02 · Focus Area 2 · Modernising legacy visualisations

Read the control charts

Every HPLC run includes control injections: a reference sample with a known answer. If the control drifts, the samples measured alongside it are suspect. A control chart plots each control result in time order against limits set from the process's own variation.

The 2021 team rebuilt these charts with one reusable function: the mean, the sample standard deviation, bands at 1, 2 and 3σ, and a printed list of runs beyond 3σ. That method is the default here. The run rules (with their false-alarm cost), the I-MR option, the disturbance tool and the optional AI explanation are additions for this revival.

Control assay

Control injection

Measurement

Data seed

Cation exchange. Charge-variant profile: acidic, main and basic peaks.

Rel.Area Basic · individuals, μ ± 3σ

n = 151

Hover or focus the chart and use the arrow keys to read individual injections. Filled red points are beyond 3σ; ringed points belong to another active rule.

Centre line

10.764

Sigma (sample SD)

0.546

Beyond 3 sigma

2 signals from 1 active rule

Explain this chartoptional · AI · your own key

Sends only this chart's numeric summary (limits, rules, flagged points, run statistics and the published false-alarm figures) to Anthropic, from your browser. What is sent

Out of control

The notebook printed one line per point beyond the 3σ limits. Same format, current chart.

Out of control:
20200217_CEX_Vega_S3: 7.854
20200217_CEX_Vega_S3: 7.645

Sequences flagged across measurements

For Control std B (system suitability), every measurement charted the 2021 way. Streams come from joining each sequence's sample injections back to the parsed lab sheet on Sample Name, as the team did.

Sequences with points beyond 3 sigma, by number of measurements affected
Sequence Name	Measurements beyond 3σ	Streams in this run
20200217_CEX_Vega_S3		Upstream
20201223_CEX_Nova_S1		Formulation
20201116_CEX_Atlas_S3		Concentrate

Limits

Phase I windowLimits from the first 40 points only

Limits from every point let a sustained shift inflate sigma and hide itself. Phase I limits avoid that but are noisier. See the study

Run rules

2021 notebook: on a 150-point in-control chart with known limits, at least one false alarm 33% of the time (95% CI 32% to 35%). Limits estimated from the data, as on this chart, behave differently (section B). Compare rule sets

WE1 1 point beyond 3σ2One point outside the control limits. This is the only rule the 2021 notebook applied.
WE2 2 of 3 beyond 2σTwo of three consecutive points beyond 2 sigma on the same side.
WE3 4 of 5 beyond 1σFour of five consecutive points beyond 1 sigma on the same side.
WE4 8 on one sideEight consecutive points on the same side of the centre line: a sustained shift.
N3 6 rising or fallingSix points in a row steadily increasing or decreasing: a drift.
N4 14 alternatingFourteen points in a row alternating up and down: over-adjustment or two alternating sources.
N7 15 within 1σFifteen points in a row within 1 sigma: limits may be too wide or data stratified.
N8 8 outside 1σEight points in a row outside 1 sigma on both sides: a mixture of two processes.

Inject a disturbance

From injection#101

Size1.5σ

Sizes are in units of this series' sample SD. Try a +1σ shift: the 3σ rule barely notices, the eight-on-one-side rule does.