Ground Improvement Techniques: A Practical Engineering Guide

A client calls with a warehouse site on 6 m of soft clay. The structural engineer has already sized the columns, the schedule is tight, and someone has suggested piles. Before you price 200 piles, ask a simpler question: does the soil have to carry the load as it is, or can we make the soil better?

That question is the whole of ground improvement. Instead of bypassing a weak layer with a deep foundation, you modify the layer itself. Sometimes that means squeezing the voids out of a loose sand. Sometimes it means installing columns of stronger material and letting the soil arch between them. Sometimes it means mixing the soil with a binder until it behaves like a weak rock.

This guide covers the techniques you are most likely to encounter, how each one actually transfers load, and how to narrow the options down for a specific site. The design values quoted here follow the framework of the FHWA ground modification reference manuals (FHWA NHI-16-027 and NHI-16-028), which remain the most complete open reference in English.

What ground improvement can and cannot do

Ground improvement changes the engineering properties of the ground: stiffness, strength, drainage, or all three at once. What it does not do is create capacity where there is no competent material within reach. If the weak layer is 30 m thick and the structure is a 20-storey tower, you are looking at piled foundations, and no amount of clever treatment will change that.

It tends to win in a specific situation:

• The problematic soil is a relatively shallow layer, often 3 to 10 m thick.
• The structure is settlement-sensitive but not extremely heavy: warehouses, embankments, industrial slabs, tank farms, low and mid-rise buildings.
• Piles would work, but the cost, the noise, the vibration, or the construction program makes them unattractive.
• The site has fill, soft alluvium, loose sands, or organic material that has to be made serviceable.

In practice the choice is between improving the ground, going to a deep foundation, or widening the shallow foundation footprint. Ground improvement competes with both, and sometimes it loses.

The five mechanisms behind every technique

Every technique in this guide works through one or more of five mechanisms. Knowing which mechanism you need is most of the selection problem, because the mechanism is dictated by the soil you have.

Load transfer is worth separating out. Techniques like rigid inclusions and geosynthetic-reinforced platforms do not strengthen the soil much at all. They redistribute load so that the soil that is already competent carries more of it.

Densification: dynamic compaction, rapid impact, and vibro

Densification only works where the soil has voids worth collapsing. Loose sands, gravels, and non-plastic fills respond well. Saturated clays and silts do not, because water cannot drain fast enough during the loading pulse and the energy goes into pore pressure instead of rearrangement.

Dynamic compaction

A crane lifts a heavy tamper, typically 6 to 20 t, and drops it from 12 to 22 m in a grid pattern, giving an energy per blow of roughly 1,000 to 3,000 kN·m. The energy per blow, not the mass or the height alone, is the design variable. The depth of influence follows Menard’s empirical relationship:

$D = n \sqrt{W H}$

where $W$ is the tamper weight, $H$ is the drop height, and $n$ is an empirical coefficient that typically falls between 0.35 and 0.6 depending on the soil type. For a 15 t tamper dropped 20 m, the depth of influence lands in the order of 6 to 10 m.

The sequence is usually a primary pass at high energy on a coarse grid, followed by an ironing pass at low energy on an offset grid to treat the surface between the print positions. Each pass is followed by levelling.

The limitations are practical. It generates strong ground vibration, so it is unsuitable near sensitive structures or buried services. It performs poorly below the water table in saturated fine soils, and it needs a site large enough to run the crane grid.

Rapid impact compaction

Rapid impact compaction uses a hydraulic hammer mounted on an excavator, dropping a smaller mass at a much higher rate. The energy per blow is lower and the depth of influence is correspondingly shallower, typically 3 to 6 m rather than the 6 to 10 m of dynamic compaction, but the output is high and the plant is far more manoeuvrable. It suits smaller areas, road widening, and sites where a crane grid is impractical.

Vibro compaction

A depth vibrator penetrates the ground and vibrates it laterally, causing loose granular soil to rearrange into a denser state. The probe is withdrawn in stages. In clean sands the improvement is substantial and can be verified directly with pre- and post-treatment cone penetration tests.

The fines content decides whether the technique applies at all. Densification generally cannot be achieved where the granular soil contains more than about 15 percent silt, or more than about 2 percent clay; above those limits the soil holds the probe open and the vibration does not rearrange the particles. Past that threshold, vibro replacement becomes the better choice.

Stone columns and vibro replacement

Stone columns are columns of compacted granular material installed in soft cohesive or mixed ground, usually with a depth vibrator or by ramming. The soil around them is displaced rather than excavated, and the columns act as reinforcement.

The load is shared between the column and the surrounding soil, and the column takes more than its share of the load because it is stiffer. That sharing is quantified with the stress concentration ratio:

$n = \frac{q_c}{q_s}$

where $q_c$ is the vertical stress on the column and $q_s$ is the stress on the surrounding soil. For stone columns $n$ typically falls in the range of 2 to 5, depending on the relative stiffness of the column material and the soil, and on the length of the column relative to the loaded width.

Two effects follow, and they are not the same effect. Bearing capacity rises because the composite material is stronger than the clay alone. Settlement of the composite is smaller because the composite is stiffer. The granular columns also act as drains, but drainage does not reduce the magnitude of consolidation settlement; it makes that settlement happen faster, because the drainage path is shortened when columns are installed at 2 to 3 m centres.

The main constraints are the presence of very soft or organic layers that cannot support the column laterally, and the need for a competent crust to build the load transfer platform on. As a working rule, undrained shear strengths below roughly 15 kPa are a warning sign that lateral confinement will be inadequate.

Rigid inclusions and aggregate piers

These two are frequently confused, and they behave differently.

Rigid inclusions, sometimes called controlled modulus columns, are stiff elements of grout or concrete installed by displacement or drilling. They are not designed as piles. The load is transferred to them partly through a load transfer platform of granular fill, sometimes reinforced with geogrid, which develops arching between column heads. The columns then carry load to a deeper, stiffer stratum by a combination of tip bearing and shaft friction.

The arching that makes the system work depends on the geometry of the load transfer platform. As a design guide, the platform thickness needed to develop a stable arch is at least half the clear span between column heads, which is why platform thickness and column spacing are designed together rather than separately.

Aggregate piers are installed by drilling and then compacting aggregate in thin lifts with high energy, which both densifies the aggregate and prestresses the surrounding soil laterally. They work well in mixed profiles and are typically shorter than rigid inclusions, since they rely on lateral confinement rather than on reach.

Grouting and mixing: jet grouting and deep soil mixing

When the soil has to be improved in place without displacing it, and the geometry is tight, cementation techniques come into play. Both of the techniques below are covered by dedicated execution standards.

Jet grouting

A high-pressure jet of grout erodes the soil while mixing it, and the excess spoil returns to the surface. The result is a column of soil-cement, often called soilcrete, whose diameter depends on the soil type, the jetting energy, and the number of jets. Single, double, and triple fluid systems give increasing diameter and control. Unconfined compressive strengths typically land between about 1 and 5 MPa.

Jet grouting can be used across a wide range of soils, from soft clays and silts to sands and gravels, and it can be installed at moderate inclinations, typically up to about 30 to 45 degrees from vertical, which makes it useful for underpinning, cut-off walls, and irregular geometries. Deep soil mixing, by contrast, is installed vertically.

Two practical caveats apply. Stiff, high-plasticity clays and boulders reduce the achievable diameter and the homogeneity of the column, and boulders can deflect the jet, so those ground conditions need site-specific trials rather than a design assumption. And because the properties depend on the installation process, the execution standard requires a production field test to confirm the actual column size, shape, verticality, homogeneity, and strength.

Execution is covered by EN 12716:2018, Execution of special geotechnical work — Jet grouting.

Deep soil mixing

A mixing tool advances into the ground while injecting binder, producing panels or columns of improved soil. Unconfined compressive strengths are typically lower than jet grouting, commonly in the range of about 0.5 to 2 MPa. It is more economical than jet grouting where a large treated volume is needed and the soil is workable, but it is not applicable in very dense, very stiff, or obstructed ground.

Execution is covered by EN 14679:2005, Execution of special geotechnical works — Deep mixing.

Both techniques are sensitive to the quality of the mixing, which is why verification relies on coring and on the strength of trial columns rather than on theoretical predictions.

Accelerating consolidation: drains and surcharge

When the problem is not strength but time, the answer is often drainage rather than reinforcement. A soft clay that would take 30 years to settle under its embankment can be brought down to a year or less by installing prefabricated vertical drains and applying a surcharge.

The governing relationship is straightforward. Consolidation time scales with the square of the drainage path:

$t = \frac{T_v \, H^2}{c_v}$

where $t$ is time, $T_v$ is the time factor for the required degree of consolidation, $H$ is the maximum drainage distance, and $c_v$ is the coefficient of consolidation. The definition of $H$ is where most quick estimates go wrong: it is the longest distance the water has to travel, which for a layer drained on both faces is half the layer thickness, not the full thickness.

Take a 6 m clay layer with drainage at top and bottom, so $H = 3$ m. Drains that reduce the maximum drainage distance to 1.5 m cut $H$ by a factor of two and the time by a factor of four. The same arithmetic applied to the full 6 m thickness would wrongly suggest a factor of sixteen.

Under real drain spacing the consolidation is dominated by radial flow to the drains rather than vertical flow to the boundaries, which is why the design is governed by the coefficient of consolidation for horizontal flow, $c_h$. Standard practice is covered by EN 15237:2007, Execution of special geotechnical works — Vertical drainage.

Vacuum consolidation applies the same logic with a vacuum applied under an impermeable membrane, which avoids the need for a large surcharge fill. The technique is effective in very soft clays where even a modest embankment would fail.

Geosynthetics: reinforcement without columns

Geosynthetics rarely appear as a headline technique, but they are present inside several of the others. A geogrid-reinforced load transfer platform is what makes rigid inclusions work over a large area. Basal reinforcement at the base of an embankment allows it to be built over ground that would otherwise fail, by resisting the outward spreading of the fill and forcing the failure surface deeper.

The design logic is the same in both cases: the geosynthetic carries tensile load that the soil cannot, and it transfers that load to the anchors or to the soil outside the critical zone.

Comparing the techniques

The following table summarises the techniques on the selection criteria that usually decide the matter: mechanism, soil suitability, depth of influence, and relative cost. Cost is relative to other ground improvement options on the same site, not absolute.

Two entries need a caveat. Dynamic compaction looks cheap per square metre until you include the trial area, the vibration monitoring, and the repeated passes. Jet grouting looks expensive until you compare it with the alternative of underpinning a structure that has already moved.

How to select a technique

Selection follows a fixed sequence, and the order matters.

1. Define the performance requirement. Allowable total settlement, allowable differential settlement, bearing capacity, and time available before the structure is loaded. Without a number, there is no way to judge whether an option is adequate.
2. Characterise the soil. Thickness and depth of the weak layer, undrained shear strength or relative density, fines content, organic content, and groundwater conditions. The fines content alone often decides between vibro compaction and vibro replacement.
3. Shortlist by soil type. Sands and gravels point to densification. Soft clays point to reinforcement, drainage, or replacement. Mixed and obstructed ground points to cementation techniques.
4. Check the depth. The weak layer has to be within the reach of the technique at a reasonable cost. If it is not, the option is a deep foundation.
5. Check constructability. Access, headroom, vibration limits, spoil disposal, and the construction programme. This step eliminates more options than engineers expect.
6. Compare cost and risk. Include the trial area, the verification testing, and the contingency for a technique whose performance cannot be predicted with the same confidence as a pile capacity.

A short worked example

Consider a single-storey warehouse on 6 m of soft clay over dense sand, with an allowable settlement of 25 mm and a column load of 900 kN. The inputs are:

The average stress increment through the clay, using the 2:1 spread, comes to roughly 28 kPa. The untreated consolidation settlement follows the classic expression:

$s = \frac{C_c}{1+e_0} H \log_{10}\left(\frac{\sigma'_{v0} + \Delta\sigma}{\sigma'_{v0}}\right)$

Substituting: $s = 0.25 \times 6 \times \log_{10}((43+28)/43) = 1.5 \times 0.219 \approx 0.33$ m, so about 330 mm untreated. That is more than thirteen times the allowance, so something has to change.

At 2.0 m centres the tributary area per column is 4.0 m², and a 0.9 m column occupies 0.64 m², a replacement ratio of 16 percent. With a column-to-soil modulus ratio of about 10, the settlement improvement factor is:

$\frac{s_{untreated}}{s_{composite}} \approx 1 + a_s\left(\frac{E_c}{E_s} - 1\right) = 1 + 0.16 \times 9 = 2.4$

The composite settlement is therefore in the order of 330/2.4 ≈ 140 mm as a floating system. Still nowhere near 25 mm.

The reason is that the columns in that first check were floating in the clay. Extending them to bear about 2 m into the dense sand, so each column is 9 m long and picks up load at the tip, changes the settlement mechanism entirely: the compressible clay no longer carries the load, and the settlement is governed by the stiffness of the dense sand and the elastic shortening of the columns, which lands in the order of 20 to 25 mm. That passes, but it passes as a load-bearing system rather than as a reinforced composite, and at 9 m of column it now has to be compared honestly against piles on cost.

The same reasoning applies well beyond this example. Ground improvement on soft clay often works by reaching past the clay rather than by reinforcing it, and where the technique cannot reach competent ground with a tight settlement allowance, no realistic column spacing will save the scheme.

What to verify before and during construction

Ground improvement is less predictable than a pile, because the properties of the improved ground depend on the installation process. Design values are estimates until a trial confirms them.

• Trial area first. Install a small area, then test it. Plug the results back into the design before committing to the full site.
• Use the right test. Cone penetration tests before and after for densification. Plate load tests for the composite. Coring and unconfined compressive strength for cementation techniques. Settlement monitoring for drainage schemes.
• Set numeric acceptance criteria in advance. A required increase in cone resistance, a plate modulus, or a minimum unconfined compressive strength, with a stated number of tests per treated area.
• Monitor during installation. Energy per blow, grout volume and pressure, mixing time, and column position all need records. A jet grouting column that produced half the expected spoil volume is not the column that was designed.
• Watch the interaction with adjacent structures. Vibration, heave, and lateral displacement are the usual causes of damage claims on these projects.

Frequently asked questions

Is ground improvement cheaper than piles? Sometimes, and it depends on depth. For a shallow weak layer, treatment is usually cheaper. For a thick weak layer, the treated depth needed to control settlement becomes large and piles win. The comparison has to include the trial area and the verification testing.

Can stone columns be used in organic soils? Rarely on their own. Peat and organic clays offer very little lateral confinement, so the columns bulge rather than carry load. They may still work as drains when combined with a surcharge, but that is a different design.

Does vibro compaction work below the water table? Yes in clean sands, where drainage is fast enough for the vibration to rearrange the particles. No in soils above roughly 15 percent silt or 2 percent clay, where the pore pressure cannot dissipate during the loading pulse and the vibrator cannot densify the ground.

How long does dynamic compaction take? The installation itself is fast, but the method needs repeated passes with levelling in between and a pause for pore pressures to dissipate. The programme is dominated by the trial area and the verification testing, not by the dropping of the weight.

What is the difference between jet grouting and deep soil mixing? Jet grouting erodes and replaces the soil with a jet of grout, producing a high-strength column and a spoil stream that must be managed. Deep soil mixing mechanically blends the soil with binder in place, produces less spoil, and is usually cheaper over large areas, but it is installed vertically and struggles in stiff or obstructed ground.

Can ground improvement replace a pile foundation on a tall building? Usually not. The stiffness of the improved ground is still far below that of a pile founded on rock, so the settlement control that tall buildings need is not achievable. Ground improvement is a shallow-to-moderate depth solution.

Where to go next

If your site is dominated by shallow layers, the next step is usually the foundation itself. Start with gross and net bearing pressure to make sure you are reading the geotechnical report correctly, and work through the step-by-step retaining wall design guide if the structure retains soil as well as supports it. For the software side of the workflow, the geotechnical software comparison guide covers what different tools are actually good at.

You can also run the numbers directly in the browser: the shallow foundation calculator handles bearing capacity and settlement for the treated-as-shallow case, and the deep foundation tools are there when the comparison goes the other way.