Deep Foundations, Piling & Underpass Construction Explained

Finding out what the load is actually going to sit on

A subsurface investigation puts boreholes, or occasionally cone penetration test (CPT) soundings, at spacing and depth set by the structure's footprint and loading, not by convenient site access. Standard Penetration Test (SPT) N-values, undrained shear strength from triaxial or vane shear tests, groundwater level, and — critically for anything near a railway or existing structure — the presence of made ground, fill, or obstructions from earlier construction all come out of this phase. For an underpass corridor, boreholes go in along the full alignment, not just at the two ends, because a single soft or obstructed pocket midway through a box-jacking drive is the difference between a routine push and a stalled one.

This is also where the foundation design basis gets set: the design loads (dead, live, wind, seismic, and for railway-adjacent work, dynamic train loading), the allowable settlement, and the governing limit states — ultimate bearing capacity and settlement, checked separately, because a pile can satisfy one and badly fail the other. Negative skin friction — where settling fill or soft clay around the pile drags downward on the shaft instead of supporting it — gets identified here if the site has recent fill or a compressible layer, because it changes both the pile's usable capacity and its structural design, and finding it out after piling starts is an expensive way to learn.

Why the structure needs to go deep at all — and which kind of deep

A shallow foundation — an isolated footing, strip footing, or raft — works when competent, adequately strong soil sits close enough to the surface to carry the load within an acceptable settlement. Deep foundations become necessary when that competent layer is too far down, when loads are too concentrated for a raft to spread economically, when the site has weak or compressible soil near the surface, or when uplift, lateral, or dynamic loads (wind on a tall structure, seismic action, a bridge pier in a scour-prone riverbed) need more than a footing can offer. A pile transfers load by end bearing (resistance at the toe, in strong soil or rock), by skin friction (shear resistance along the shaft), or almost always by some combination of both — and which one dominates drives the entire pile type decision that follows.

From there, the choice is really about installation method matched to ground conditions and site constraints. Driven piles — precast concrete, steel H-piles, or steel pipe piles — displace or shear through the soil as they're driven, densifying loose granular soil around them, but the vibration and noise make them a poor fit close to existing structures or in an urban/live-corridor setting. Bored cast-in-situ piles replace soil by excavation before concreting, which avoids most of that vibration but depends entirely on shaft-wall stability during the bore. Continuous Flight Auger (CFA) piles combine boring and concreting into one continuous operation and are fast and low-vibration, but leave far less room to inspect the ground actually encountered. Micropiles and helical/screw piles earn their place in restricted access, low headroom, or underpinning work where nothing larger can physically get in.

Getting the pile to refusal, not just to depth

Precast concrete, steel H-section, or steel pipe piles are driven with an impact hammer (diesel, hydraulic, or air) or, in cohesionless soils, sometimes vibrated in with a vibratory hammer for faster, lower-noise installation before final seating with an impact hammer. Driving is tracked against a dynamic formula or, on larger projects, a wave equation analysis (GRLWEAP) predicting blow count versus depth — but the real acceptance criterion in the field is a set (the pile's penetration per blow) at final drive, checked against the design set from a driving formula or, better, from a dynamic load test on a trial pile. A pile that reaches design depth without reaching the required set hasn't reached refusal; it's just stopped moving for now, and needs re-driving or a design review, not sign-off.

Pile driving analyzer (PDA) instrumentation — strain gauges and accelerometers near the pile head — measures force and velocity during driving on selected piles, feeding a CAPWAP analysis that estimates static capacity and checks for pile damage from over-driving, both things a blow count alone can't tell you. Splicing is common on longer piles (welded for steel, mechanical or wet-jointed for precast concrete) and each splice is a point the driving record and the visual inspection both need to specifically confirm, because a bad splice fails invisibly, underground, exactly where nobody can see it again.

Holding a hole open long enough to fill it properly

A rotary bored pile is excavated with a rotary rig using a bucket, auger, or (in hard ground or rock) a core barrel, with the shaft supported either by temporary steel casing or by bentonite/polymer drilling slurry that maintains hydrostatic pressure against the borehole wall until concrete replaces it. The base is cleaned — critically — before the reinforcement cage goes in, because sediment or loose material left at the toe becomes a soft, compressible layer directly under the point meant to carry the most load. Concrete is placed by tremie pipe from the bottom up, displacing slurry or water upward and out rather than letting concrete free-fall through it, which would segregate the mix and trap slurry inside the shaft.

Continuous Flight Auger (CFA) piling drills and concretes in one continuous operation: the auger bores to depth, then concrete is pumped through the hollow stem under pressure as the auger is withdrawn, with the reinforcement cage pushed or vibrated into the wet concrete afterward. It's fast and low-vibration, but the entire process happens without the engineer ever seeing the excavated shaft — so real-time monitoring of auger torque, rotation, penetration rate, and concrete pressure/volume against depth (an automated monitoring system on the rig) is the only real-time check that the shaft is continuous and hasn't necked or been contaminated by collapsing soil.

What to build when the site rules out the standard answer

Micropiles (drilled, small-diameter, typically 100–300mm, reinforced with a steel bar or casing and pressure-grouted) go where headroom, access, or proximity to sensitive existing structures rules out a conventional rig — inside basements, under low bridges, for underpinning an existing foundation while it stays in service. Helical (screw) piles are mechanically screwed into the ground on their own helical plates, installed and load-tested essentially in the same operation, with torque during installation correlating directly to ultimate capacity — genuinely fast for light-to-medium structures and temporary works, but limited in the loads and ground conditions they suit. Jet grouting uses high-pressure fluid jets to erode and mix soil in place with cementitious grout, forming a soil-cement column — used less as a primary pile type and more to strengthen ground locally, seal a cut-off, or underpin foundations from a tight access point.

The genuinely modern shift across all of these is instrumentation replacing judgement calls: real-time drilling-parameter recording (torque, thrust, penetration rate) on micropile and CFA rigs, automated monitoring on jet-grouting columns, and installation torque logging on helical piles all now generate a continuous digital record per pile rather than a handwritten log — which matters enormously on underpinning or restricted-access work, where a problem caught during installation is a design conversation, and the same problem caught after cladding goes back on is a demolition conversation.

Proving the pile actually does what the calculation says it does

A static load test — applying a known load through a hydraulic jack against a reaction system (kentledge, anchor piles, or a reaction beam) and measuring settlement — remains the most direct proof of capacity, run either to a working load multiple (commonly 1.5–2.5× design load) or, less often now, to failure. An Osterberg cell (O-cell) test embeds a hydraulic jack inside the pile shaft itself, loading the pile against its own end bearing and shaft friction simultaneously in opposite directions, which avoids the need for a large surface reaction system entirely and is increasingly the practical choice for large-diameter or deep piles. Dynamic load testing (PDA, analyzed via CAPWAP) estimates capacity from the pile's response to a hammer blow — far faster and cheaper than a static test, and standard practice for verifying a sample of production driven piles against the trial-pile design.

Integrity testing is a different question entirely — not "how much load can it carry" but "is the shaft actually sound." Low-strain Pile Integrity Testing (PIT), striking the pile head with a small hammer and reading the reflected stress wave, is fast enough to run on essentially every cast pile and picks up major defects — necking, voids, cracks — though it can miss subtle ones. Cross-hole sonic logging (CSL), used on larger bored piles with pre-installed access tubes, sends an ultrasonic signal between tube pairs down the shaft and gives a far more detailed defect profile, at the cost of needing tubes cast into the pile from day one — which means the decision to CSL-test a pile has to be made before concreting, not after.

Tying individual piles into one structural unit

Pile heads are broken down (chipped back on driven/bored concrete piles, or cut to level on steel piles) to expose sound concrete or a clean steel section for the cap connection — a step that matters because the top of any cast pile is where laitance, contamination, and any minor construction defects concentrate. The pile cap or grade beam then ties the pile group together structurally, distributing the column or wall load across the group rather than relying on any single pile perfectly matching its neighbours. Reinforcement detailing at this connection — dowel bars from pile into cap, adequate embedment length, and cap thickness for punching shear around each pile — is where a structural engineer's cap design and a geotechnical engineer's pile design actually meet, and it's worth both disciplines reviewing the same drawing before it's built.

The decision that defines everything after this point — and who has to approve it

Cut-and-cover means exactly what it says: excavate an open trench, build the underpass structure inside it, then backfill and reinstate the surface. It's the most straightforward method structurally and usually the cheapest per metre, but it requires the surface above to be closed or heavily diverted for the duration — workable where a road can be detoured, close to impossible where a railway line can't simply stop running.

Box pushing / jacking — sometimes called the box-pushing technique (BPT) or pipe-push technique in railway engineering — builds the finished RCC box structure in a jacking pit alongside the live corridor, then hydraulically pushes it through the embankment beneath the road or track while traffic continues to run on top, with the track or pavement itself carried on a temporary load-transfer structure during the push. It's slower and more specialised than cut-and-cover, but it's the standard method for eliminating level crossings and building road-under-bridges (RUBs) beneath live railway lines precisely because it avoids a traffic or train possession for anything beyond short, scheduled blocks.

Microtunneling / pipe jacking bores a tunnel using a remotely steered microtunnel boring machine (MTBM) from a launch shaft to a reception shaft, jacking pipe or box sections in behind it — no open trench, minimal settlement when properly controlled, but limited mostly to circular or near-circular cross-sections and best suited to utility crossings, pedestrian subways, or smaller-diameter road underpasses rather than a full multi-lane carriageway box.

Holding the excavation walls up before the permanent structure can

A diaphragm wall — excavated in slurry-supported panels and concreted by tremie, much like an oversized bored pile turned into a continuous wall — gives the highest stiffness and water-tightness of the common options, and can double as a permanent structural wall, not just temporary support, which often justifies its higher cost on deep or long-duration excavations. Secant and tangent pile walls — overlapping (secant) or touching (tangent) bored piles forming a wall — are faster and cheaper to install than a diaphragm wall and secant walls give reasonable water-tightness through the overlap, though rarely as good as a properly built diaphragm wall. Sheet piling — interlocking steel sections driven or vibrated in — is the fastest and most economical option in soils that allow driving, but brings back the vibration and noise concerns from Phase 3, and offers materially less stiffness against wall deflection than either concrete option.

Whichever wall type, the excavation itself is braced as it deepens — internal struts, ground anchors (where permitted; anchors crossing under a live railway or adjacent property line often aren't), or a top-down construction sequence where permanent floor slabs are cast and used as bracing as excavation proceeds beneath them. Wall deflection and ground settlement behind the wall are predicted at design stage (typically by finite-element modelling) and then checked against real inclinometer and settlement-marker readings throughout excavation — the comparison between predicted and observed behaviour, known as the observational method, is standard practice on any excavation adjacent to sensitive structures or live infrastructure.

Building the box in the open, then closing the ground back over it

With the retaining walls from Phase 9 in place, excavation proceeds in stages down to formation level, with dewatering (Phase 13) running alongside as needed. The underpass structure itself — usually a single or multi-cell reinforced concrete box, occasionally with precast segments for speed — is cast against the retaining walls or founded on its own piles/footing where the walls aren't designed as permanent structural elements. Waterproofing is applied to the outside of the box before backfilling — a membrane system, or increasingly a fully-bonded/crystalline waterproofing approach — because once backfill and pavement go back on top, the waterproofing layer is no longer accessible for repair without breaking everything back open.

On a live-road project, this phase is very often staged in halves: build and open one carriageway while traffic runs on a temporary diversion alongside, then switch traffic onto the completed half and repeat on the other, which keeps some capacity open throughout but roughly doubles the construction duration compared to a full-width closure. Backfill is placed and compacted in controlled layers around and over the completed box exactly like an earthwork subgrade, because settlement of poorly compacted backfill above a rigid box shows up as a dip or bump right at the structure's edges — one of the most common visible defects on an otherwise well-built underpass.

Pushing a finished structure through the ground beneath a line that never stops running

Work starts well back from the live corridor: a jacking pit is excavated and lined on one side, with a thrust wall (or thrust block) built at its rear face to react against the full jacking force — and behind that, the RCC box itself is cast, either in one piece or in sequential segments, exactly as if it were a normal structure, just built sideways-on to its final position. A cutting edge with a steel canopy is fixed to the leading face of the box, and — most critically for anything under a live railway — the track (or road) directly above the drive alignment is first transferred onto a temporary load-bearing structure, commonly a steel/timber crib or a rail-supported girder system, so it keeps carrying live loads throughout the push even as the ground beneath it is progressively excavated and replaced by the advancing box.

Excavation ahead of the cutting edge proceeds by hand or small excavator in a controlled, incremental sequence — sometimes literally described as tunnelling by hand under the canopy — with the box advanced by hydraulic jacks (commonly in the 200–650 tonne range, ganged together) reacting against the thrust wall, typically in short strokes of a few hundred millimetres before the jacks are reset. Track or pavement settlement is monitored continuously during the push — automated total stations or precise levelling at frequent intervals — against a strict trigger-level regime, because the entire method's premise is that the live corridor above never knows the ground under it changed; any settlement trending toward the trigger level halts advance for grouting or re-support before it becomes visible at the surface.

Boring instead of pushing, for smaller crossings and tighter tolerances

A microtunnel boring machine (MTBM) — remotely operated and steered, with a slurry or earth-pressure-balance face to control ground and groundwater pressure at the cutting head — advances from a launch shaft toward a reception shaft, with precast concrete or steel pipe (or, for pedestrian/small-vehicle underpasses, box sections) jacked in behind it by the same hydraulic jacking principle used in box pushing, just around a bored circular or near-circular face rather than an excavated rectangular one. Line and level are maintained by a laser guidance system referencing the machine's position against the design alignment continuously, correcting steering as it goes rather than in periodic manual surveys — which is what makes the method capable of genuinely tight tolerances over long drives.

Ground loss — the small volume of soil that inevitably goes missing between what the cutting head removes and what the pipe displaces — is the parameter the entire method lives or dies by; kept to a fraction of a percent of the excavated volume through correct face pressure balance, it produces negligible surface settlement, but an unbalanced face (too little pressure lets ground collapse in, too much pressure heaves the surface) shows up immediately in settlement monitoring above. Because the finished bore is a fixed circular or elliptical shape, microtunneling suits utility crossings, pedestrian subways, and smaller vehicular underpasses well, but for a full-width road or rail underpass box, the rectangular efficiency of box jacking or cut-and-cover usually wins on cost per usable cross-sectional area.

Controlling the one thing that undoes almost everything else on this list

Where excavation or boring goes below the water table, groundwater has to be controlled before it controls the job. Wellpoint dewatering — a ring of small-diameter wells connected to a header pipe and vacuum pump — suits shallower excavations in sandy or silty soils. Deep wells with submersible pumps handle larger volumes and greater depths. Where dewatering risks drawing down groundwater under adjacent structures or a live railway embankment (settling foundations that were never meant to move), a cut-off wall — the diaphragm wall or sheet piling from Phase 9, extended below formation level into a low-permeability layer — combined with sump pumping inside the excavation controls water without lowering the regional water table at all.

Every dewatering system needs a discharge plan (to a drain, watercourse, or treatment before discharge, depending on local regulation and water quality) and, on any project near existing structures, a monitoring regime for ground settlement caused by dewatering itself — not just the settlement from excavation or jacking. It's a genuinely common and avoidable failure mode: an underpass built with textbook-perfect excavation support still settles an adjacent building because nobody checked what pumping the water table down 4 metres for six months would do beneath a structure two piles' width away.

Proving the ground did what the design assumed, throughout — not just at the end

Beyond the pile-specific testing in Phase 6, instrumentation across a foundation or underpass project typically includes inclinometers (lateral wall/soil movement), settlement markers and precise levelling (vertical movement of the ground surface, adjacent structures, and — on box-jacking or tunnelling work — the track or road itself), piezometers (pore water pressure, telling you whether the dewatering and excavation support are actually working as designed), and strain gauges on struts or anchors where bracing loads need direct verification rather than calculation alone. On live-corridor work, this data usually flows to an automated monitoring dashboard with real-time alerts against pre-agreed trigger levels, because a monthly report is far too slow a feedback loop for something moving day by day underneath a railway.

Concrete quality control (cube/cylinder strength tests, slump tests at point of placement) and reinforcement inspection (bar size, spacing, cover, lap lengths) apply throughout exactly as on any concrete structure — but on underground and underwater work specifically, this is genuinely the last chance to catch a defect before it's permanently inaccessible, which is worth repeating because it's the single biggest difference between quality control here and on an above-ground building: there is no second inspection once concrete goes around a tremie pipe or a box gets pushed past the point of no return.

Ground movement doesn't necessarily stop the day construction does

As-built drawings — pile positions and levels as actually installed, retaining wall alignments, and the underpass structure's final geometry — are handed over along with the complete QA/QC and instrumentation record set. Where long-term settlement monitoring points were installed (common on box-jacked or tunnelled underpasses beneath a live railway, at the railway authority's insistence), monitoring frequently continues for months or years after construction completes, since consolidation settlement in clay soils in particular can continue well after loading is applied. A maintenance plan for the finished structure covers joint and waterproofing inspection, drainage sump/pump servicing where the underpass sits below the surrounding water table (a permanent, not temporary, condition for most underpasses), and periodic structural inspection consistent with the asset owner's bridge/structure inspection regime.