neqsim-control-authority-screening

Control Authority Screening

Use this skill to test whether a regulatory loop still has authority. A final control element driven onto its stop can no longer reject a disturbance, so ordinary unchanged variability begins passing straight through to the controlled variable and into a trip. The presenting symptom — "the process got worse" — points outward at the disturbance, while the cause sits inside the loop.

The screening is cheap: it needs a controller output and its process value, two tags that are almost always historised. Run it as a default check before an investigation commits to an external cause.

When to Use

• A controlled variable has degraded and the disturbance behind it looks unchanged.
• A control valve is reported wide open, bottomed out, or "already doing everything it can".
• An investigation into an external disturbance is being proposed, and the loop has not been ruled out.
• A root-cause analysis presents the symptom "the controlled variable got worse".
• A trip margin must be converted into the time an operator actually has to intervene.

Inputs

• controller_output: controller output history, same units as output_min and output_max.
• controlled_variable: process value history, aligned sample for sample with the output.
• output_min, output_max: output stops, default 0 and 100.
• saturation_band: distance from a stop that still counts as saturated, default 1.
• sample_interval_h: sampling interval in hours, default 1.
• set_point, set_point_tolerance: optional, for the off-set-point fraction.
• disturbance: optional aligned disturbance history, for the regime gain fits.
• limit_value: optional trip or alarm limit, for time-to-limit.

compare_periods takes a sequence of PeriodObservation(label, saturation_fraction, disturbance_std) covering comparable periods, for example the same season in successive years.

Outputs

• saturation_fraction, upper_saturation_fraction, lower_saturation_fraction, modulating_fraction.
• saturated and modulating: RegimeStatistics with sample count, mean, standard deviation, disturbance gain, and the gain fit r-squared.
• variance_ratio: controlled-variable variance while saturated over variance while modulating — the direct measure of rejection lost.
• gain_ratio: disturbance-to-process gain while saturated over the gain while modulating.
• off_set_point_fraction, max_rate_per_hour, margin_to_limit, time_to_limit_h.
• authority_status: authority-retained, authority-marginal, or authority-lost.
• warnings, assumptions.
• From compare_periods: saturation_change, saturation_relative_change, disturbance_change, disturbance_relative_change, verdict (control-authority-loss, disturbance-growth, both, no-significant-trend, indeterminate) and a one-paragraph narrative.