When a simple PID is not enough, structures that combine loops are used. Cascade control uses a fast inner loop to absorb disturbances before they reach the main variable, ratio control keeps a proportion between two flows, and feedforward corrects a measured disturbance before it affects the process. This guide explains when each one fits and what each requires.
Why a simple PID sometimes is not enough
A feedback PID controller only reacts when the error already exists. If a disturbance takes time to show up in the main variable, the loop corrects late. If the process has a long dead time, the loop is slow by nature. To improve without changing the base instrumentation, loops are combined in known ways: cascade, ratio and feedforward. All three rely on the same instruments, but change the structure of the calculation.
Cascade, ratio and feedforward are standard process control architectures. Cascade combines two PIDs: one adjusts the setpoint of the other. It is a classic improvement in dynamic performance.
Cascade control
Structure
Cascade uses two controllers and two measurements for a single manipulated variable:
- Primary (master) loop. It controls the variable that really matters, for example the product temperature. Its output does not go to the valve, but to the setpoint of the secondary loop.
- Secondary (slave) loop. It controls an intermediate variable that responds quickly to the actuator, for example the steam flow. Its output does go to the valve.
The slave handles the disturbances that affect it directly, such as a change in steam pressure, before they show up in the product temperature. The master only adjusts the slave’s target.
When it is justified
| Condition | Why it matters |
|---|---|
| The inner loop is much faster than the outer one | Otherwise the two compete and the system oscillates |
| A measurable intermediate variable exists | Without a measurement there is no slave loop |
| The most frequent disturbances enter through the inner loop | That is where cascade pays off |
| There is nonlinearity or friction in the actuator | The inner loop compensates for it |
Common examples: temperature of a jacketed reactor (the master sets the jacket temperature), tank level with outlet flow as the inner loop, motor speed with an inner current loop (the one every variable frequency drive already includes).
Tuning and commissioning
The order matters:
- Tune the secondary loop first, with the primary in manual.
- With the secondary stable and in automatic (local mode), tune the primary. For the primary, the secondary is part of the process.
- Switch to cascade mode, where the slave’s SP is set by the master.
The secondary must respond clearly faster than the primary; a practical rule used in engineering is that its time constant be several times smaller. If it is not, the benefit disappears.
There is one essential detail: when the secondary saturates, the primary must find out and stop integrating, or the integral windup phenomenon described in another guide of the series occurs. Most industrial controllers have a tracking connection for this.
Ratio control
Idea
In many processes what matters is keeping a proportion between two flows: air and fuel in a boiler, ingredients in a blend, additive and water flow in a treatment process. Ratio control measures the “free” flow, which is set by production, and sets the SP of the second flow as:
SP₂ = R × Q₁
where R is the desired ratio and Q₁ the measured flow of the free stream.
Variants
- Manually adjusted ratio: the operator changes R.
- Ratio adjusted by a controller: another loop, for example the oxygen loop in flue gas, corrects R. It resembles a cascade in that the master output modifies the ratio.
- Cross-limiting: in combustion it is restricted so that, as load rises, air increases first and then fuel, and as load falls the reverse happens. It avoids fuel-rich mixtures during transients.
Points of attention
- Both flows must be in compatible units, and the measurement of both must be corrected for density, pressure or temperature when they vary.
- A badly scaled ratio produces a wrong mixture that the control does not detect, because each individual loop “works fine”.
- The free flow has noise and fast changes; if they are passed as is to the second SP, the actuator will move too much. A filter or a rate limit is used.
Feedforward control
Idea
Feedback control waits for the error. Feedforward measures the disturbance before it affects the process and immediately corrects the manipulated variable, based on a model of how the disturbance affects the output. It is not based on the error but on knowledge of the process; therefore, if an incorrect action is sent, it cannot correct itself.
The classic example is the heat exchanger: if the product inlet flow increases, the feedforward controller immediately raises the steam by the proportion the energy balance indicates, without waiting for the outlet temperature to fall.
Combination with feedback
Pure feedforward depends on the accuracy of the model and on the disturbance being measured. That is why it is almost always used together with a PID: feedforward does most of the correction and the PID picks up what the model did not foresee. The final output is the sum of both contributions.
What is needed
- A measurable and significant disturbance: load flow, inlet temperature, composition.
- A model of its effect, at least static (gain) and, if needed, dynamic (lead and lag to synchronize the correction with the disturbance).
- Maintenance: if the process changes, the model goes out of date. The cost of developing and maintaining the model is the reason feedforward is used less than a simple PID.
When to use each
| Problem | Strategy that usually fits |
|---|---|
| Disturbances at the actuator or in the actuator supply | Cascade |
| Two flows that must keep a proportion | Ratio |
| Measurable load disturbance, with slow effect on the output | Feedforward + PID |
| Large dead time and measurable disturbances | Feedforward |
| A single stable loop without frequent disturbances | Simple PID |
Combining all three in one plant
The structures are not mutually exclusive. A steam boiler, for example, can have a cascade of steam pressure over fuel flow, an air-fuel ratio with oxygen trim and a feedforward that uses steam demand to move the fuel before the pressure drops. Each layer adds a point of failure and a piece of documentation, so it is added only when the measurable benefit justifies it. A sensible practice is to start with the simple PID, record its performance and add structure only where the trend shows that disturbances are not rejected in time.
Numerical example (illustrative)
The numbers are made up. A blending tank receives a water flow that changes with production and an additive is added in a proportion of 2 % of the water flow.
- Measured water flow: 50 m³/h. Additive SP: 0.02 × 50 = 1.0 m³/h.
- If the flow rises to 60 m³/h, the additive SP immediately moves to 0.02 × 60 = 1.2 m³/h.
- A standalone controller on the final concentration would correct the ratio only after the mixture had left specification. Ratio adjusts beforehand.
- If the concentration measured at the outlet reads 2.2 % instead of 2 %, an outer loop can correct R from 0.020 to 0.0182 to recover the target.
The ratio does the immediate work and the concentration loop corrects the residual error, which is the general pattern of these structures.
Common mistakes
- Cascade with loops of similar speed. The two controllers interact and oscillate.
- Tuning both loops at once. You must start with the inner one.
- Not configuring tracking or anti-windup in cascade. The primary keeps integrating while the secondary is saturated.
- Ratio without compensating for density or temperature. The real proportion drifts away from the desired one.
- Feedforward without feedback. Any model error is left uncorrected.
- Disturbance measured with noise. It passes the noise directly to the actuator.
- Lack of documentation of the strategy. When someone puts a loop in manual, nobody knows what dependencies it had.
Checklist
- Main disturbances identified and their measurements available.
- Cascade intermediate variable measured and much faster than the main one.
- Tuning order defined: secondary first.
- Tracking and anti-windup configured between loops.
- Units and compensations of the ratio flows verified.
- Feedforward model documented, with its date and source data.
- Filter or rate limit on the measured disturbance.
- Operating modes tested: local, cascade, manual, failure of an instrument.
- Fail-safe behavior of each strategy on loss of measurement.
Engineering support
Choosing the strategy, distributing the loops among controllers and documenting the dependencies is design work. To define the control architecture and the instrumentation, see engineering design and studies, delivered remotely and worldwide. To check each operating mode and the response to failures before handover, see testing and commissioning, in Colombia or elsewhere on request. You can write to us through contact. To set the PID gains, the Ziegler-Nichols method and its alternatives are explained in Ziegler-Nichols tuning and alternatives.
Related guides
- ABB ACS880 drive
- Siemens SINAMICS drive
- Schneider Altivar drive
- WEG CFW500, CFW11 and MVW3000 drives
- Motor control center (MCC)
- PID windup and limits: causes and remedies
- Control valves: types and positioners
References
- Woolf, P., Chemical Process Dynamics and Controls, University of Michigan, chapter on control architectures (CC BY 3.0).
- Kuphaldt, T. R., Lessons in Industrial Instrumentation, chapter on basic control strategies (CC BY 4.0).
- Wikipedia, Feed forward (control) and PID controller (two-PID cascade).
Frequently asked questions
What advantage does cascade have over a single PID?
The inner loop corrects disturbances, such as a pressure change upstream of a valve, before they affect the main variable. The main loop no longer has to wait for the effect to show in the final measurement, so the response is faster and more stable.
When does cascade not work?
When the inner loop is not clearly faster than the outer one, when there is no measurable intermediate variable or when the disturbance enters directly into the outer loop. In those cases it only adds complexity and can cause oscillations.
Does feedforward control replace the PID?
No. Feedforward compensates for what is known and measured, but it depends on a model and does not correct model errors or unmeasured disturbances. That is why it is almost always combined with a feedback PID that corrects the rest.
Is ratio control the same as cascade?
No. In ratio control a free flow, the one imposed by the process, is measured and the SP of a second flow is calculated by multiplying by the desired ratio. The ratio can be adjusted by another controller, and in that case it does resemble a cascade.