interpreted with attention to measurement error and variation.
• Relationship management: Supplier capability and process interfaces affect final quality.
• Continual improvement: Improvement is a normal operating activity, not a crisis response.

TQM is often associated with Deming's system view, Juran's quality trilogy of planning, control, and improvement, and Crosby's emphasis on conformance and prevention. These approaches differ in language but share the idea that management systems strongly influence quality outcomes.

ISO 9001

ISO 9001 specifies requirements for a quality-management system (QMS). It does not prescribe a particular product design or guarantee that every product is excellent. Instead, it requires an organization to establish controlled processes capable of consistently meeting customer and applicable statutory requirements.

A practical ISO 9001 system generally addresses organizational context, leadership, planning, support, operation, performance evaluation, and improvement. Important evidence includes controlled documented information, competence records, calibration control, supplier evaluation, internal audits, nonconformity control, corrective action, and management review.

Certification is not a substitute for engineering judgment. A company may possess a certified QMS and still produce poor designs if customer requirements are misunderstood or technical controls are inadequate. Effective implementation links the QMS to risk analysis, process capability, maintenance, design reviews, and measurable objectives.

4. Variation and the seven basic quality tools

No industrial process produces exactly identical outputs. Variation may be common-cause, arising from the normal system, or special-cause, arising from an assignable disturbance such as a broken tool, incorrect material, or operator error. The distinction matters: common-cause variation usually requires system-level redesign, whereas special causes require investigation and removal.

The seven basic quality tools are practical methods for collecting, displaying, and analyzing process information.

1. Check sheet: A structured form for recording events, locations, times, or defect types. It converts observations into organized data.
2. Histogram: A graphical display of the frequency distribution of numerical data. It helps reveal center, spread, skewness, and multiple populations.
3. Pareto chart: A descending bar chart, usually accompanied by a cumulative-percentage line. It supports prioritization of the relatively few categories contributing most to a problem.
4. Cause-and-effect diagram: Also called an Ishikawa or fishbone diagram. It organizes possible causes under headings such as man/machine/material/method/measurement/environment, though categories should be adapted to the process.
5. Scatter diagram: A plot of paired observations used to explore association between variables. Association does not prove causation.
6. Control chart: A time-ordered display with a center line and statistically derived control limits. It distinguishes routine variation from evidence of a process change.
7. Flowchart or process map: A visual representation of process steps, decisions, inputs, outputs, and interfaces. It reveals delays, loops, handoffs, and opportunities for error.

A run chart is also widely used. It displays observations over time but normally lacks statistically calculated control limits. A control chart therefore provides stronger evidence for distinguishing signal from background variation.

Root-cause analysis

Root-cause analysis (RCA) seeks to identify causes whose removal prevents recurrence. A useful RCA sequence is:

1. Define the problem in observable, measurable terms.
2. Contain the immediate risk without assuming containment is the permanent solution.
3. Map the process and collect evidence.
4. Generate possible causes using a fishbone diagram, 5 Whys, or fault-tree analysis.
5. Test causes against data and physical knowledge.
6. Implement corrective action and verify effectiveness.
7. Standardize the improvement and monitor recurrence.

The “5 Whys” is not a requirement to ask exactly five questions. It is a disciplined search from symptom to mechanism. Statements such as “operator error” are generally inadequate stopping points unless the system explains why the error was possible and why controls did not detect it.

5. Statistical process control

Statistical process control (SPC) uses statistical methods to monitor and improve process stability. Control limits are not specification limits. Specification limits come from design, customer, or regulatory requirements; control limits are estimated from process behavior.

For a variable measured as individual observations, a basic Shewhart chart has:

Here, μ is the process mean and σ is the standard deviation. In practice, these parameters are estimated from rational subgroups or historical data. The ±3σ convention is effective for many approximately stable processes, but it is not a universal law and should not replace domain knowledge.

For subgroup means of size n, if the process standard deviation is known, the standard deviation of the mean is σ/√n, so:

An x̄ chart monitors process location. An R chart or s chart monitors within-subgroup variation. Control-chart constants such as A₂, D₃, and D₄ are commonly used when σ is estimated from subgroup ranges:

The constants depend on subgroup size and assume an appropriate subgrouping model. They should be obtained from a reliable statistical reference or software implementation.

Signals of potential special causes include a point outside a control limit, a long run on one side of the center line, a sustained trend, cyclic behavior, or unusual clustering. Rules should be defined before examining results to reduce selective interpretation. An out-of-control signal requires investigation; automatically deleting the point is unacceptable unless a documented measurement or recording error is established.

Measurement-system considerations

SPC is only as credible as the measurement system. Resolution, bias, linearity, stability, repeatability, and reproducibility should be evaluated. A gauge repeatability and reproducibility (Gage R&R) study separates variation due to the instrument and appraiser from part-to-part variation. If measurement error is large relative to process variation or tolerance, process conclusions may be false.

6. Process capability

Capability compares the natural spread and centering of a stable process with specification limits. Capability indices should not be interpreted before stability has been demonstrated.

For lower specification limit LSL, upper specification limit USL, estimated within-process standard deviation σ̂, and process mean μ̂, the potential capability index is:

C_p measures potential capability if the process is centered. The centered capability index is:

These indices are dimensionless. A larger value indicates more specification width relative to estimated process spread. C_pk cannot exceed C_p for the same data because off-center operation reduces effective capability.

Worked example: capability

Suppose a stable shaft-diameter process has USL = 20.10 mm, LSL = 19.90 mm, estimated mean μ̂ = 20.04 mm, and σ̂ = 0.025 mm. Then:

The two one-sided indices are:

Therefore, C_pk = 0.80. The process has adequate potential spread relative to the tolerance but is poorly centered toward the upper specification. Centering alone may improve C_pk, but the organization must also verify stability, distributional assumptions, and the source of the mean shift.

Capability indices are not universal pass/fail criteria. Common targets such as 1.33 are organizational conventions, not physical laws. Non-normal data, autocorrelation, mixed populations, and unstable processes require alternative analysis or transformation.

7. Acceptance sampling

Acceptance sampling determines whether to accept or reject a lot from a sample rather than inspecting every item. It is useful when 100% inspection is destructive, expensive, slow, or less reliable than a planned sample. It does not control the production process and should not replace process improvement.

A sampling plan is defined by sample size n and acceptance number c for an attributes plan. If a sample contains no more than c defectives, the lot is accepted; otherwise it is rejected. The operating characteristic (OC) curve shows the probability of accepting lots with different quality levels.

For a finite lot of N items containing D defectives, sampling n items without replacement gives the hypergeometric probability of observing d defectives:

where C(a,b) is the binomial coefficient. If sampling is treated as independent because the lot is very large relative to the sample, the binomial approximation may be used with defective fraction p:

A sound plan specifies producer's risk (rejecting a good lot) and consumer's risk (accepting a poor lot), along with acceptable quality levels and limiting quality levels. These risks must be communicated rather than hidden behind a single sample result.

8. Design of experiments

Design of experiments (DOE) is a structured method for changing input factors and observing their effects on a response. Unlike one-factor-at-a-time experimentation, factorial designs can reveal interactions.

For two factors, A and B, each at two levels, a full factorial design has four treatment combinations: low-low, high-low, low-high, and high-high. An interaction exists when the effect of one factor depends on the level of the other. Replication estimates experimental error, randomization reduces the influence of unknown time-related effects, and blocking handles known nuisance variation.

A first-order model for two factors may be written:

Here, y is the response, x_A and x_B are coded factor levels, β terms are model coefficients, β_AB represents interaction, and ε is random error. The model is useful only when its assumptions and residual behavior are checked.

DOE applications include optimizing cutting speed, feed, depth of cut, heat-treatment conditions, welding parameters, cooling flow, and assembly torque. Practical experimentation must also consider safety, equipment limits, confounding, repeatability, and the difference between statistical significance and engineering significance.

9. Six Sigma and DMAIC

Six Sigma is a data-driven approach for reducing process variation and defects while improving customer-relevant performance. In practice, the term refers both to a performance aspiration and to a structured improvement methodology. A “sigma level” should not be quoted without stating the defect definition, opportunity definition, sampling period, and convention used for long-term shifts.

DMAIC consists of:

• Define: Establish the problem, customers, CTQs, scope, business case, and project charter.
• Measure: Define operational measures, validate the measurement system, and establish baseline performance.
• Analyze: Identify and verify relationships between inputs and outputs using process knowledge and data.
• Improve: Develop, test, optimize, and implement solutions, often using DOE or mistake-proofing.
• Control: Standardize the improved process, monitor performance, document reaction plans, and transfer ownership.

For an attributes metric, defects per million opportunities (DPMO) is:

A defect is a failure to meet a requirement, whereas a defective unit may contain one or more defects. Confusing these terms produces invalid performance comparisons.

Lean Six Sigma combines waste reduction with variation reduction. Lean tools such as value-stream mapping, 5S, standardized work, and quick changeover complement, but do not replace, statistical analysis.

10. Applied case: illustrative precision-pump seal problem

Background

A manufacturer of industrial pumps observes increasing warranty claims for leakage at a rotating-shaft seal. The design specification requires leakage below 2 mL/h after a defined endurance test.

Problem

Warranty data show 8.5% of tested units exceeding the limit. Initial inspection finds no single obvious defect. Management proposes 100% final inspection, but this would increase test time and would not explain the cause.

Analysis

A process map identifies seal molding, material storage, shaft finishing, assembly, and endurance testing. A check sheet and Pareto chart show that most failures occur on one molding press and during humid weeks. A fishbone analysis identifies possible causes involving material moisture, mold temperature, shaft roughness, assembly force, and measurement variation.

A Gage R&R study shows that the leakage test is sufficiently repeatable for screening, although its fixture requires improved sealing. A control chart for mold temperature reveals intermittent excursions associated with a faulty thermocouple. Subsequent experiments use mold temperature and material-drying time as factors, with leakage as the response. The experiment shows a strong temperature effect and an interaction: longer drying time helps substantially only at the lower temperature setting.

Approach and solution

The plant replaces the thermocouple, adds automatic temperature alarms, establishes a validated material-drying window, and introduces a poka-yoke preventing assembly before the drying record is complete. The process is then monitored with control charts, while periodic capability studies verify performance against the leakage specification.

Outcome

In this illustrative case, the defect rate falls from 8.5% to 1.1% during the verification period. The improvement is not attributed solely to inspection; it results from measurement validation, verified causal analysis, equipment control, standardized work, and ongoing monitoring.

Learning points

The case demonstrates that customer complaints often require system analysis rather than a search for individual blame. It also shows why capability analysis should follow process stabilization and why corrective action must address mechanisms, not merely symptoms.

11. Challenges, ethics, technology, and best practice

Quality data can be manipulated through selective sampling, deletion of inconvenient observations, changing definitions, or reporting only averages. Such practices are technically dishonest and can create safety risks. Engineers have a duty to report uncertainty, preserve traceability, protect users, and escalate serious nonconformities.

Regulatory and industry requirements may apply to pressure equipment, aerospace, automotive, medical devices, or safety-critical machinery. The applicable standard, legal requirement, and customer contract must be identified before accepting a product or process. Records must be secure, attributable, retrievable, and protected from unauthorized alteration.

Digital manufacturing enables automated data collection, machine-vision inspection, sensor-based SPC, traceability, and predictive maintenance. Artificial intelligence can help classify defects or detect unusual patterns, but it does not remove the need for representative data, validated measurement systems, explainability where decisions affect safety, cybersecurity, human review, and monitoring for model drift. An algorithm trained on biased or incomplete defect data may systematically miss important failures.

Best practice is to:

• translate customer needs into measurable CTQs;
• control the measurement system before trusting process data;
• use prevention and mistake-proofing before relying on inspection;
• distinguish specification limits from control limits;
• investigate special causes without tampering with data;
• verify root causes experimentally or through strong evidence;
• connect corrective action to process ownership;
• use capability indices only for stable, appropriately modeled processes;
• design sampling plans around stated risks; and
• sustain gains through standard work, audits, feedback, and management review.

Chapter Summary

Quality is multidimensional and includes performance, reliability, conformance, durability, serviceability, safety, and customer perception. Effective quality management begins with prevention and process design rather than depending exclusively on final inspection. The cost of quality combines prevention, appraisal, internal-failure, and external-failure costs.

TQM provides an organization-wide philosophy based on customer focus, leadership, employee involvement, process thinking, evidence, and continual improvement. ISO 9001 provides requirements for a quality-management system but does not guarantee technical excellence by itself.

The seven basic quality tools help engineers collect data, prioritize problems, explore causes, understand distributions, and visualize processes. SPC distinguishes common-cause from special-cause variation, while capability indices compare stable process performance with specification limits. Acceptance sampling provides a lot decision with stated risks but does not replace process control. DOE evaluates factor effects and interactions efficiently. Six Sigma and DMAIC provide a structured approach from problem definition through control and sustainment.

Professional quality practice requires statistical competence, measurement discipline, ethical reporting, regulatory awareness, and respect for the people affected by engineering decisions.

Key Terms

Quality: Degree to which characteristics satisfy requirements.

Conformance: Agreement of a product or process with specified requirements.

Critical-to-quality characteristic: Measurable characteristic important to customer or process performance.

Cost of quality: Costs of achieving quality plus costs caused by poor quality.

Prevention cost: Expenditure intended to prevent defects.

Appraisal cost: Cost of measuring, inspecting, or auditing quality.

Internal failure: Nonconformity discovered before delivery.

External failure: Nonconformity discovered after delivery.

Total Quality Management: Organization-wide philosophy of customer-focused continual improvement.

Quality-management system: Coordinated processes used to direct and control quality.

Common cause: Routine source of process variation.

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

Statistical process control: Use of statistical methods to monitor and improve processes.

Control limit: Statistically calculated boundary indicating expected process behavior.

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

Process capability: Relationship between stable process variation and specification limits.

Acceptance sampling: Decision about a lot based on a sample.

Design of experiments: Planned investigation of factor effects on a response.

Root-cause analysis: Systematic search for causes whose removal prevents recurrence.

DMAIC: Define, Measure, Analyze, Improve, and Control improvement cycle.

Review Questions

Short Answer Questions

1. Distinguish quality from grade.
2. Name and explain the four categories of cost of quality.
3. What is the difference between a control limit and a specification limit?
4. Why must process stability be examined before capability is interpreted?
5. State the five phases of DMAIC.

Long Answer Questions

1. Explain the dimensions of quality using examples from mechanical engineering.
2. Discuss the principles of TQM and the role of ISO 9001 in a quality-management system.
3. Describe the seven basic quality tools and explain how they can be combined in an improvement project.
4. Explain the construction, interpretation, and limitations of Shewhart control charts.
5. Compare acceptance sampling, SPC, capability analysis, and DOE as quality-management methods.

Analytical / Application Questions

1. A stable process has USL = 50.20 mm, LSL = 49.80 mm, μ̂ = 50.12 mm, and σ̂ = 0.05 mm. Calculate C_p and C_pk, then identify the primary improvement priority.
2. A Pareto chart shows that 62% of pump failures are associated with seal leakage, 18% with bearings, 12% with electrical faults, and 8% with other causes. Propose an evidence-based improvement sequence and explain why Pareto concentration alone does not prove root cause.
3. Design a small DOE for investigating cutting speed and feed rate on surface roughness. Specify the factors, levels, response, randomization or blocking considerations, and how you would use the results in a control plan.

Multiple Choice Questions with Answers

Question 1

Which cost category includes warranty claims discovered after delivery?

A. Prevention cost

B. Appraisal cost

C. Internal-failure cost

D. External-failure cost

Correct Answer: D. External-failure cost

Question 2

What is the principal purpose of a control chart?

A. To establish customer specification limits

B. To distinguish routine variation from possible special causes

C. To replace measurement-system validation

D. To determine product selling price

Correct Answer: B. To distinguish routine variation from possible special causes

Question 3

For the same stable data, which statement is generally correct?

A. Cpk is greater than Cp when the process is off-center

B. Cpk cannot exceed Cp

C. Cp measures only process centering

D. Cpk is calculated without specification limits

Correct Answer: B. Cpk cannot exceed Cp

Question 4

Which tool ranks defect categories from most frequent to least frequent?

A. Pareto chart

B. Scatter diagram

C. Flowchart

D. Histogram

Correct Answer: A. Pareto chart

Question 5

In DMAIC, measurement-system validation primarily belongs to which phase?

A. Define

B. Measure

C. Improve

D. Control

Correct Answer: B. Measure

Question 6

A process has USL = 10.10, LSL = 9.90, and σ̂ = 0.025. What is Cp?

A. 0.67

B. 1.00

C. 1.33

D. 2.67

Correct Answer: C. 1.33

Question 7

Which statement about acceptance sampling is correct?

A. It guarantees that every accepted unit is defect-free

B. It replaces process improvement

C. It makes a lot decision using sample evidence and stated risks

D. It always has lower total cost than inspection

Correct Answer: C. It makes a lot decision using sample evidence and stated risks

Question 8

What does an interaction in DOE mean?

A. A factor has no effect on the response

B. The effect of one factor depends on the level of another factor

C. The response has no measurement error

D. All experimental runs produce identical results

Correct Answer: B. The effect of one factor depends on the level of another factor

Question 9

Which action is most consistent with root-cause analysis?

A. Blaming the last operator involved

B. Deleting an unusual data point without investigation

C. Testing proposed causes against process evidence

D. Increasing final inspection without examining the process

Correct Answer: C. Testing proposed causes against process evidence

Question 10

Which statement best describes ISO 9001?

A. It prescribes one exact product design

B. It is a quality-management-system requirements standard

C. It guarantees zero defects

D. It eliminates the need for engineering judgment

Correct Answer: B. It is a quality-management-system requirements standard