Tiegel home Talk to a metallurgist

Process control for metallurgists: PID loops, step tests and lambda tuning

PID control for metallurgists: loop structure, interpretation of a step test for gain, dead time and time constant, and lambda tuning with a cyclone density example.

· 6 minute read · By

A control loop automates the task of an operator who watches a gauge and adjusts a valve: if the cyclone feed is too dense, the water valve is opened slightly, and if it is too dilute, the valve is closed slightly. The controller performs this adjustment continuously according to the PID algorithm.

The algorithm has three terms. The proportional term responds to the current size of the error. The integral term continues to act while any error remains, so that the loop eventually reaches the setpoint exactly. The derivative term responds to the rate of change of the error and is rarely used on plant loops, because instrument signals are noisy. Tuning is the selection of the strength of each term.

This article describes the elements of a loop, the conduct and interpretation of a step test, and loop tuning by the lambda method, using a cyclone feed density loop as the example. It is written for metallurgists who work alongside control engineers, not as a substitute for them.

Feedback control loopThe controller compares the setpoint with the measured process variable and moves the final control element, such as a valve, to remove the error. A disturbance acts on the process.SetpointSPControllerPIDValve orpump speedProcesse.g. cyclone feedTransmittere.g. density gaugeSPOPPVdisturbance: ore hardness, feed rate
Every plant loop has the same parts. When a loop performs badly, the cause is as often the valve or the transmitter as the tuning.

The PID equation

u=Kc[e+1Ti∫e dt+Tddedt]e=SP−PV

where

u
controller output, for example valve position or pump speed (%)
e
error: setpoint minus process variable
Kc
controller gain
Ti
integral (reset) time (s or min)
Td
derivative time (s or min); usually zero on plant loops

The controller action must have the correct sign. Opening the dilution water valve lowers the density, so when the density rises above setpoint the controller must open the valve further. A controller whose output rises as the PV rises is called direct acting; most loops, in which opening the valve raises the PV, require reverse action instead. An incorrect action drives the loop to a limit.

Control loops in a mineral processing plant

LoopMeasured (PV)Adjusted (OP)
Mill feed rateBelt weightometerFeeder speed
Mill feed water ratioWater flow, with setpoint from ore feed rateWater valve
Cyclone feed sump levelLevel transmitterCyclone feed pump speed
Cyclone feed densityDensity gaugeSump dilution water valve
Thickener underflowBed level or bed pressure, underflow densityUnderflow pump speed, flocculant dose
Leach pH and cyanidepH probe, cyanide analyserLime and cyanide addition
Dissolved oxygenDO probeAir or oxygen flow

Many loops are combined. A ratio loop sets the mill water in proportion to the ore feed. A cascade loop has an outer controller, such as sump level, setting the setpoint of an inner controller, such as pump speed. Feedforward adjusts the output as soon as a measured disturbance changes, before the PV moves.

Step test interpretation

With the loop in manual and the process steady, the output is moved by a small step and the PV recorded. The response of most self-regulating plant loops is a delay followed by an exponential curve.

Step test: dilution water valve from 40% to 45% open at 20 sDensity holds at 1.620 for 15 seconds after the valve moves, then falls; it reaches 1.595, 63.2% of the change, at 80 seconds and settles at 1.580.1.571.581.591.601.611.621.63020406080100120140160180200Time (s)Density (t/m³)valve movesDensitydensity starts to move at 35 s: dead time 15 s63.2% of the change at 80 s: τ = 45 s
Reading a step test: the dead time is the delay before the density moves; the time constant is the further time to 63.2% of the total change. Illustrative figures.
Kp=ΔPVΔOPθ=dead timeτ=time to 63.2% of the change

where

Kp
process gain, with PV and OP both in % of their ranges
θ
dead time: delay before the PV starts to move (s)
τ
time constant (s), measured from the end of the dead time

Worked example: the dilution water valve is stepped from 40% to 45% open. The density gauge, ranged 1.0 to 2.0 t/m³, falls from 1.620 to 1.580 t/m³, a change of 4% of span for a 5% output step, so Kp = −4 ÷ 5 = −0.8. The density begins to move 15 s after the step (θ = 15 s) and reaches 63.2% of its total change 45 s after that (τ = 45 s).

Two or three steps should be made in each direction at normal operating conditions. Unequal responses up and down usually indicate a sticking valve or a non-linear process, neither of which tuning can correct.

Lambda tuning

Kc=τKp(λ+θ)Ti=τ

where

λ
chosen closed-loop time constant: how fast the loop should respond; a common, robust starting point is three times the larger of θ and τ

Worked example: with λ = 3 × 45 = 135 s, the controller gain is 45 ÷ (0.8 × (135 + 15)) = 0.375 and the integral time is 45 s, with the controller set to direct action so that it opens the valve on high density. A faster choice, λ = 45 s, gives a gain of 45 ÷ (0.8 × 60) = 0.94.

Response to a setpoint change of −2% of spanWith lambda 45 seconds the density reaches the new setpoint in about 3 minutes; with lambda 135 seconds it takes about 9 minutes; neither overshoots.-2.0-1.5-1.0-0.50.00100200300400500600Time (s)Change in density (% of span)λ = 45 s (λ = τ)λ = 135 s (3τ)
Lambda tuning trades speed for robustness. The slower setting stays stable even if the process gain or dead time doubles, which is common as ore and wear change. Illustrative figures.

Both settings reach the new setpoint without overshoot; the faster settles in about 3 minutes and the slower in about 9, close to the expected 4λ. The slower setting remains stable even if the process gain or dead time doubles, as occurs with changes in ore, flows and equipment wear. Interacting loops, such as sump level and cyclone density, should have lambdas that differ by a factor of three or more so that they do not oppose each other.

Checks before retuning

Valve stiction or backlash appears as a slow cycle in the PV and can be tested with small output steps. The transmitter should be checked for calibration, range, filtering and installation, for example a density gauge on a line containing entrained air. A valve that operates nearly shut or fully open has no range in which to control.

Integrating loops, such as levels, behave differently from self-regulating loops and use different tuning rules; averaging level control often calls for a slow loop that allows the level to float.

Glossary

PV (process variable)
The measured value the loop controls, such as density or level.
SP (setpoint)
The target value for the PV.
OP (output)
What the controller adjusts, such as valve position or pump speed.
Process gain
Change in PV divided by the change in OP that caused it, in % of range.
Dead time
The delay between an output change and the first movement of the PV.
Time constant
The time after the dead time for the PV to reach 63.2% of its final change.
Lambda (λ)
The chosen closed-loop time constant in lambda tuning.
Cascade control
An outer loop that sets the setpoint of an inner, faster loop.

Frequently asked questions

Is the derivative term needed?

Rarely on mineral processing loops. Derivative action responds to the rate of change, and noisy plant signals make it erratic. Most loops operate as PI controllers.

What lambda should be chosen?

A common robust starting point is three times the larger of the dead time and the time constant. A faster setting should be used only if the loop requires it and the process is consistent.

Why does a loop cycle even after tuning?

Often because of a sticking valve, an interacting loop or a transmitter problem. These should be checked with the loop in manual before the tuning is changed.

What is the difference between direct and reverse action?

A direct-acting controller raises its output when the PV rises; a reverse-acting controller lowers it. Reverse action suits loops in which more output raises the PV, such as a feeder and weightometer. The naming and application of the setting should be checked in the particular control system.

Discuss your plant with a metallurgist

Tiegel is a Perth metallurgy company. We help with testwork, plant optimisation, metal accounting and investigations, and we build the Tiegel Suite software.

Talk to a metallurgist

References and further reading

  1. Seborg, D.E., Edgar, T.F., Mellichamp, D.A. and Doyle, F.J. (2016). Process Dynamics and Control, 4th edition. Wiley.
  2. ISA InTech (June 2016). Basics column on lambda tuning. Link

Worked examples and charts marked illustrative use invented figures, not data from any operation.