A raft foundation, also called a mat foundation, is a single continuous slab carrying all the columns and walls of a structure and spreading their loads over the whole footprint. It is used where individual footings would cover so much of the plan area that a continuous slab becomes simpler, or where the ground is too weak or too variable for separate footings to perform acceptably.
Designing one is a soil-structure interaction problem rather than a bearing capacity problem. The slab distributes load according to its own stiffness relative to the ground beneath it, and the ground responds according to how much the slab deflects at each point. Neither can be resolved without the other.
That mutual dependence is usually handled by representing the soil as a bed of springs, and the quality of a raft design follows almost entirely from how those springs were derived.
A raft does not apply a pressure to the ground. It negotiates one.
When Is a Raft the Right Foundation?
Three situations, and a fourth that is about performance rather than geometry.
When individual footings would occupy roughly half the plan area or more, the excavation, formwork and reinforcement of separate pads stop being cheaper than a continuous slab. When the soil is weak enough that the required footing area becomes impractical, the raft spreads load over the whole footprint and reduces the applied pressure accordingly. When the bearing stratum is variable across the site, a stiff slab bridges the weaker zones and evens out what individual footings would expose.
The fourth case is differential settlement. Even where separate footings would carry the load, a raft can be chosen because it ties the structure together and reduces the relative movement between columns, which is usually what damages a building rather than the total settlement.
What Governs Raft Design: Bearing Capacity or Settlement?
Settlement, in nearly every case.
Because the load is spread over a large area, the applied pressure is low and the bearing capacity check is rarely close. The large dimensions also raise the capacity itself: the width term of the bearing capacity equation grows with foundation width, so a raft, particularly on granular soil, has a considerably higher ultimate bearing capacity than an isolated footing on the same ground. A low applied pressure set against a higher capacity leaves a wide margin.
The same large area, however, means the stress increase reaches deep into the profile, so compressible layers well below the slab contribute to the settlement that would never have mattered under an isolated footing.
This reverses the usual intuition. A raft reduces the pressure and increases the depth of influence, and the second effect is the one that governs. A layer at a depth equal to the raft width is inside the stressed zone; under a two-metre footing on the same site it is not.
Why Does Differential Settlement Matter More Than Total?
Because a structure that settles uniformly is not distorted by settling, although uniform settlement still has limits of its own.
Design codes and project specifications set limiting values for total settlement as well as for differential movement, because total settlement affects connections to services, entrance levels and relationships with adjacent structures. Differential settlement imposes distortion on the frame, and it is that distortion which cracks partitions, jams doors and overstresses members.
A raft reduces differential settlement by being stiff, but not to zero. The classic pattern on uniform ground is a dished profile with more settlement at the centre than at the edges, because the stress from the whole loaded area accumulates beneath the middle. The resulting curvature is what the slab has to be designed for, and it is a function of the slab stiffness relative to the soil rather than of the loads alone.
How Is the Soil Represented in the Structural Model?
As a bed of independent springs, one under each node of the slab.
Each spring has a stiffness expressed as the modulus of subgrade reaction, the pressure required to produce unit settlement:
ks = q / s
where q is the contact pressure and s the settlement at that point. The structural analysis then solves the slab on that elastic foundation, producing contact pressures, bending moments and deflections together.
The model is called a Winkler foundation, and its defining simplification is that the springs are independent: loading one does not deflect its neighbours. Real soil is a continuum, so a load at one point settles the ground around it as well. That is the central limitation, and most of the practical difficulties in raft design come from it.
Why Is a Single Subgrade Modulus Wrong?
Because the value is not a soil property, and because the continuum effect is not uniform.
Two points follow from this. First, ks depends on the loaded area. A plate load test on a 300 mm plate gives a stiffness for a 300 mm plate, and applying that value to a raft twenty metres across overestimates the stiffness severely, because the larger area stresses far more soil. The correction between the two is not optional.
Second, ks is not constant across the raft. Because independent springs cannot represent the way real soil transfers stress sideways, a uniform spring bed produces uniform settlement under uniform load, whereas the real profile is dished. The usual remedy is to stiffen the springs at the edges and corners and soften them towards the centre, which recovers approximately the settlement shape the continuum would give.
Both corrections push in the same direction: a single value taken from a table and applied everywhere will underestimate the differential settlement, and therefore the bending moments, that the slab has to carry.
How Should the Subgrade Modulus Be Derived?
From the settlement the raft is actually expected to undergo.
The consistent approach is to calculate settlement for the raft using a method that treats the soil as a continuum, through the layered profile and the stress distribution the raft produces, and then to back out the spring stiffness from that settlement at each location. The springs then reproduce a settlement field that a proper settlement analysis already established, and the structural model inherits the correct behaviour.
The alternative routes are the Vesić expression, which relates the modulus to the soil stiffness and the foundation geometry, and a plate load test corrected for size. Both are defensible; both should be checked against a calculated settlement rather than trusted alone.
A useful discipline is to run the structural model, sum the spring reactions, compare the resulting settlement with the geotechnical prediction, and iterate if they disagree. If the structural model settles 15 mm where the settlement analysis said 40 mm, the springs are wrong and the moments derived from them are too.
Rigid or Flexible: Which Assumption Applies?
It depends on the relative stiffness of slab and soil, not on the slab thickness alone.
A raft that is stiff compared with the ground rotates and translates roughly as a plane, distributing pressure in a way that can be computed from statics. A raft that is flexible compared with the ground follows the load: pressure concentrates beneath columns and the slab between them lifts relatively.
Real rafts sit between the two. The practical consequence is that the pressure distribution assumed for a preliminary bearing check, usually uniform, is not what the structural analysis will produce, and the difference is largest at edges and corners, where the continuum effect concentrates pressure.
What Is a Compensated Raft?
A raft founded deep enough that the weight of excavated soil offsets part of the structure’s weight.
The net pressure applied to the ground is the gross pressure minus the overburden removed. Where a basement is deep and the structure relatively light, that difference can approach zero, and a fully compensated raft applies no net increase in stress at founding level. Settlement then becomes a question of heave and recompression rather than of virgin compression, and the magnitudes are very different.
Two cautions travel with it. The heave occurring during excavation has to be accounted for, since the ground rebounds before the structure is built and recompresses afterwards. And the compensation only holds for the loading it was calculated with; a later change of use that adds weight removes the compensation the design relied on.
When Does a Raft Become a Piled Raft?
When settlement rather than capacity requires help.
In a piled raft, the slab and the piles share the load, and the piles are used to control settlement rather than to carry everything. A smaller number of piles placed where the settlement profile is worst, usually towards the centre, can reduce the dished curvature more efficiently than piling the entire footprint.
The analysis is more demanding, because the slab, the piles and the soil interact and the load share depends on the relative stiffness of all three. Where the design intent is settlement control rather than capacity, that interaction is the whole point and cannot be avoided by assuming the piles take everything.
Common Errors in Raft Foundation Design
| Error | Consequence |
|---|---|
| Plate load test modulus applied without size correction | Slab stiffness overestimated, moments understated |
| One subgrade modulus used across the whole raft | Dished settlement not reproduced, differential movement missed |
| Settlement depth of influence taken as for an isolated footing | Deep compressible layer omitted from the calculation |
| Structural and geotechnical settlements never reconciled | Springs and settlement analysis describe different foundations |
| Uniform contact pressure assumed for the structural design | Edge and corner concentrations missed |
| Excavation heave ignored in a compensated raft | Recompression settlement unaccounted for |
| Piles in a piled raft assumed to carry the full load | Load share and settlement interaction not represented |
Where SETAF2018 Fits
SETAF2018 designs raft foundations within the same model that carries the ground data, the other foundation types and the reporting.
The raft is defined as a rectangular, circular or polygonal element which can include an unlimited number of openings, placed on the layered soil profile built from the project boreholes. Bearing capacity follows Terzaghi, Meyerhof and Vesić with the bearing capacity factors and correction factors each defines, and SPT-based methods are available in granular soils.
Settlement is calculated through the profile by the Boussinesq stress distribution for shallow foundations, combining elastic and one-dimensional consolidation components with Terzaghi correction for the construction period, so the depth of influence follows the raft dimensions rather than a fixed multiple. The modulus of subgrade reaction is then derived from that settlement, from the Vesić expression or from a plate load test, which means the springs handed to the structural engineer come from the settlement analysis rather than from a separately chosen value.
Where the raft is supported on piles, micropiles, deep mixing or jet grout columns, those elements are placed in the same model and the improved zone is re-layered with composite properties, so a raft, a piled raft and a shallow alternative are compared on one profile. Calculation reports print the equations, the layer contributions and the intermediate values behind both the settlement and the derived modulus, within the wider foundation analysis workflow.
What the software does not do is solve the slab itself as a structural element. The spring stiffnesses, contact pressures and settlement field are produced for the structural model rather than replacing it.
Frequently Asked Questions
What is a raft foundation?
It is a single continuous slab carrying all the columns and walls of a structure and spreading their loads over the whole footprint. It is also called a mat foundation, and it is used where individual footings would cover much of the plan area, where the soil is weak, or where variable ground would cause differential settlement.
Does bearing capacity or settlement govern raft design?
Settlement, in nearly every case. The large loaded area keeps the applied pressure low and raises the bearing capacity, but it also extends the stress increase deep into the profile, so compressible layers that would not matter under an isolated footing contribute to the settlement of a raft.
What is the modulus of subgrade reaction used for?
It is the stiffness of the springs that represent the soil in the structural model of the slab, defined as contact pressure divided by settlement. The structural analysis solves the raft on that spring bed to obtain contact pressures, bending moments and deflections.
Why is a single subgrade modulus not sufficient?
Because the value depends on the loaded area and is not constant across the raft. Independent springs cannot transfer stress sideways as real soil does, so a uniform spring bed fails to reproduce the dished settlement profile. Edge and corner springs are normally stiffened and central springs softened to recover it.
Can a plate load test value be used directly?
No. A plate load test gives the stiffness for the plate size tested, and a raft stresses vastly more soil. Applying the uncorrected value to a large raft overestimates the stiffness and understates both differential settlement and the bending moments derived from it.
What is a compensated raft foundation?
A raft founded deep enough that the weight of excavated soil offsets part of the structure’s weight, reducing the net pressure increase at founding level. Where the offset is complete, settlement becomes a matter of heave and recompression rather than virgin compression.
When should a piled raft be used instead of a raft?
When settlement rather than capacity needs reducing. Piles placed where the settlement profile is worst, usually towards the centre, reduce the dished curvature more efficiently than piling the whole footprint, with the slab and piles sharing the load rather than the piles carrying everything.
How can the spring stiffnesses be checked?
By comparing the settlement the structural model produces with the settlement the geotechnical analysis predicts. If the two disagree, the springs are wrong, and the bending moments derived from them are wrong in the same proportion.
