How MRT Tunnels Are Built Using Mathematics

A tunnel boring machine can disappear underground for kilometres. Yet when it breaks through, the hole must still be where the railway map said it would be.

That is not luck. It is geometry, surveying, geotechnics, probability, structural mechanics, numerical modelling, monitoring and repeated World Return.

An MRT tunnel begins as a line on a plan.

Then the line has to become a three-dimensional path through real ground.

It may pass beneath roads, canals, expressways, buildings, utilities and existing railway tunnels.

The path must bend horizontally.

It must rise and fall vertically.

It must enter stations at the correct orientation and elevation.

It must stay inside a structural envelope large enough for trains, tracks, equipment and maintenance access.

It must remain far enough from other structures and utilities.

And while the tunnel is being excavated, the surrounding ground must not move so much that buildings, roads or operating railways above are damaged.

So the construction problem is not:

dig from A to B

It is:

construct a continuous three-dimensional railway path through uncertain ground while keeping geometry, ground movement, structural capacity, existing infrastructure and future train operation inside compatible limits.

Singapore’s current rail programme gives us unusually good public examples.

Cross Island Line Phase 1 uses large tunnel boring machines, including a 12.6-metre-diameter machine on selected stretches so one large tunnel can eventually carry two tracks. Cross Island Line Phase 2 includes twin bored tunnels extending about 2.4 kilometres beneath the Sungei Pandan canal and major infrastructure, reaching depths of up to around 40 metres in one contract.

In March 2026, LTA awarded the final Downtown Line 2 Extension civil contract for twin bored tunnels of about 1.8 kilometres each beneath Woodlands Road and existing infrastructure including the Kranji Expressway flyover. LTA said real-time monitoring of ground stability will be used during the works.

The Thomson-East Coast Line used an even wider family of methods: a retractable micro-tunnel boring machine near Orchard Boulevard, 24/7 settlement and movement monitoring around Orchard MRT station, non-invasive ground investigation using vibrations or electrical methods in difficult built-up areas, and ground freezing to create an “ice wall” before tunnelling at Marina Bay.

And in January 2022, Circle Line 6 completed its final tunnel breakthrough after boring twin tunnels about two kilometres from Prince Edward Road station towards Cantonment station.

Each example tells the same story:

the railway exists above ground as coordinates
before it exists underground as concrete

This article continues the eduKateSG MRT mathematics cloud. The permanent whole-system owner is How MRT Works | It’s Mathematics. Closely connected pillars include How MRT Track Inspection Works Using Mathematics, How MRT Tunnel Ventilation and Airflow Work Using Mathematics, How MRT Signalling and Train Regulation Work Using Mathematics and How MRT Network Resilience Is Measured Using Mathematics.

The RFE — Why Does Tunnel-Construction Mathematics Exist?

The weakest objective would be:

dig as fast as possible

A tunnel bored quickly but off alignment is not a railway tunnel.

A tunnel on perfect alignment that causes unacceptable settlement above is not a successful construction project.

A tunnel that is structurally sound but too small for the railway structure gauge is unusable.

A tunnel that reaches the station but leaves no room for track geometry, cables, walkways or ventilation also fails its future job.

The Reason for Existence is:

convert a designed railway alignment
into a physical underground structure
with sufficient geometric accuracy
while controlling ground movement
and preserving surrounding infrastructure
so track, trains and systems can operate for decades

The public mathematical system can be written:

alignment geometry
+
survey control
+
ground model
+
TBM steering
+
excavation support
+
structural lining
+
settlement monitoring
+
construction correction
+
as-built survey
= usable railway tunnel

The RFE is not to create a hole. It is to preserve the future railway geometry while the ground is temporarily being asked to stop behaving like untouched ground.

Prompt 1 — How Does a Railway Alignment Become a 3D Curve?

A rail alignment is a three-dimensional centreline.

Let distance travelled along the alignment be chainage s.

The tunnel centreline is:

r(s) = [x(s), y(s), z(s)]

The horizontal alignment controls x and y.

The vertical alignment controls z.

Horizontal tangent and curvature

Let horizontal heading angle be θ(s).

dx/ds = cosθ
dy/ds = sinθ

Curvature is:

κ = dθ/ds

For a circular curve:

κ = 1/R

where R is curve radius.

A smaller R means greater curvature.

Railway alignments generally avoid jumping instantly from zero curvature on a straight to full curvature on a circular arc.

A transition curve can increase curvature gradually.

For an ideal clothoid:

κ(s) = s/A²

where A is the clothoid parameter.

This gives:

straight
→ gradually increasing curvature
→ circular curve

The tunnel boring machine therefore follows not merely a series of GPS-like waypoints but a designed continuous geometry.

Vertical alignment

Define gradient:

g = dz/ds

A constant gradient gives a straight vertical profile.

To connect two gradients smoothly, a parabolic vertical curve is often used conceptually.

If initial gradient is g₀, final gradient is g₁, vertical-curve length is L and Δg=g₁−g₀:

z(s)
= z0 + g0 s + (Δg/2L)s²

and local gradient becomes:

g(s)=g0+(Δg/L)s

This matters because tunnel depth, drainage, station elevation, train energy and neighbouring infrastructure all depend on the vertical profile.

Geometry has many owners

A curve is not chosen only for the tunnelling machine.

It must also work for:

train speed and ride comfort,
wheel–rail forces and noise,
station locations,
available land,
existing tunnels and utilities,
geology,
construction access,
future maintenance.

That is why LTA’s public explanations of rail construction repeatedly connect alignment to existing infrastructure and ground conditions.

The tunnel centreline is the compromise line where urban geometry, train dynamics, geology and construction finally agree.

Prompt 2 — How Does Surveying Put an Invisible Line Underground?

The designed alignment exists in a coordinate system.

The construction machine exists underground.

Surveying connects them.

A survey control network consists of known points with coordinates:

Pi = (Xi,Yi,Zi)

Angles, distances and elevations are measured between these points to establish and check the coordinate frame used by construction.

From bearing and distance to coordinates

If horizontal distance between two survey points is d and bearing is θ:

ΔX = d cosθ
ΔY = d sinθ

So:

X2 = X1 + d cosθ
Y2 = Y1 + d sinθ

Vertical measurement adds Z.

A long underground traverse therefore builds coordinate knowledge step by step.

Why errors accumulate

Every measured angle and distance has uncertainty.

Suppose one measured distance is d±σ_d.

Suppose bearing is θ±σ_θ.

Approximate coordinate variance can be propagated using derivatives.

σX²
≈ (∂X/∂d)²σd²
+ (∂X/∂θ)²σθ²

Since:

X=d cosθ

we get:

σX²
≈ cos²θ σd²
+ d² sin²θ σθ²

Long underground traverses need redundancy and repeated checks because small errors can accumulate.

Closure error

If a survey route returns to a known point, the computed endpoint may not land exactly on the known coordinate.

eclosure
= √[(ΔX)²+(ΔY)²+(ΔZ)²]

The residual is distributed and analysed through survey adjustment rather than ignored.

Least-squares adjustment chooses the most probable corrected coordinates given redundant measurements and their uncertainties.

minimise vᵀPv
subject to observation equations

where v contains measurement residuals and P contains weights.

The public lesson is:

Surveying does not make uncertainty disappear. It measures uncertainty well enough that the tunnel can be steered without pretending the coordinates are perfect.

Prompt 3 — How Does a Tunnel Boring Machine Know Where to Go?

A tunnel boring machine excavates, supports the face and installs tunnel lining while advancing through the ground.

LTA describes Earth Pressure Balance machines as using excavated material to support the tunnel face before that material is transported away.

Other TBM types are chosen for different ground and groundwater conditions.

The machine must also steer.

Target versus actual axis

At chainage s, target tunnel centreline is:

rtarget(s)

Measured TBM reference position is:

rTBM(s)

Position error vector is:

e(s) = rTBM(s) − rtarget(s)

More useful than total error alone are:

horizontal cross-track error,
vertical error,
heading error,
pitch error.

A controller or operator makes small steering corrections so these errors remain bounded.

Why angular error is dangerous

Suppose the machine has a small constant heading error Δθ.

After distance L, lateral drift is approximately:

e ≈ L tanΔθ

For small angles:

tanΔθ ≈ Δθ

so:

e ≈ LΔθ

Suppose fictional heading error is only 0.01°.

0.01° = 1.745×10⁻⁴ rad

After one kilometre, if never corrected:

e ≈ 1000×1.745×10⁻⁴
  ≈ 0.175 m

About 17.5 centimetres.

The example is fictional and not an MRT tolerance.

It shows why continuous steering correction matters.

Curved advance

A TBM cannot turn like a car.

It bends its trajectory gradually through controlled thrust and articulation while rings are installed behind.

If each installed ring advances length ℓ and produces small heading increment Δθ_i, accumulated direction becomes:

θn = θ0 + Σi Δθi

Approximate ring-centre coordinates advance by:

xn+1 = xn + ℓ cosθn
yn+1 = yn + ℓ sinθn

The physical tunnel is therefore a discrete sequence of lining rings approximating a continuous designed curve.

After breakthrough, LTA engineering specifications require accurate as-built surveys of tunnel rings and rail-support geometry so actual construction can be compared with design and any out-of-tolerance locations identified.

A TBM does not aim once and hope. It repeatedly measures where it is, compares that with the invisible centreline and makes thousands of small corrections before the tunnel becomes permanent.

Prompt 4 — How Can Two Tunnel Drives Meet?

A tunnel breakthrough is a geometry exam.

Imagine one tunnel drive starts at Station A and another target lies at Station B.

Both are referenced to the same control network and design alignment.

At breakthrough, desired endpoint is:

P* = (X*,Y*,Z*)

Actual endpoint is:

P = (X,Y,Z)

Breakthrough position error is:

ebreak
= √[(X−X*)²+(Y−Y*)²+(Z−Z*)²]

But total magnitude alone is incomplete.

Engineers care about:

horizontal offset,
vertical offset,
heading mismatch,
profile mismatch,
remaining structural and track clearance.

Uncertainty budget

Breakthrough error can arise from several sources:

survey control error
+ shaft-transfer error
+ underground traverse error
+ TBM guidance error
+ structural installation error

If independent standard uncertainties are σ₁, σ₂ … σ_n, a root-sum-square estimate is:

σtotal
≈ √(σ1²+σ2²+...+σn²)

Correlation can make the real calculation more complex.

The important planning principle is that total tolerance must be budgeted across contributing systems.

CCL6 as a public breakthrough example

LTA marked CCL6 tunnelling completion in January 2022 with the final breakthrough from Prince Edward Road station into Cantonment station. The final stretch included about two kilometres of twin bored tunnels.

Behind the ceremony was a long chain of coordinates:

survey control
→ launch position
→ repeated TBM guidance
→ lining ring geometry
→ as-built position
→ breakthrough

That is why “meeting underground” is not the last-minute part of tunnel construction.

It is the accumulated result of every previous measurement.

A tunnel breakthrough is not when two underground spaces finally discover each other. It is when years of coordinate decisions are tested against one physical point.

Prompt 5 — How Do Engineers Model Ground They Cannot Fully See?

The ground is not a manufactured material.

It changes from place to place.

One metre may contain soft clay.

The next may contain weathered rock.

A boulder can appear inside weaker soil.

Groundwater pressure changes with depth and permeability.

The tunnel therefore begins with investigation.

At borehole or investigation location i, observations may include:

zi = depth
soil/rock type
strength
stiffness
water level
permeability
other geotechnical parameters

But observations are sparse compared with the full tunnel volume.

Engineers interpolate and build a geological model.

That model has uncertainty.

Random-variable view

Instead of saying soil stiffness is exactly E, represent it as:

E ~ probability distribution

For example:

E ~ Normal(μE,σE²)

where appropriate.

Then predicted settlement, face stability and structural loading also become uncertain.

This produces scenario analysis:

soft ground scenario
expected ground scenario
stiff ground scenario
high-water scenario
mixed-face scenario

TBM method and operating parameters can then be selected and adapted within authorised engineering envelopes.

Mixed ground

LTA’s public Cross Island Line environmental documentation notes that settlement risk can be more challenging when a TBM encounters interfaces between different geological strata, mixed soil-and-rock faces, extended stoppages or abrasive conditions.

Why?

Because the excavation face can respond unevenly.

One side may cut easily while another resists strongly.

Ground support and material extraction become harder to balance.

This is a spatially varying boundary problem.

Non-invasive investigation

On TEL, LTA used non-invasive techniques in places where conventional drilling was difficult. Ground vibration or electrical-response methods can infer subsurface properties without digging a borehole at every location.

The inverse problem is:

measure surface response
→ infer likely underground material properties

This is the same structure found across science:

hidden state
→ observable signal
→ model
→ estimated hidden state

The ground model is never the ground itself. Tunnelling remains safe by keeping that distinction visible and letting new observations update the model continuously.

Prompt 6 — Why Does Tunnelling Make the Ground Settle?

Before tunnelling, the ground occupies a volume.

A tunnel removes material and replaces it with a lining.

If excavation and support perfectly preserved the original ground volume and stress field, there would be no movement.

Real construction creates small changes.

Ground may move slightly towards the tunnel.

At the surface, this can create a settlement trough.

Gaussian settlement trough

A classic empirical representation is the Peck Gaussian trough:

S(x)
= Smax exp[−x²/(2i²)]

where:

S(x) is surface settlement at offset x from tunnel centreline,
S_max is maximum settlement above the centreline,
i is trough-width parameter.

Area under the settlement trough per unit tunnel length is:

Vs = √(2π) i Smax

If tunnel excavation area is A_exc, a conceptual volume-loss ratio is:

VL = Vs/Aexc

A fictional settlement example

Suppose:

Smax = 8 mm = 0.008 m
i = 8 m

Then:

Vs
= √(2π)×8×0.008
≈ 0.160 m²

For a fictional 6 m diameter circular tunnel:

Aexc = π(3²)
     ≈ 28.27 m²

So:

VL ≈ 0.160/28.27
   ≈ 0.00566
   ≈ 0.57%

This is a fictional educational example, not a Singapore construction acceptance limit.

The key point is that a very small percentage of underground volume loss can become measurable movement at the surface.

Differential settlement matters more than one number

A building does not care only how far one point moves downward.

It cares how differently nearby points move.

For two points separated by distance L:

angular distortion
≈ |S2−S1|/L

Differential movement can create structural distortion even when absolute settlement is modest.

This is why LTA repeatedly emphasises real-time ground-stability and settlement monitoring in current tunnel contracts.

The TEL Orchard example used 24/7 monitoring around an operating MRT station. DTL2e and CRL contracts similarly specify real-time monitoring and mitigation to protect nearby infrastructure.

Tunnel settlement is the construction version of World Return: the ground is continuously asked whether the excavation is behaving the way the model promised.

Prompt 7 — How Does the Tunnel Lining Carry the Ground?

Behind many bored-tunnel TBMs, precast concrete segments are assembled into rings.

The rings create the permanent tunnel lining.

A simple circular ring under uniform external pressure p gives an introductory membrane-force scale.

For tunnel radius R, circumferential compression per unit tunnel length is approximately:

Nθ ≈ pR

If effective lining thickness is t, idealised hoop stress scale is:

σθ ≈ pR/t

Real tunnel loading is rarely uniform.

The lining experiences combinations of:

axial force N,
bending moment M,
shear V,
joint forces,
ground reaction,
water pressure,
construction loads.

A familiar structural-stress expression is:

σ = N/A ± My/I

where y is distance from the neutral axis.

Modern tunnel design uses finite-element and soil-structure-interaction models to represent non-uniform ground, segment joints and construction stages.

Ground and lining interact

A tunnel lining is not a pipe sitting in empty space.

The surrounding ground provides support.

A simplified ground spring is:

pground = k u

where u is lining displacement and k is ground-reaction stiffness.

Soft ground produces different interaction from stiff rock.

Groundwater adds pressure.

Adjacent structures change the stress field.

This is why tunnelling method cannot be separated from geotechnical context.

Why large-diameter tunnels are different

CRL Phase 1 includes a 12.6 m diameter TBM on selected stretches, the largest TBM deployed on an LTA rail project when announced.

The larger tunnel can carry two tracks in one bore.

That can reduce the number of drives, but larger diameter changes:

excavated area,
ground disturbance potential,
face pressure requirements,
lining forces,
machine logistics,
spoil volume.

Excavated circular area scales with diameter squared:

A = πD²/4

Increase diameter from 6 m to 12 m and area becomes four times as large.

Large-diameter tunnelling therefore exchanges fewer bores for a much larger excavation cross-section.

A tunnel lining works because concrete, joints and ground become one load-sharing system after excavation has temporarily disturbed the original balance.

Prompt 8 — How Does Construction Control Itself While It Is Still Happening?

A tunnel design is a prediction.

Construction tests that prediction every metre.

Before excavation, engineers predict:

ground movement,
TBM forces and pressures,
structural loading,
water behaviour,
settlement at nearby structures.

During excavation, monitoring returns:

surface settlement,
building movement,
tunnel movement,
track movement where existing rail is nearby,
groundwater observations,
TBM operational data,
survey alignment.

A generic observational loop is:

predict
→ excavate small increment
→ measure
→ compare
→ adjust construction within authorised controls
→ excavate again

This is the observational method.

LTA’s older railway-protection guidance publicly illustrates this philosophy with automated tiltmeters, settlement cells, crack meters, prisms, vibration sensors and repeated track surveys around sensitive existing railway works.

Current DTL2e and CRL contracts continue the same broad idea through real-time ground-stability monitoring and mitigation.

Prediction error

Let predicted settlement at sensor i be Ŝ_i.

Measured settlement is S_i.

ei = Si − Ŝi

Across n instruments:

RMSE
= √[(1/n)Σei²]

A growing systematic residual suggests the model is drifting from reality.

The engineering question is not merely:

Is settlement below a limit?

It is also:

Is the trend behaving the way we expected?
Is the rate changing?
Are several instruments telling the same story?
What construction event preceded the change?

Trend detection can reveal a problem earlier than waiting for a single threshold crossing.

Ground freezing as boundary control

TEL’s Marina Bay construction used ground freezing where high soil permeability created water-seepage risk.

The method changes the material itself.

unfrozen water-bearing ground
→ remove heat
→ water freezes
→ frozen soil gains stiffness and lower permeability
→ excavation proceeds inside temporary frozen boundary

A heat-conduction model begins with:

ρc ∂T/∂t
= k∇²T + heat sources/sinks

Freezing changes temperature, phase and engineering properties.

This is a striking example of engineering not merely adapting to ground conditions but temporarily redesigning them.

Tunnel construction becomes safest when measurement is allowed to interrupt the plan. The design predicts; the ground replies; construction listens.

A Complete Fictional Tunnel-Building Example

Consider a fictional 1.5 km bored MRT tunnel between two station boxes.

All values are educational abstractions and do not represent Singapore MRT construction tolerances or live works.

Step 1 — Horizontal alignment

First 600 m is straight.

Then a 100 m transition enters a circular curve of radius:

R = 900 m

Curve curvature:

κ = 1/900
  ≈ 0.001111 m⁻¹

For a 100 m clothoid transition ending at that curvature:

κ(L)=L/A²

A = √(L/κ)
  = √(100/0.001111)
  ≈ 300 m

Step 2 — Vertical profile

Suppose tunnel descends at 1.5 per cent for 400 m.

Δz = gL
   = −0.015×400
   = −6 m

The centreline is now six metres deeper than the starting tangent before the vertical curve changes gradient.

Step 3 — Survey uncertainty

Suppose the combined standard uncertainty budget at breakthrough is made of fictional components:

surface control: 8 mm
shaft transfer: 10 mm
underground traverse: 14 mm
TBM guidance: 18 mm

Assuming independence:

σtotal
= √(8²+10²+14²+18²)
= √684
≈ 26.2 mm

Again, these values are invented.

The useful lesson is that every stage consumes part of the total uncertainty budget.

Step 4 — TBM heading correction

At 700 m chainage, survey finds horizontal cross-track error:

eperp = +18 mm

and heading error:

Δθ = +0.006°

If that angular error remained for another 500 m:

Δθ = 0.006×π/180
   ≈ 1.047×10⁻⁴ rad

drift ≈ 500×1.047×10⁻⁴
      ≈ 52 mm

The guidance system therefore corrects gradually rather than waiting for the endpoint.

Step 5 — Settlement prediction

Predicted trough parameters:

Smax = 7 mm
i = 9 m

At lateral offset x=9 m:

S(9)
= 7 exp[−9²/(2×9²)]
= 7e^(−0.5)
≈ 4.25 mm

A monitoring point at that offset later records 5.1 mm.

eS = 5.1−4.25
   = +0.85 mm

One small residual is not automatically a problem.

The trend across many instruments and excavation steps determines whether the ground model needs revision.

Step 6 — Lining load scale

Suppose fictional equivalent uniform external pressure is:

p = 250 kPa

and lining mean radius:

R = 3.0 m

Ideal hoop-force scale:

Nθ ≈ pR
   = 250×3
   = 750 kN/m

Real segment design would include non-uniform ground pressure, joints, water, bending, construction and load combinations.

Step 7 — Breakthrough

At the receiving station, fictional measured offsets from target are:

ΔX = +14 mm
ΔY = −19 mm
ΔZ = +8 mm

3D endpoint error:

ebreak
= √(14²+19²+8²)
= √621
≈ 24.9 mm

That value has meaning only relative to the validated project tolerance and remaining track/structure geometry, which are deliberately not published here.

Step 8 — As-built World Return

The tunnel is surveyed ring by ring.

Actual centreline is compared with design.

Then track designers use the real constructed geometry rather than pretending construction followed the design perfectly.

design alignment
→ construct
→ as-built survey
→ compare
→ verify clearances
→ final track geometry
→ operate
→ inspect again

The tunnel’s construction model has finally returned to the railway-operating model.

The MRT Tunnel Deletion Tests

Remove the design centreline

The TBM can dig but has no railway geometry to follow.

Remove survey control

The designed coordinates cannot be transferred reliably underground.

Remove uncertainty

Every measured coordinate becomes falsely exact and the breakthrough error budget disappears.

Remove steering correction

Tiny heading errors accumulate into large endpoint drift.

Remove the geological model

The TBM encounters changing ground without any predictive framework for support, wear or settlement risk.

Remove settlement monitoring

The ground can move away from prediction without the construction team seeing the divergence early.

Remove the lining

The excavated opening has no permanent structural system maintaining the tunnel void.

Remove groundwater

Permeability and water pressure disappear from a problem where they can control method and stability.

Remove as-built survey

Track and system designers are forced to pretend the constructed tunnel equals the drawing exactly.

Remove World Return

The ground, TBM and lining can disagree with the model indefinitely without changing the construction strategy.

The MRT Tunnel Paradoxes

Paradox 1 — The most accurate tunnel begins with an uncertain ground model

Precision is achieved not by pretending uncertainty is absent but by measuring and updating it.

Paradox 2 — A bigger tunnel can reduce the number of tunnels needed

CRL’s large-diameter TBM can create one bore containing two tracks, exchanging bore count for much larger excavation area.

Paradox 3 — The TBM must continuously deviate from its present heading to remain on the intended alignment

Curved tunnel geometry requires constant controlled steering rather than “straight” machine behaviour.

Paradox 4 — Tunnelling can be safest when construction pauses

If measurements disagree with prediction, stopping to investigate can preserve the project better than continuing at maximum production.

Paradox 5 — Freezing the ground can make excavation possible

At Marina Bay, temporary frozen ground was used to change permeability and stability before tunnelling.

Paradox 6 — Very small volume loss can create visible surface settlement

Large tunnel area multiplied over long distance means sub-percent ground losses can still move measurable volumes.

Paradox 7 — A tunnel is finished structurally before it is finished geometrically

After lining completion, as-built surveying and track/system installation still have to convert the concrete tube into a railway.

Paradox 8 — The tunnel is underground, but much of its safety is monitored from the surface

Settlement prisms, building instruments and other monitoring points reveal how underground excavation is affecting the world above.

The Tunnel Construction Audit

What passenger and network job will this tunnel eventually perform?
What is the designed 3D centreline?
What horizontal curvature and transition geometry are required?
What vertical gradients and curves are required?
What structures, utilities and existing railways constrain the alignment?
What survey control establishes the coordinate system?
What uncertainty exists in that control?
How is the coordinate frame transferred underground?
What is the TBM’s measured position, heading and pitch?
What is the cross-track and vertical error?
What happens to the endpoint if heading error persists?
What geological units are expected?
Where are ground-model uncertainties highest?
What groundwater and permeability conditions matter?
What tunnel-boring method fits the ground?
What settlement is predicted?
What surface structures lie inside the settlement influence zone?
What monitoring instruments observe those structures?
Are observed movements matching prediction?
What trend would trigger engineering review?
What loads act on the lining?
How does ground stiffness influence lining behaviour?
What temporary ground-improvement method is required?
What is the accumulated breakthrough uncertainty?
What does the as-built survey say the tunnel actually became?
Can the final track and structure gauge fit the real tunnel?
What operational monitoring continues after trains begin running?

How the Mathematics Grows from School to Research

Primary Mathematics

distance,
coordinates,
angles,
circles,
area and volume.

Secondary Mathematics and Physics

trigonometry,
coordinate geometry,
gradients,
forces and pressure,
statistics and measurement error.

Junior College

calculus and curvature,
vectors,
matrices,
probability distributions,
mechanics,
differential equations.

University and Research

geodesy and engineering surveying,
least-squares estimation,
geotechnical engineering,
soil and rock mechanics,
finite-element analysis,
soil–structure interaction,
tunnel boring machine control,
probabilistic ground modelling,
computer vision and automated monitoring,
digital twins and observational design.

A student learns that a straight line has a gradient.

A tunnelling engineer asks how that line becomes a kilometre-long three-dimensional excavation through uncertain ground while every millimetre of error, every settlement reading and every lining ring changes the next move.

The World Return — When the Ground Answers the Drawing

The drawing says the tunnel should be here.

The survey says where it actually is.

The geotechnical model predicts settlement.

Instruments say how much the ground actually moved.

The structural model predicts lining forces.

Instrumentation, inspection and as-built geometry return the physical result.

alignment prediction
→ TBM advance
→ survey residual

settlement prediction
→ excavation
→ movement residual

structural prediction
→ lining installed
→ inspection and monitoring residual

Let design centre coordinate at chainage s be r̂(s).

As-built centre is r(s).

ealignment(s)=r(s)−r̂(s)

Let predicted settlement be Ŝ(x,t) and observed settlement S(x,t).

esettlement(x,t)=S(x,t)−Ŝ(x,t)

These residuals become construction intelligence.

They show whether the next ring should be steered differently.

Whether ground support needs reassessment.

Whether monitoring should intensify.

Whether the geological model was wrong.

And once the tunnel opens, track inspection continues the same loop.

design tunnel
→ build tunnel
→ survey tunnel
→ install track
→ run trains
→ inspect geometry
→ maintain
→ compare again

The tunnel drawing is only the first hypothesis. The final railway is the geometry that survived construction, measurement and operation.

RFE Return — What Does a Good MRT Tunnel Owe the Passenger?

The passenger should not know whether the tunnel was bored through marine clay, granite, weathered rock or mixed ground.

They should not know which survey control network guided the TBM.

They should not know how many settlement instruments watched the ground above.

They should experience:

correct railway alignment
+
smooth track geometry
+
sufficient structure gauge
+
stable surrounding ground
+
dry and durable underground space
+
ventilation and systems that fit
+
long-term inspectability and maintenance

The tunnel also owes the city something.

It must arrive without unnecessarily damaging the world above it.

The tunnel is therefore a hidden civic compromise:

move millions of future passengers
through ground already supporting today's city

The RFE of MRT tunnelling is to make a future railway occupy underground space precisely enough that the existing city can continue above while the new city begins moving below.

Conclusion — The Tunnel Is a Coordinate Made Permanent

A railway planner draws an alignment.

A surveyor turns it into control points.

A geotechnical engineer turns sparse boreholes and geophysics into a ground model.

A TBM begins to excavate.

The machine moves a few metres.

The survey says it is slightly right of target.

The next steering correction moves it back.

A settlement instrument above changes by a fraction.

The ground model is checked.

The lining ring is installed.

Another ring follows.

The tunnel curves.

The vertical grade changes.

The machine crosses beneath existing infrastructure.

Thousands of measurements accumulate.

Then one day, the cutterhead reaches the receiving station.

Breakthrough.

But the mathematical job is not finished.

The tunnel is surveyed again.

The track is laid to the real as-built geometry.

Clearances are verified.

Systems are installed.

Trains begin running.

Inspection starts the next phase of the same measurement loop.

map
→ 3D centreline
→ survey control
→ ground model
→ TBM guidance
→ lining rings
→ settlement monitoring
→ breakthrough
→ as-built survey
→ track
→ trains
→ inspection

The passenger sees a dark window between stations.

The surveyor sees coordinates.

The geotechnical engineer sees uncertain ground.

The structural engineer sees ring forces.

The TBM operator sees steering error.

The monitoring engineer sees settlement trends.

The track engineer sees the final centreline the train must inherit.

An MRT tunnel works when a line imagined in coordinates survives geology, excavation and uncertainty strongly enough to become a permanent path through the city.

Key Equations

r(s)=[x(s),y(s),z(s)]
3D alignment centreline

dx/ds=cosθ, dy/ds=sinθ
Horizontal tangent geometry

κ=dθ/ds
Curvature

κ=1/R
Circular-curve curvature

κ(s)=s/A²
Ideal clothoid transition curvature

g=dz/ds
Vertical gradient

z(s)=z0+g0s+(Δg/2L)s²
Parabolic vertical curve

ΔX=d cosθ, ΔY=d sinθ
Survey coordinate increment

eclosure=√(ΔX²+ΔY²+ΔZ²)
3D survey closure error

minimise vᵀPv
Least-squares survey-adjustment objective

e(s)=rTBM−rtarget
TBM alignment-error vector

e≈LΔθ
Small-angle drift from heading error

θn=θ0+ΣΔθi
Accumulated ring heading

xn+1=xn+ℓcosθn
yn+1=yn+ℓsinθn
Discrete ring-centre advance

ebreak=√[(X−X*)²+(Y−Y*)²+(Z−Z*)²]
Breakthrough position error

σtotal≈√(Σσi²)
Independent uncertainty budget

S(x)=Smax exp[−x²/(2i²)]
Gaussian settlement trough

Vs=√(2π)iSmax
Settlement-trough volume per unit tunnel length

VL=Vs/Aexc
Conceptual volume-loss ratio

angular distortion≈|S2−S1|/L
Differential settlement measure

Nθ≈pR
Ideal circular-lining hoop-force scale

σθ≈pR/t
Ideal hoop-stress scale

σ=N/A±My/I
Combined axial and bending stress

pground=ku
Simplified ground-reaction spring

A=πD²/4
Circular excavation area

ρc∂T/∂t=k∇²T+...
Heat equation for ground freezing concept

RMSE=√[(1/n)Σ(S−Ŝ)²]
Settlement prediction error

ealignment=r−r̂
As-built alignment residual

Reader-safety note: All fictional survey uncertainties, alignment errors, settlement values, lining pressures and construction examples are educational abstractions. This article does not reproduce live Singapore MRT tunnel coordinates, exact protection-zone vulnerabilities, construction-control thresholds, TBM operating parameters, ground-treatment recipes, structural design details or other security- or safety-sensitive construction information.

Continue the MRT Mathematics Cloud

How MRT Works | It’s Mathematics — the permanent whole-system hub.
How MRT Timing Works Using Mathematics — train movement in time.
How MRT Braking Works Using Mathematics — stopping geometry.
How MRT Headway Works Using Mathematics — safe spacing and frequency.
How MRT Station Dwell Time Works Using Mathematics — station time.
How MRT Delays Propagate and Recover Using Mathematics — railway recovery.
How an MRT Timetable Is Built Using Mathematics — event geometry.
How MRT Energy Use Is Optimised Using Mathematics — traction energy.
How MRT Passenger Capacity Is Calculated Using Mathematics — passenger throughput.
How MRT Routes and Transfers Are Optimised Using Mathematics — network geometry.
How MRT Network Resilience Is Measured Using Mathematics — redundancy and recovery.
How MRT Predictive Maintenance Works Using Mathematics — degradation prediction.
How MRT Track Inspection Works Using Mathematics — geometry after construction.
How MRT Wheel–Rail Contact Works Using Mathematics — train-track physical interface.
How MRT Noise and Vibration Work Using Mathematics — vibration through tunnel and train.
How MRT Power Supply Works Using Mathematics — traction electricity.
How MRT Signalling and Train Regulation Work Using Mathematics — movement authority.
How MRT Platform Screen Doors Work Using Mathematics — precision station interface.
How MRT Tunnel Ventilation and Airflow Work Using Mathematics — the air inside the tunnel.
How MRT Depots and Fleet Operations Work Using Mathematics — the operating fleet.

The next natural pillar is How MRT Stations Are Built Using Mathematics: excavation geometry, retaining walls, groundwater, structural loads, interchange connections, vertical circulation, construction staging and how a station has to be both an underground building and a machine for moving thousands of people every hour.

How MRT Tunnels Are Built Using Mathematics: How Two Underground Drives Still Meet the Same Railway Map