Pavement Layers, Subgrade and Soil-Supported Structures

A pavement is a stress-spreading stack

A wheel load applied to a surface spreads outward with depth, so the pressure reaching any level is lower than the contact pressure above it and lower still further down. A pavement exists to spread it far enough that what arrives at the SUBGRADE is within what the subgrade can take repeatedly without deforming.

That makes the subgrade the controlling input. A strong one needs little spreading and a thin pavement; a weak one needs a great deal and a thick pavement, for exactly the same traffic. Two roads carrying identical loads in the same town can differ substantially in construction depth because their ground differs, and no amount of traffic data explains that difference.

The layers are also not interchangeable. Each is assigned a coefficient describing how much structural work a unit of its thickness does, and the total is a weighted sum — so a given structural requirement can be met by more of a cheap granular layer or less of an expensive bound one, and the choice is an economic one within the same structural answer.

Drainage sits underneath all of it. A granular layer that saturates loses much of its contribution, and a subgrade that saturates loses strength directly, so the drainage design is a structural element rather than a separate discipline. Most premature pavement failure is a water story before it is a loading one.

Every measure of subgrade strength is an index, and one needs correcting

Subgrade strength is not measured directly. It is expressed through indices — a bearing ratio against a standard crushed stone, a modulus from a repeated-load test, a blow count from a driven sampler — and each is a proxy with its own conditions.

The standard penetration test needs the most care, because its raw result is systematically misleading. A blow count depends on the confining pressure at the depth it was taken, so the SAME soil at the same density reads higher deeper down. Comparing an N-value from ten metres with one from two metres without correcting for overburden compares depth as much as density.

There is a second correction that matters more than most people realise. The test's energy delivery varies with the hammer and release mechanism used, by large factors between rig types, so results are normalised to a standard energy ratio before anything is done with them. A borehole log that does not state its energy ratio is missing an input.

And the corrections do not commute with the correlations. The published relationships between blow count and friction angle, density or bearing capacity are written for a particular corrected form, and feeding a raw or differently-corrected value into them produces an answer that is wrong in a direction the number itself does not reveal. The pages here state which form they expect.

Stabilisation changes the material, and only for the right material

Adding a binder to a subgrade is not the same as adding a layer. It changes what the soil IS, and which binder works depends on what the soil is made of.

Lime works on plastic clays. Calcium ions exchange onto the clay minerals, which flocculates them and immediately reduces plasticity and moisture sensitivity; a slower pozzolanic reaction then forms cementitious compounds and raises strength over weeks. On a non-plastic granular soil there are no clay minerals to exchange with and lime does essentially nothing.

Cement works by hydrating, as it does in concrete, and suits granular and low-plasticity soils — the reverse case. Choosing the wrong binder for the soil is not a matter of a smaller improvement; it is a matter of no improvement at all.

The dose is determined by test rather than by rule of thumb. The initial consumption of lime is found by raising the dose until the mixture's pH stabilises, which is the point at which the exchange demand is satisfied, and the structural dose is then set above it and confirmed by strength testing. A percentage taken from a table is a starting point for a trial, not a specification.

Two field conditions govern whether any of it works. Mixing has to be uniform through the full treated depth, and the material has to be compacted within a limited window after mixing, because a cement-stabilised layer that stiffens before it is compacted cannot then be compacted at all.

Air voids in asphalt: a narrow band, and the best predictor of life

Asphalt compaction is not specified against a laboratory maximum dry density in the way soil is. It is specified against the mixture's MAXIMUM THEORETICAL density — the density it would have with no air in it at all — and the quantity controlled is the percentage of air voids remaining.

The acceptable band is narrow, and it is bounded on both sides for different reasons. Too many voids and the mat is permeable: water and air reach the binder, it oxidises and hardens, and the surface ravels and cracks years early. Too few and there is no room for the binder to expand in hot weather, so the mix flushes to the surface and ruts under traffic.

The relationship with life is steep and well documented: a mat left a few percentage points above its target void content can lose a large fraction of its service life, and no maintenance recovers it. This makes in-place density the single most predictive acceptance test on a paving job, and the reason rolling patterns, mat temperature and the time available to compact are controlled so tightly.

Temperature is the constraint that ties it together. Asphalt can only be compacted while it is hot enough to move, so the cessation temperature sets a hard time limit after placement — which is why a cold day, a long haul, a thin lift or a breakdown in the rolling train all show up later as a density failure rather than as an immediate problem.

On a rigid pavement the edge is the worst place, not the middle

A concrete pavement is a slab on an elastic foundation, and the stress a wheel load produces depends on WHERE on the slab it sits. A load in the interior is supported all round; a load at a free edge has support on one side only; a load at a corner has least of all.

The classical solutions make the ranking explicit: for the same load and slab, edge stress exceeds interior stress substantially, and corner conditions are more severe again. Cracking on concrete pavements consequently starts at edges and corners rather than in the middle of a panel, which is where intuition puts it.

This is why edge support is worth more than thickness. A tied concrete shoulder, a widened lane that moves the wheel path away from the free edge, or a thickened edge all reduce the governing stress more efficiently than adding depth across the whole slab — and a widened lane does it by geometry alone.

Load transfer across joints does the same job at the other free edges. Dowels carry shear from one slab to the next so that a wheel at a joint is not loading a free edge at all; without them, every joint is two free edges facing each other. That is what dowel bars are for, and it is why their alignment matters — a misaligned dowel locks the joint instead of transferring across it.

Airfield pavements are the same analysis with harsher inputs: far heavier wheel loads, multi-wheel gear whose footprints interact, and higher tyre pressures, all applied to slabs whose failure consequences are different. The method is recognisable; the numbers are not transferable.

A flexible culvert is a soil structure

A large corrugated pipe has very little bending strength of its own. Loaded from above it tries to flatten — and as it flattens it pushes outward at its sides into the backfill, which pushes back. That lateral resistance is what carries the load.

So the design is really the BACKFILL. The stiffness of the material beside the pipe, and how well it is compacted, dominate the prediction of how much the pipe will deflect; the pipe's own stiffness is a smaller term. A perfectly manufactured culvert in poorly compacted haunch material will deflect and eventually buckle, and the pipe was never the problem.

The haunch — the zone under the springline, beside the bottom of the pipe — is the hardest place on a site to compact and the most important. It is confined, awkward to reach with a roller, and it is where the lateral support starts.

Deflection is therefore the acceptance criterion rather than a strength check, and it is measured after backfilling and again later, because part of it develops over time as the soil consolidates. A limit expressed as a percentage of diameter is a serviceability criterion with a real margin to buckling behind it, which is why exceeding it is treated as a defect to be remedied rather than a number to be argued about.

Separation, filtration, and what has to be tested

A geotextile between a granular layer and a soft subgrade has two jobs that are easy to confuse. SEPARATION stops the aggregate being pushed down into the soil and the fines being pumped up into the aggregate — which is what destroys a granular layer's strength over time without anything visibly failing. FILTRATION lets water through while retaining soil particles, which is a pore-size question about the fabric against a grading question about the soil.

The overlap at fabric joints is set by how weak the subgrade is, and the relationship is the reverse of intuitive: a WEAKER subgrade needs a WIDER overlap, because the fabric moves more during placement and the joint has more opportunity to open. On the softest ground the sheets are sewn rather than overlapped.

Riprap over a filter is the same system in coarse form, and it fails the same way: stone sized correctly on a filter that is wrong lets the underlying soil wash out through the voids, and the armour subsides into the hole it created. The filter criteria — a graded layer or a geotextile, sized against both the soil below and the stone above — are what keep it in place.

Everything on this page is an estimate that a test supersedes. Subgrade strength comes from a CBR, a plate load or a laboratory modulus; asphalt density from cores or a calibrated gauge; stabilisation dose from a pH and strength series; culvert deflection from a measured survey after backfill. The calculators size and screen, and every page says which test produces the number that governs.