November 18, 2015
Capability Scoring

Capability Scoring

I recently helped apply for and received a patent (US9182758) for "Computer-implemented system and method for capability zone-based manufacturing quality control". I'm very proud of this work, and I'd like to briefly explain it.

Overview

This patent is for a computer system for improving manufacturing quality control by analyzing how close measured features are to their nominal (target) values. This means it looks at not just whether they fall within tolerance. It assigns each measurement to a “capability zone” and computes capability scores for suppliers, machines, operators, etc.

Problem It Solves

Conventional SPC methods often fail when:

  • Parts are within tolerance but cluster near a tolerance limit.
  • Multiple suppliers manufacture parts intended to mate or interact.
  • Small deviations from nominal reduce assembly yield or cause fit issues.

The invention provides a more nuanced view of quality by examining distance from nominal.

How It Works

1. Inputs

  • Feature specification: nominal value and tolerance
  • A dataset of actual measured values
  • Optional attributes: manufacturer, machine, operator, shift, etc.

2. Measurement Processing

For each measurement, we:

  • Compute deviation from nominal
  • Convert deviation to a fraction (or percentage) of the tolerance
  • Assign the measurement to a capability zone based on this distance

3. Capability Zones

Zones typically correspond to regions like:

  • 6-sigma zone: closest to nominal
  • 5-sigma, 4-sigma, …
  • Out of tolerance zone: this is the worst and has significant impact on score

Zones are defined as ranges of percent-of-tolerance from nominal.

4. Capability Scoring

  • Each zone has a weight (higher is worse)
  • Weighted counts are aggregated to produce a 0–100 capability score for a supplier, operator, feature, machine, etc.
  • Scores compare performance across any combination of attributes

5. Visualization & UI

The system produces:

  • Bar charts showing measurement distribution across zones
  • Color-coded displays
  • Ranked lists of suppliers/operators by capability
  • Dashboards highlighting where processes drift toward tolerance limits

Benefits

  • Provides insight beyond simple pass/fail metrics
  • Helps choose suppliers whose output clusters closer to nominal
  • Detects subtle but important process drift
  • Reduces risk of assembly issues caused by asymmetric or biased measurements
  • Works well in high-precision industries (for example, aerospace, where we focus)

Example

If tolerance is ±0.010 cm and a measurement is 0.002 cm from nominal, it is 20% of tolerance. This places it in a specific zone (e.g., 5-sigma). Many such measurements form a distribution that determines a capability score.

Implementation Notes

  • The system may run on servers or cloud platforms
  • Measurements exactly equal to nominal may be excluded to avoid skewing results. In manufacturing and high-precision industries, these are rarely valuable pieces of data since they are often fabricated or put is as defaults
  • The system can map attributes to causes to recommended corrective actions.