Signalise

Product benchmark

What a supervisory autopilot is worth on your plant

A matched A/B on the Cerro Corona (autobench) flowsheet — the circuit on its existing PID, then under the Signalise autopilot, same ore and seed. Measured over 6,120 steps (~102 h of plant time).

PID vs. Signalisesettled A/B · 6,120 steps
Throughput+13.8%
662
753
Cu recovery+2.4%
87.0
89.1
Au recovery+2.8%
67.5
69.3
Grind energy−3.7%
23.56
22.67
SAG overload−1.9%
38.2
37.5

How the value shows up

Read top to bottom. Incoming ore hardness (the disturbance) drives the mill. PID holds the SAG-speed setpoint at a static nominal; Signalise rides it up as hard ore loads the mill — capturing throughput and recovery the fixed setpoint leaves on the table, without tipping the mill into overload.

ORE HARDNESS — THE DISTURBANCE24 BWI6SAG SPEED SETPOINT — THE CONTROL ACTION81 rpm75MILLED THROUGHPUT — PRIMARY OUTCOME994 t/h411RECOVERY (CU + AU) — SECONDARY OUTCOME162 %52
Shadow · PID baselineSignalise · autopilot

PID vs. Signalise, stat by stat

Settled-state averages over the second half of the matched run, so the comparison is like-for-like. Bars grow from parity toward the better arm; the plant runs smoother, not just harder — recovery and throughput up, energy and overload down.

662
Throughput t/h
+13.8%
753
87.0
Cu recovery %
+2.4%
89.1
67.5
Au recovery %
+2.8%
69.3
23.56
Grind energy kWh/t
−3.7%
22.67

lower is better

38.2
SAG overload % of time
−1.9%
37.5

time above trip load

How it's measured

A matched A/B, not a demo

Both arms run the same simulated ore body (a four-mode hardness distribution) from an identical seed and warm-up. The PID arm is the regulatory baseline; the Signalise arm adds the supervisory autopilot on top of that same PID, so the A/B isolates the autopilot's effect. Percentages are computed from settled-state means — no cherry-picked window.

This report is a single representative seed. The same pipeline runs on your plant from historian data: it establishes your PID baseline, quantifies the achievable uplift, and reports it alongside the world model's accuracy — before a single setpoint is dispatched. Headline economics are computed per plant from your grades, prices, and operating hours.