1. Tool Overview

The Process Capability Analysis Tool evaluates continuous measurement data against specification limits, including process spread, centering, and estimated nonconformance risk. The current release supports Single, Subgroup, and Historical Capability Analysis.

2. Quick Start

1. Select an analysis mode from the top bar, or load a runnable dataset from Demo Cases.
2. Upload an Excel/CSV file or paste measurement data.
3. Review the detected data type and valid value count; confirm field mapping when required.
4. Enter the product or process name, LSL, Target, USL, and capability requirement.
5. Select Calculate or Run Analysis, then review the Dashboard and Report tabs.

3. Analysis Modes

4. Interface and Three-Step Workflow

• Measurement Data: Upload or paste data, review automatic recognition, and confirm measurement or historical field mapping when required.
• Analysis Setup: Enter report identity, LSL, Target, USL, capability requirement, report language, and decimal precision.
• Run: Start calculation, then review Dashboard and Report.

Demo Cases provides runnable Single, Subgroup, and Historical datasets. Local Save/Restore uses browser localStorage and does not synchronize across devices.

5. Data Formats and Minimum Requirements

Single-column data

Subgroup data

Historical data

Comma-separated, line-separated, tabular, and common spreadsheet formats are supported. Blank cells are excluded and nonnumeric content is not included in calculations. Excel import requires SheetJS to load successfully; CSV and pasted data do not depend on that library.

• At least 2 valid values with sample standard deviation greater than zero.
• Subgroup mode requires at least 2 subgroups with at least 2 values each.
• Samples below 30 can be calculated but receive a caution message.

6. Specifications and Settings

• LSL / USL: Lower and upper specification limits. USL must be greater than LSL.
• Target: Engineering target used to evaluate centering and draw the target line; it is not part of the Cp/Cpk formulas.
• Capability Requirement: Threshold used to classify capability indices, such as 1.330.
• Prepared By / Team: Report metadata only; it does not affect calculations.

7. Single Capability Analysis

1. Select Single Analysis and Single Measurement Data.
2. Upload or paste data; verify the measurement column, valid count, and missing-value status.
3. Enter specifications and the capability requirement, then Calculate.
4. Review data validity, centering, Cp/Cpk, Pp/Ppk, PPM, and normality in that order.

Without rational subgroups, the same sample standard deviation may estimate within and overall variation. Cp=Pp or Cpk=Ppk does not prove that short- and long-term variation are equal.

8. Subgroup Capability Analysis

Subgroups should represent observations produced under approximately the same short-term conditions. Long format uses Subgroup/Measurement columns; matrix format treats each row as one subgroup. Within SD is estimated from pooled within-group sums of squares.

• Cp/Cpk use pooled within SD and describe short-term capability.
• Pp/Ppk use overall sample SD and describe current overall performance.
• When Cpk materially exceeds Ppk, investigate time, lot, setup, line-change, or maintenance effects.
• Unequal subgroup sizes are supported but should be reviewed for sampling consistency.

9. Historical Capability Analysis

1. Import a historical table with headers.
2. Review high/medium/low recognition confidence; low confidence requires field correction.
3. Confirm group and measurement columns plus optional characteristic, part, and lot fields.
4. Select one characteristic when the source contains several.
5. Review valid/excluded rows and group count before running.

Historical output includes group N, mean, sample SD, Pp, Ppk, minimum, maximum, status, warnings, and Ppk/mean/SD trends. Historical groups are not necessarily rational subgroups, so Cp/Cpk are intentionally omitted.

10. Excel Recognition

• Scans multiple worksheets and selects the most likely data block.
• Recognizes single-column, long subgroup, wide subgroup, and row-matrix layouts.
• Ignores common date, sequence, operator, and remarks columns.
• Can detect LSL, USL, and Target from metadata rows.
• Requests user confirmation for multiple candidates or low-confidence detection.

If Excel import is unavailable, save as CSV or paste from Excel. Do not place unit text inside measurement cells.

11. Calculation Methods

Overall SD is the sample standard deviation with n-1 denominator. Subgroup within SD uses pooled within-group sums of squares. Estimated PPM uses normal-tail probabilities from the current mean and overall SD.

12. Understanding Results

• Cp: Potential capability based on within standard deviation; it compares tolerance width with short-term variation.
• Cpk: Accounts for within variation and the mean position relative to the nearest specification limit.
• Pp / Ppk: Use overall standard deviation and describe the long-term performance represented by the current data.
• Within / Overall SD: A material difference may indicate lot-to-lot, time-related, or other long-term variation.
• Estimated PPM: A model-based estimate from the current mean, overall standard deviation, and fitted normal distribution; it is not the observed defect count.
• Normality Assessment: Uses the Anderson-Darling test. Failure to reject normality does not prove process stability or sample representativeness.

Results from fewer than 30 observations require caution. Capability cannot be calculated when fewer than 2 valid values exist or the standard deviation is zero.

13. Dashboard, Plot, and Report

The Dashboard provides KPI values, a capability distribution plot, process statistics, and an engineering summary. The plot includes observed data, within/overall distributions, specification limits, target, mean, and a Sigma scale.

The Report provides an engineering-document preview, HTML export, and PDF export through the browser print dialog. Report language follows the selected Chinese/English interface. An optional company logo is stored locally and scaled into a fixed 160 × 40 px area.

Before formal export, verify language, specifications, requirement, warnings, and raw-data appendix. For PDF, use A4 and disable browser-generated headers and footers.

14. Local Save, Logo, and Privacy

Local Save writes workspace settings to browser localStorage. Restore reads them in the same browser. Clearing site data, using private browsing, or changing devices can make saved content unavailable.

The company logo also remains local and is aspect-fit into a fixed area. Measurement values, specifications, filenames, column names, and report content are not uploaded.

15. Troubleshooting

• No result: Check valid count, SD, LSL/USL, and capability requirement.
• Subgroup error: Confirm at least 2 groups and 2 values per group.
• Historical mapping not ready: Select a numeric measurement field and valid group field.
• Excel library unavailable: Use CSV or pasted data.
• PDF did not open: Allow the site to open the print window.
• Saved workspace missing: Use the same browser and URL and confirm site data was not cleared.

16. Limitations and Recommended Practice

• Confirm the measurement system and process stability using appropriate evidence before production release.
• Historical Analysis currently reports Pp and Ppk only because historical groups are not necessarily rational subgroups.
• One-sided specifications, nonnormal or mixed distributions, and material outliers require additional engineering judgment.
• Report conclusions support decisions but do not replace the organization's quality approval process.

For formal use, retain source data, sampling context, specification revision, warnings, report, and tool version, then complete a controlled review by authorized personnel.