Passive, semi-active and active control

Three categories, one distinguishing question

Passive. The device's properties are fixed. It responds to motion and supplies force from the motion itself. No power, no sensors, no control. Isolators, viscous dampers, tuned mass dampers, friction dampers, yielding braces.

Semi-active. The device's properties can be changed in real time — an orifice that opens and closes, a magnetorheological fluid whose viscosity changes with an applied field, a variable-stiffness brace. It cannot add energy to the structure; it can only choose how much to take out and when.

Active. An actuator applies a commanded force, drawing energy from an external supply. It can do things a passive device cannot, including pushing the structure in the direction it is already going, which is occasionally the right thing to do.

The question that separates them: where does the force come from? Passive — from the motion. Semi-active — from the motion, but modulated. Active — from a power supply.

What semi-active buys

A passive damper is a compromise: sized for one condition, it is wrong for the others. A semi-active damper can be soft when the structure is moving one way and stiff when it is moving the other, adapting to the event as it happens.

The attraction is the risk profile. Because it cannot add energy, a semi-active device cannot destabilise the structure. If the control system fails, the device reverts to a passive damper — perhaps not the best passive damper, but a safe one. It gets most of the benefit of active control with a fraction of the consequence when it fails.

Power requirements are also modest: enough to operate a valve or energise a coil, not to move the building. Battery backup is realistic.

This is why semi-active control is the category that has actually been deployed in significant numbers.

What an active system needs

Sensors. Accelerometers, displacement transducers, sometimes a wind or ground-motion input. They must be reliable, calibrated and maintained for the building's life.

A control algorithm. It decides the force to apply, from the sensed state and a model of the structure. There is a great deal of theory here — optimal control, sliding mode, fuzzy, neural — and the practical constraint is that the algorithm must work with the structure the building actually is, not the one in the model.

Actuators. Hydraulic rams, or a driven mass, capable of the required force and stroke. They are large, and they need real power.

Power. Continuous, and available during exactly the events for which the system exists. An earthquake that takes out the grid takes out the control system unless it has independent supply.

Maintenance. For decades, on a system whose failure may not be evident until the day it is needed.

Robustness and stability

An active system relies on a model of the structure to decide what force to apply. The real structure is not that model.

If the model's frequency is wrong, the commanded force can arrive at the wrong moment in the cycle. A force applied in phase with the velocity adds energy rather than removing it — the control system driving the structure instead of restraining it. In the worst case the closed loop is unstable and the system fails by exciting the very motion it was installed to suppress.

Control theory has well-developed tools for this — gain margin, phase margin, robust design against a family of plausible structures rather than a single nominal one. The engineering point is that the robustness question is not optional, and it is a different question from whether the algorithm is good.

Fail-safe behaviour

This is the requirement that governs everything.

A structural control system must fail into a state that is safe. Not merely 'not helpful' — safe. Specifically:

• loss of power must leave a passive device, not a free one;
• a failed sensor must be detected, and the system must degrade rather than act on bad data;
• a runaway actuator must be limited mechanically, not only in software;
• the structure must be adequate WITHOUT the control system operating, for the design event.

That last one is the decisive constraint, and it is worth stating plainly: an active control system is generally not permitted to be the reason a structure is safe. It is permitted to improve comfort, to reduce damage, to protect contents. If the building would be unsafe with the system switched off, then the safety of the building depends on a computer, a power supply and a set of sensors continuing to work through an earthquake.

That constraint, more than cost or complexity, is why active control is rare in buildings and common in aerospace, where the same reasoning leads to the opposite answer because there is no passive alternative.

Choosing

Passive for almost everything. Simple, reliable, no power, no maintenance beyond inspection. Start here and only leave it for a specific reason.

Semi-active where the excitation varies enough that a single passive setting is a poor compromise, and where the fail-safe argument can be made cleanly.

Active where the performance genuinely cannot be achieved passively and the consequence of failure is comfort rather than safety. In buildings this means, in practice, occupant comfort in wind on very tall towers.