CQE cheat sheet

Cost of quality = prevention + appraisal + internal failure + external failure; prevention and appraisal are conformance costs, failures are nonconformance costs.

Total quality cost is minimized where spending on prevention and appraisal balances failure costs; driving either category alone toward zero raises total cost.

Juran Trilogy: quality planning, quality control, quality improvement; Juran also popularized applying Pareto's 'vital few, trivial many' rule to quality problems.

Crosby: 'quality is free,' zero defects, cost of nonconformance; Deming: 14 points, PDSA cycle, drive out fear, eliminate dependence on mass inspection.

Herzberg two-factor theory: hygiene factors (pay, policy, working conditions) prevent dissatisfaction but do not motivate; motivators (achievement, recognition, responsibility, growth) create satisfaction.

McGregor Theory X assumes workers dislike work and require close control; Theory Y assumes workers seek responsibility and self-direction.

Tuckman team stages: forming, storming, norming, performing, adjourning; teams commonly regress to earlier stages when membership or goals change.

Lewin change model: unfreezing, changing, refreezing; force-field analysis holds that change occurs only when driving forces exceed restraining forces.

Hersey-Blanchard situational leadership matches style to follower readiness: telling (low readiness), selling, participating, delegating (high readiness).

ISO 9001:2015 has ten clauses: scope; normative references; terms and definitions; context of the organization; leadership; planning; support; operation; performance evaluation; improvement.

Seven QMS principles: customer focus, leadership, engagement of people, process approach, improvement, evidence-based decision making, relationship management.

ISO 9001:2015 mandates risk-based thinking (clause 6.1) but requires no formal risk register, no named management representative, and no mandatory quality manual.

Audit types: first party = internal audit; second party = customer or purchaser auditing a supplier; third party = independent certification body.

Certification cycle: stage 1 readiness and documentation review, stage 2 implementation audit, annual surveillance audits, full recertification every three years.

Documented information hierarchy: quality policy/manual, system procedures, work instructions, records and objective evidence.

Sector standards: AS9100 (aerospace), IATF 16949 (automotive), ISO 13485 (medical devices), TL 9000 (telecommunications), ISO 14001 (environmental).

Baldrige Criteria: seven categories totaling 1000 points with the largest share in results; ISO 9001 is a conformance standard, Baldrige an excellence framework.

Nonconformity grading: major = system breakdown or missing required element; minor = isolated lapse; observation/opportunity for improvement is not a nonconformity.

Design control sequence: design inputs, design activities, design outputs, verification, validation, design transfer; formal design reviews are held at each phase gate.

Verification confirms that outputs meet inputs (built right); validation confirms that the product meets intended use and user needs (built the right thing).

QFD house of quality translates customer requirements (whats) into technical requirements (hows) via a relationship matrix, then cascades to parts, process, and production planning.

FMEA risk priority number = Severity x Occurrence x Detection; DFMEA addresses design failure modes, PFMEA addresses process failure modes.

Series reliability Rs = R1 x R2 x ... x Rn; parallel (redundant) reliability Rp = 1 - (1-R1)(1-R2)...(1-Rn).

Exponential model R(t) = e^(-lambda t) for constant failure rate lambda; MTBF = 1/lambda; reliability at the mean time is e^-1, about 0.368.

Bathtub curve phases: infant mortality (decreasing failure rate, screened by burn-in), useful life (constant rate), wear-out (increasing rate).

Tolerance stack-up: worst case sums tolerances arithmetically; statistical root-sum-square = sqrt(t1^2 + t2^2 + ...) applies only to independent, centered dimensions.

Robust design (Taguchi): parameter design selects levels least sensitive to noise, tolerance design tightens only where loss is greatest; loss L = k(y - T)^2.

Chart selection: X-bar and R for subgroups n = 2-10, X-bar and s for n > 10, individuals and moving range for n = 1; attributes use p (variable n), np (constant n), c and u.

X-bar chart limits = grand mean +/- A2 times average range; R chart UCL = D4 times average range and LCL = D3 times average range; limits derive from process variation, not from specifications.

Attribute limits: p-bar +/- 3 sqrt(p-bar(1-p-bar)/n); c-bar +/- 3 sqrt(c-bar); u-bar +/- 3 sqrt(u-bar/n).

Out-of-control signals: one point beyond 3 sigma; two of three beyond 2 sigma on the same side; four of five beyond 1 sigma on the same side; eight consecutive points on one side; trends and cycles.

Cp = (USL - LSL)/6 sigma measures potential capability; Cpk = min[(USL - mean)/3s, (mean - LSL)/3s] measures actual capability and penalizes off-centering.

Pp and Ppk use the overall long-term standard deviation; Cp and Cpk are valid only for a stable, in-control, approximately normal process.

Capability benchmarks: Cpk at least 1.33 for established processes, 1.67 for new or critical characteristics, 2.00 for safety-critical; six sigma is 3.4 ppm with a 1.5 sigma shift.

Gage R&R percent GRR = 100 x GRR/total variation; below 10% acceptable, 10-30% marginal, above 30% unacceptable; number of distinct categories should be at least 5.

Acceptance sampling: AQL is the quality level with high acceptance probability (producer's risk alpha); LTPD/RQL has low acceptance probability (consumer's risk beta); the OC curve plots probability of acceptance against lot fraction defective.

PDCA/PDSA: Plan (define problem, target, root cause), Do (pilot the change), Check/Study (verify results against target), Act (standardize or adjust).

DMAIC (define, measure, analyze, improve, control) improves existing processes; DMADV (define, measure, analyze, design, verify) creates new designs under DFSS.

Eight wastes: transportation, inventory, motion, waiting, overproduction, overprocessing, defects, unused talent (TIMWOOD plus skills).

5S: seiri (sort), seiton (set in order), seiso (shine), seiketsu (standardize), shitsuke (sustain).

Kaizen is small, continuous, incremental improvement by the people doing the work; a kaizen blitz is a concentrated three-to-five-day event targeting one process.

Root cause tools: 5 Whys, cause-and-effect (fishbone) with 6M categories (man, machine, material, method, measurement, environment), Pareto prioritization, fault tree analysis.

Overall equipment effectiveness = availability x performance x quality; world-class benchmark is approximately 85%.

Theory of constraints: identify the constraint, exploit it, subordinate everything else to it, elevate it, then repeat; prevent inertia from becoming the new constraint.

Poka-yoke prevents or detects errors at the source; SMED reduces changeover to single-digit minutes; Taguchi loss L = k(y - T)^2 rises with the square of deviation from target.

Descriptive statistics: mean = sum x/n; sample variance s^2 = sum(x - x-bar)^2/(n - 1); coefficient of variation CV = s/x-bar, expressed as a percentage for relative variability.

Normal distribution: about 68.27% of values lie within +/-1 sigma, 95.45% within +/-2 sigma, 99.73% within +/-3 sigma; standard score z = (x - mu)/sigma.

Central limit theorem: for n of roughly 30 or more the sampling distribution of the mean is near-normal with standard error sigma/sqrt(n), regardless of the population shape.

Confidence interval for a mean: x-bar +/- z(alpha/2) times sigma/sqrt(n); substitute t for z when sigma is unknown and the sample is small.

Hypothesis testing: reject H0 when the p-value is less than alpha or the test statistic exceeds the critical value; alpha = P(Type I error), beta = P(Type II error), power = 1 - beta.

ANOVA tests equality of k means using F = mean square between groups / mean square within groups; assumptions are independent samples, normality, and equal variances.

Discrete probability: binomial P(x) = C(n,x) p^x (1-p)^(n-x); Poisson P(x) = e^(-lambda) lambda^x / x!; independent events multiply, P(A and B) = P(A) P(B).

Design of experiments: a full 2^k factorial estimates all main effects and interactions in 2^k runs; center points detect curvature; randomization, replication, and blocking control nuisance variation.

Regression: least-squares line y = a + bx; r^2 is the proportion of variation in y explained by x; correlation r ranges from -1 to 1 and does not establish causation.

Risk is characterized by likelihood (probability) times consequence (severity); scores are plotted on a likelihood-impact matrix, commonly 5x5, to rank and prioritize risks.

ISO 31000 process: establish context, risk identification, risk analysis, risk evaluation, risk treatment, with communication and consultation plus monitoring and review throughout.

Risk treatment hierarchy: avoid the activity, reduce likelihood, reduce impact, transfer or share (insurance, contract), or accept and monitor; ALARP means reduce to as low as reasonably practicable.

FMEA risk priority number = Severity x Occurrence x Detection; severity of 9 or 10 requires design or process change regardless of RPN; actions should attack occurrence before detection.

Fault tree analysis is deductive and top-down using AND/OR gates to find minimal cut sets; event tree analysis is inductive and forward-propagating from an initiating event.

HAZOP applies guide words (no, more, less, as well as, part of, reverse, other than) to process parameters at each node to identify deviations and their consequences.

A risk register records risk description, cause, potential effect, owner, likelihood, impact, score, planned response, and review date.

ISO 9001:2015 clause 6.1 requires determining risks and opportunities and planning actions to address them, but does not mandate a documented risk register or a formal FMEA.

Poka-yoke and other error-proofing reduce the occurrence of a failure mode; added inspection reduces the detection score without reducing the underlying risk.