๐Ÿ“– This is a static documentation page. Open the interactive version with live canvas demos โ†’

Quality Metrics

The application calculates a set of statistical quality metrics for every operation. These metrics describe how predictable and stable the operation's cycle time is โ€” not the quality of the product itself, but the quality of the process timing.

Quality Level (QL)

The sigma-equivalent of timing stability. Range: 3.0 โ€“ 5.8.

QL = 3.0 + 2.8 ร— (0.6 ร— mechanisation/5 + 0.4 ร— productionType/3)

How Mechanisation Affects QL

Mechanisation describes the automation level of the operation:

  • Manual Labour (0) โ€” all work performed by human hands
  • Manual Mechanised (1) โ€” hand tools powered by external energy
  • Machine Manual (2) โ€” machine performs work, human controls each cycle
  • Machine (3) โ€” machine cycle is automatic, human loads/unloads
  • Automated (4) โ€” automated cycle with human supervision
  • Automatic (5) โ€” fully autonomous, no human intervention per cycle

Higher mechanisation โ†’ more predictable timing โ†’ higher QL.

How Production Type Affects QL

Production type describes the batch volume context:

  • Small (0) โ€” prototype or job-shop
  • Average (1) โ€” small series
  • Large (2) โ€” established mass production
  • Mass (3) โ€” high-volume continuous production

Higher volume โ†’ more established process โ†’ higher QL.

Reading the Formula

The formula weights mechanisation at 60% and production type at 40%. Both inputs are normalised to 0โ€“1 scale, then combined and scaled to the 3.0โ€“5.8 output range.

  • QL = 3.0 โ€” manual work in a job shop. Maximum timing variability.
  • QL = 5.8 โ€” fully automatic line in mass production. Maximum predictability.

QL does not judge whether the operation is good or bad. It quantifies process capability from a time perspective. A Q = 3.0 operation may be perfectly correct โ€” it simply has wider natural timing variation.


DPMO โ€” Defects Per Million Opportunities

The expected number of timing deviations beyond rejection limits per 1 000 000 operation executions.

DPMO is derived from QL using the standard normal distribution. As QL increases, the distribution tails narrow and DPMO drops.

  • QL = 3.0 โ†’ DPMO โ‰ˆ 66 807
  • QL = 4.0 โ†’ DPMO โ‰ˆ 6 210
  • QL = 5.0 โ†’ DPMO โ‰ˆ 233
  • QL = 5.8 โ†’ DPMO โ‰ˆ 3

A timing "defect" means the observed execution time fell outside the rejection limits โ€” the operation took significantly longer or shorter than expected. It is not a product defect.


Stability (S)

A normalised score from 0 to 1 measuring how consistent the process timing is:

S = clamp(0โ€ฆ1, 0.6 ร— mechanisation/5 + 0.4 ร— productionType/3)

The formula is the same weighted combination as QL, but without the 3.0 offset and 2.8 scale:

  • S = 0 โ€” least stable (manual, small batch)
  • S = 1 โ€” most stable (fully automatic, mass production)

Stability directly influences the coefficient of variation used in simulation. Lower stability produces wider time distributions, meaning more spread in the sampled execution times.


Rejection Limits

At the operation level, the result includes two rejection limits:

  • rjl_min โ€” lower rejection boundary
  • rjl_max โ€” upper rejection boundary

These are derived from the physical minimum and maximum times, adjusted by the quality level. Any observed execution time outside these limits counts as a DPMO event.

The physical bounds themselves are computed from the base time and standard deviation:

  • sigma = coefficient_of_variation ร— base_time
  • lo = base โˆ’ k ร— sigma (physical minimum)
  • hi = base + k ร— sigma (physical maximum)

Where k depends on the stability level.


Where These Metrics Appear

On the Result Block

The expanded result block for each operation shows: Q (quality level), D (DPMO), and S (stability) alongside the time components.

In the Simulation System

The simulation uses QL and S to generate realistic time samples. Each call to `sample()` produces:

  • A sampled time value drawn from a constrained normal distribution
  • A metadata object containing sigma, lo, hi, stab, QL, and dpmo

This means you can run many samples to observe the expected distribution for a given process configuration.


Improving Quality Metrics

To improve QL, S, and DPMO:

  • Increase the mechanisation level of the operation (replace manual work with tooling or automation)
  • Correctly classify the production type to reflect actual batch volumes
  • Both changes take immediate effect in the calculation โ€” no recalculation delay

Quality metrics are automatic. You do not enter them manually. They are derived from the operation's mechanisation and production type settings. Changing those two parameters is the only way to influence Q, S, and DPMO.