Free CQE Exam Flashcards

Juran defined quality as fitness for use, judged by the customer's actual application. Crosby defined it as conformance to requirements, with zero defects as the performance standard. Deming focused on the system, arguing that management, not the worker, owns most of the variation.

Exactly one person is Accountable: the single owner answerable for the outcome. Responsible names whoever performs the work, and more than one person may be Responsible. Consulted parties give two-way input before the decision. Informed parties get one-way updates after it.

Holding paramount the safety, health, and welfare of individuals, the public, and the environment. The code groups expectations under acting with integrity and honesty, demonstrating responsibility, respect, and fairness, and safeguarding proprietary information while avoiding conflicts of interest. Disclose a conflict rather than judging it harmless yourself.

Structure that neutralizes dominant voices. Members generate ideas silently in writing, share them round-robin one at a time, clarify without debate, then rank or vote to produce a prioritized list. Brainstorming only generates ideas openly and produces no ranking.

Internal compares units inside your own organization. Competitive compares you with a direct rival in your market. Functional compares the same function across a different industry. Generic compares a broadly similar process anywhere. Competitive benchmarking is the direct-rival comparison.

Prevention, appraisal, internal failure, and external failure. Prevention plus appraisal is the cost of conformance, money spent to get quality right. Internal plus external failure is the cost of nonconformance. External failure is normally the costliest per incident, because the customer found the defect.

Level 1 is the quality manual stating policy and scope. Level 2 is procedures saying who does what and when. Level 3 is work instructions saying how a task is performed. Level 4 is records and completed forms, the objective evidence that the system actually ran.

First-party is internal: the organization audits itself. Second-party is a customer auditing its own supplier, or an agent doing so on the customer's behalf. Third-party is an independent body such as a registrar or regulator, and only a third-party audit can lead to certification.

A product audit examines a finished item or service against its specification. A process audit examines whether one specific process is executed as documented and is capable. A system audit examines the entire quality management system against a standard such as ISO 9001.

ISO 9001 is a requirements standard: you either conform to its clauses or you do not, and a registrar can certify you. Baldrige is a non-prescriptive performance-excellence framework, scored across seven categories for an award and for self-assessment. Baldrige confers no certification.

Level 1 reaction, did participants like it. Level 2 learning, did they gain the knowledge. Level 3 behavior, are they applying it on the job. Level 4 results, did business measures improve. Only levels 3 and 4 show that the training changed the organization.

Verification asks whether you built the product right, confirming that output meets the specified requirements. Validation asks whether you built the right product, confirming it meets user needs in the intended operating environment. Verification checks against specification; validation checks against actual use.

Installation qualification confirms the equipment was installed and configured as specified. Operational qualification confirms it functions correctly across its intended operating range, including the limits. PQ, which the 2022 BoK labels process qualification and pharma guidance calls performance qualification, confirms it produces conforming output consistently under actual production conditions, with real materials and operators.

A basic dimension is theoretically exact and carries no tolerance of its own. Its tolerance comes from the feature control frame that references it, or from general notes. Contrast a directly toleranced dimension, which carries its own plus-minus limits right on the drawing.

RPN equals severity times occurrence times detection, each rated 1 to 10, so RPN ranges from 1 to 1000. The detection scale is inverted: 10 means current controls will almost certainly miss the failure before it reaches the customer, and 1 means detection is nearly certain.

MTBF applies to repairable items and measures mean operating time between successive failures. MTTF applies to non-repairable items that are discarded on failure. When the failure rate is constant, meaning the exponential model applies, MTBF equals 1 divided by lambda.

Availability equals MTBF divided by the sum of MTBF and MTTR. It rises when you extend mean time between failures through reliability improvement, or shorten mean time to repair through maintainability improvement such as better diagnostics, spares stocking, and physical access.

In series, system reliability is the product of the component reliabilities, so it is always lower than the weakest component. In active parallel, system reliability equals 1 minus the product of the component unreliabilities, that is 1 minus the product of each (1 minus R).

Producer's risk is alpha, the probability of rejecting a lot that is actually good, a Type I error. Consumer's risk is beta, the probability of accepting a lot that is actually bad, a Type II error. The producer loses acceptable product; the consumer receives defective product.

AQL is the worst quality level still treated as acceptable, conventionally placed where the probability of acceptance is about 0.95, so producer's risk is about 0.05. LTPD, also called RQL, is the poor quality level where probability of acceptance is conventionally 0.10, the consumer's risk.

Increasing the sample size steepens the curve, so the plan discriminates more sharply between good and bad lots. Increasing the acceptance number shifts the curve to the right and flattens it, making the plan more lenient. Sample size, not lot size, drives discrimination.

Z1.4 is attributes sampling: you count conforming and nonconforming units. Z1.9 is variables sampling: you measure a continuous characteristic, and it assumes that measurement is normally distributed. Z1.9 reaches equivalent protection with a smaller sample, but it covers only one characteristic per plan.

An unbroken, documented chain of comparisons back to a recognized national or international standard such as NIST, with the measurement uncertainty stated at every link. A calibration sticker alone is not traceability, and the chain is broken if any intermediate standard's uncertainty is undocumented.

Repeatability is equipment variation: one operator remeasuring the same part with the same gage. Reproducibility is appraiser variation between operators. Under 10% GRR is generally acceptable, 10 to 30% may be acceptable depending on the application and cost, and over 30% is unacceptable.

Use as is, rework, repair, scrap, or return to supplier. Rework restores the item to full conformance with the original specification. Repair makes it usable and safe but leaves it non-conforming, so repair needs a documented concession or deviation approval before shipment.

Flowcharts, Pareto charts, cause-and-effect (fishbone) diagrams, control charts, check sheets, scatter diagrams, and histograms. These handle numerical and process data. Pareto separates the vital few causes; the cause-and-effect diagram organizes hypotheses before you have data to test them.

Affinity diagrams with force-field analysis, tree diagrams, process decision program charts, matrix diagrams, interrelationship digraphs, prioritization matrices, and activity network diagrams. These organize verbal, qualitative, and planning information, in contrast to the seven basic tools, which handle numerical data.

PDCA is a lightweight, repeating improvement loop suited to incremental, local changes. DMAIC is a chartered Six Sigma project with a defined baseline, a validated measurement system, statistical root-cause analysis, and a control plan. Use DMAIC when the cause is unknown and the stakes justify the rigor.

Defects, overproduction, waiting, non-utilized talent, transportation, inventory, motion, and extra processing. Overproduction is often called the worst waste because it creates inventory, transportation, and waiting on top of itself. Non-utilized talent is the eighth waste, added to Ohno's original seven.

Takt time equals available production time divided by customer demand for that period. It is the pace the customer requires. Cycle time is how long the process actually takes per unit. When cycle time exceeds takt time, the line cannot meet demand.

OEE equals availability times performance times quality, each expressed as a rate. Because the three multiply, a line running 90% on all three yields only about 73% OEE. Roughly 85% is the commonly cited world-class benchmark for discrete manufacturing.

Correction fixes the specific nonconforming item, such as rework or scrap. Corrective action eliminates the cause of a nonconformity that has already occurred, so it does not recur. Preventive action eliminates the cause of a potential nonconformity, so it never occurs. Only the latter two address causes.

A prevention device makes the error physically impossible, such as an asymmetric fixture that accepts the part only one way. A detection device allows the error but signals it immediately, such as a part-present sensor or a light curtain. Prevention is always preferred.

That the sampling distribution of the sample mean approaches normal as subgroup size grows, even when the population is skewed. The standard error equals sigma divided by the square root of n. This is why X-bar charts tolerate non-normal data but individuals charts do not.

A Type I error rejects a true null hypothesis, and its probability is alpha, the significance level. A Type II error fails to reject a false null hypothesis, and its probability is beta. Power equals 1 minus beta, the chance of detecting a real effect. Larger samples raise power.

Binomial: a fixed number of independent trials at constant probability, counting nonconforming units. Poisson: counts of nonconformities in a constant area or interval with a known mean. Hypergeometric: sampling without replacement from a small finite lot, where the probability changes with every draw.

F equals the mean square between groups divided by the mean square within groups. A significant F says at least one treatment mean differs from the others. It does not say which one; that needs a post-hoc comparison such as Tukey. Non-significance does not prove the means are equal.

The value r runs from minus 1 to plus 1 and gives the direction and strength of a linear association. The value r-squared runs from 0 to 1 and gives the proportion of variation in the response explained by the model. An r of 0.7 explains only about 49% of the variation.

Common cause is the inherent random variation of a stable process. Special cause is assignable variation from something outside that usual system. Rational subgrouping forms subgroups so only common-cause variation appears within a subgroup, which forces special causes to show up between subgroups.

A p chart plots the proportion nonconforming and tolerates varying subgroup size. An np chart plots the number nonconforming and requires a constant subgroup size. A c chart counts nonconformities in a constant-size inspection unit. A u chart plots nonconformities per unit when inspection-unit size varies.

Control limits are calculated from the process itself, normally three standard errors of the plotted statistic, and describe what the process does. Specification limits come from the customer and describe what is wanted. Plotting specs against subgroup means compares customer limits for individual parts to averages, which is invalid.

Cp equals (USL minus LSL) divided by 6 sigma and measures potential capability, ignoring where the process sits. Cpk equals the smaller of (USL minus x-bar) over 3 sigma and (x-bar minus LSL) over 3 sigma, penalizing an off-center mean. Cpk never exceeds Cp, and they match only when perfectly centered.

Cp and Cpk use a within-subgroup, short-term sigma estimated from the control chart, typically R-bar divided by d2. Pp and Ppk use the overall sample standard deviation of all the data, which absorbs between-subgroup drift. A large gap between Cpk and Ppk signals an unstable process.

A two-level full factorial with k factors needs 2 to the k runs and estimates every effect separately. A fractional factorial runs 2 to the (k minus p) and confounds effects with each other. Resolution names the cost: III aliases main effects with two-factor interactions, IV keeps main effects clear of them, and V keeps two-factor interactions clear of each other too.

Inherent risk is the exposure before any control is applied. Residual risk is what remains after the chosen treatment and controls are in place and working. Risk acceptance decisions and risk register sign-offs are made against residual risk, never against inherent risk.

Enterprise risk is strategic and organization-wide, such as regulatory or business-model risk. Operational risk sits in execution, such as supplier, supply chain, project, or quality system risk. Product risk attaches to the item itself, covering design, process, use, and safety risk.

A risk matrix plots severity against likelihood on a grid and assigns each cell an acceptability level, so two dimensions drive the ranking. An FMEA RPN multiplies severity, occurrence, and detection, so it also credits your ability to catch the failure. A risk matrix ignores detectability.

Avoid, by not doing the activity at all. Mitigate, by reducing severity or likelihood with controls. Transfer or share, by moving the financial exposure through insurance or contract terms. Accept, by consciously retaining the residual risk. Transfer moves the cost, not the underlying hazard.

The risk description and category, the owner, severity and likelihood ratings, existing controls, the resulting risk level, the planned treatment with a due date, the residual risk after treatment, and current status. Without a named owner and a date, a register records risk but does not manage it.

Complaint tracking, trending of field and warranty data, and post-market surveillance. All three test whether the assumed occurrence and severity ratings still hold in real use. A rising complaint trend is the signal to reopen the FMEA and revise the risk assessment.