Designing an isolated building

The moat

An isolated building needs a gap around it, at every level below the isolation plane, wide enough for the building to move without striking anything. That gap is the moat, and it is covered by plates that slide.

The design width is not the analysis displacement. It is larger, and for several accumulating reasons:

Bidirectional motion. The analysis displacement is usually reported per direction. The bearing moves in a plane, and the resultant of two near-simultaneous peaks exceeds either.

Torsion. Any eccentricity between the centre of mass and the centre of isolator stiffness rotates the building on its bearings, and the corner bearings move furthest.

Record-to-record scatter. The displacement from one record is one sample. The design value comes from an envelope or a mean over a set, and the scatter on displacement is wide.

Bearing property variation. Rubber stiffness varies with temperature, with age, with rate, and with scragging history. Design codes require the analysis to be run with bounding upper and lower stiffness values, and the low-stiffness bound governs the displacement.

The consequence is that a computed 105 mm might become a required moat of 200 mm or more, and a moat is expensive at every level of the basement.

Impact against a moat wall is not a serviceability event. It reintroduces, in a single instant, the high-frequency acceleration the isolation existed to remove — and it does so at the base of a building detailed on the assumption that those accelerations were gone.

Everything that crosses the plane

At the isolation plane, the building above moves and the ground below does not. Every single thing that spans it must accommodate the full design displacement:

• water, drainage, gas and fire mains — flexible loops or articulated joints;
• electrical and data — service loops with slack, not taut runs;
• lifts — the shaft moves with the building, so the pit detail and the guide rails have to be resolved at the plane;
• stairs and ramps entering below the plane — sliding or hinged connections;
• any structure attached to both sides, which is generally simply not allowed.

A single rigid connection across the isolation plane can short-circuit the entire system. This has happened: a services contractor bracing a pipe to both the building and the ground, in good faith, because nothing on the drawing said not to. Isolation planes are labelled and inspected for exactly this reason.

Wind restraint

Here is the tension at the heart of the design. The isolation layer is deliberately soft, so a building that is soft enough to be isolated from an earthquake is also soft enough to sway noticeably in wind.

The resolution is a system that is stiff under service loads and soft under severe ones:

• lead cores in elastomeric bearings, which are stiff until the lead yields;
• friction in sliding systems, which does not move until the shear exceeds the friction force;
• separate wind restraints designed to fail at a defined force — a deliberate fuse.

The criterion is comfort as well as strength: the isolated period is around 3 s, which is close to where people are most sensitive to horizontal motion.

Re-centring

After the earthquake the building must return to somewhere near its original position. Elastomeric bearings do this naturally — they are springs. Sliding systems do not, unless they are given a restoring mechanism, which is why friction pendulum bearings have a curved surface: gravity supplies the restoring force, and the radius of curvature sets the period.

A system with poor re-centring leaves a residual offset at the isolation plane, and that offset consumes moat width in any subsequent event.

When isolation is the wrong answer

Soft soil sites. Deep soft deposits amplify long-period motion, so the site's own spectrum may peak near 3 s — exactly where the isolated building has been placed. Isolation can make things worse rather than better, and the site spectrum has to be checked at the isolated period before anything else is decided.

Tall, flexible buildings. A 30-storey building already has a period of 3 s or more. There is no shift available; the bearings would have to be absurdly soft, and uplift on the windward bearings becomes the governing problem.

Near-fault sites. Forward-directivity pulses have long periods and large displacements, and they drive isolated buildings hard. Isolation is still used near faults, but the displacement demands are much larger and the analysis has to use pulse-type records.

Where the moat cannot be built. Tight urban sites, party walls, buildings extended from existing structures. This is a practical constraint and it is often decisive.

Isolation suits stiff, low-to-medium-rise structures on firm ground, especially where the CONTENTS matter more than the structure — hospitals, data centres, museums, emergency facilities. That is the profile it was invented for, and it is where it is still most convincing.