SPC 术语表：中文词条、英文原名与操作性定义

Margin of a stable process with an accepted distribution model relative to specifications. In the dashboard, it determines when C indices are appropriate.

Observed behavior relative to specifications when stability has not yet been demonstrated. In that case, the dashboard retains P nomenclature.

Compares tolerance width with six standard deviations of total variation. Does not consider whether the mean is centered.

Compares the distance to the nearest specification limit with three standard deviations of total variation. Penalizes off-centering.

Width index calculated from short-term dispersion within rational subgroups. It is diagnostic and does not replace the index based on total variation.

Centering index based on dispersion within rational subgroups. A large difference from the total index can reveal variation between cavities, batches, or time.

Pp summarizes width and Ppk incorporates centering using total variation. The dashboard displays them when a process study does not justify asserting C capability.

Machine study indices; Pm measures width and Pmk the least favorable side. Evaluated against specific machine study targets.

Method expressing width and location via the central distribution interval. Under the normal model, the dashboard uses standard s-based formulas.

Method transforming out-of-specification fractions into standard normal distances. Can differ from .G when the distribution is non-normal.

Dispersion estimated within subgroups collected under comparable conditions. Feeds Cw, Cwk, and subgroup control chart limits.

Dispersion of all observations combined; includes changes within and between subgroups. Serves as the basis for general indices shown by the calculator.

Permissible interval between LSL and USL. Must not be confused with control limits, which are estimated from the process.

Lower acceptable limit defined by product or process requirements; not calculated from data.

Upper acceptable limit defined by product or process requirements; not calculated from data.

Minimum value against which applicable indices are compared. Depends on study type, class, and agreed conditions.

Target from the primary table when the study reaches the corresponding base sample size.

Corrected target for a reduced sample size and specific confidence level, only where applicable tables provide for it.

Confidence degree chosen to adjust the target for a reduced sample. A higher level produces a more demanding adjusted target.

Uncertainty range of the estimated index. The lower limit confirms whether the result still meets the target when accounting for sampling error.

Study with fewer observations than the base sample. The dashboard applies an adjusted target only for sizes and levels covered by its tables.

Observations collected under comparable conditions to estimate short-term variation without mixing shifts, cavities, batches, or time.

Number of observations forming each subgroup. Determines which variation statistic and chart limits are applicable.

Total number of observations in the study. Affects result eligibility, uncertainty, and applicable target.

Sufficient coverage of actual sources of variation, such as time, batches, shifts, and tooling. A large N alone does not guarantee it.

Measurements arriving one by one that do not form rational subgroups. In the dashboard, they retain the I-MR pathway.

Continuous numerical measurements, such as diameter, weight, or temperature. Enable study of location and dispersion.

Counts or classifications, such as nonconforming units or defects. Require p, np, c, or u charts depending on the denominator.

Smallest increment distinguished by the gauge or appearing in data. Coarse resolution can conceal variation and distort diagnostics.

Confirmation that the measurement system provides sufficient quality for the SPC decision. Must be reviewed separately from capability calculation.

Time-ordered series with center line and limits calculated from the process to detect signals requiring investigation.

Chart evaluating each sample statistic against control limits and patterns. Direct for relatively large shifts.

Central reference on the chart, estimated from the process or fixed by chart design. Not a specification limit.

Lower and upper limits calculated from process variation. Crossing them generates a control signal, not an automatic conformity decision.

Chart pair for means and standard deviations of rational subgroups. Variation is checked before interpreting the means chart.

Chart pair for means and ranges of small rational subgroups. Range estimates within-subgroup variation.

Chart for genuine individual observations and consecutive moving ranges. Must not be constructed by flattening rational subgroups.

Chart for the proportion of nonconforming units. Accommodates variable sample sizes by computing limits for each subgroup.

Chart for the number of nonconforming units per sample. Appropriate when sample size is held constant.

Chart for the count of nonconformities when the inspected area of opportunity remains comparable across samples.

Chart for nonconformities per unit when the number of units or inspected opportunity varies.

Chart accumulating deviations from target to expose small, persistent shifts earlier than an isolated signal.

Chart combining each data point with weighted memory of previous ones. Limits vary during startup.

Chart linked to tolerance for processes with accepted, systematic location shifts. Not equivalent to an ordinary Shewhart chart.

State in which the sequence is consistent with expected common variation and shows no unexplained decisive signals.

Variation inherent in the system during routine operation. Reducing it requires process improvement, not adjustment after each point.

Shift not belonging to routine behavior. Must be investigated before adjusting; a point is not removed merely for being extreme.

Historical review used by the dashboard to decide whether a final study may adopt C nomenclature. Distinct from displaying operational signals.

Point or pattern flagged by a chart criterion. Prompts verification and investigation; does not by itself identify the physical root cause.

Intervening on common-cause variation as if it were a special cause. Typically adds variation and degrades process performance.

Symmetric model used by calculator for indices, ppm, and limits. If not fitting, those results remain indicative until an appropriate model is used.

Description of how location and variation change over time. Guides appropriate chart selection and capability interpretation.

Observation not belonging to the studied population for a justifiable reason. Being extreme or failing a statistical test is not sufficient to delete it.

Measure of distribution asymmetry. Helps detect when the normal model may poorly represent a characteristic.

Measure related to tail weight and peakedness of the distribution. The dashboard uses it alongside skewness.

Statistical test based on skewness and kurtosis. In the calculator, p < 0.05 blocks conformity claims under the normal model.

Bars showing where measurements concentrate. Allows comparing shape, center, and dispersion against specifications.

Plot contrasting ordered data against expected values under normality. Systematic curvature or extreme tails guide model review.

Plan defining how to confirm a signal, investigate the cause, act, and verify effectiveness without improvising adjustments.

Point, zone, trend, or run patterns flagged by the dashboard on charts. Their operational and retrospective roles are evaluated separately.