How MRT Rail Grinding and Wheel Profiling Work Using Mathematics

Railways sometimes extend component life by deliberately removing material from the component.

Grinding rail and reprofiling wheels can remove damaged surface layers, restore geometry, reduce roughness and reset the contact conditions that would otherwise accelerate wear, noise and fatigue.

A wheel wears because it touches the rail.

A rail wears because wheels pass over it.

If the shapes remained perfect forever, maintenance would be simpler.

They do not.

Wheel profiles can become uneven.

Rail heads can develop corrugation, local roughness and profile change.

Those changes alter future contact.

contact
→ wear and roughness
→ geometry changes
→ contact changes
→ future wear changes

This is a feedback loop.

Rail grinding and wheel profiling interrupt it.

SMRT publicly describes both as preventive maintenance. Rail grinding is used to smooth uneven track surfaces, while wheel profiling restores wheel roundness and diameter so trains run more smoothly and quietly. SMRT also notes that rail grinding is conducted during limited non-revenue engineering hours, and that individual track sections are typically ground on a recurring maintenance cycle.

LTA has similarly cited rail grinding as one mitigation for higher noise on curved sections of the Thomson-East Coast Line.

This article owns surface and profile restoration as a maintenance-optimisation problem. The contact mechanics belong to How MRT Wheel–Rail Contact Works Using Mathematics. Acoustic consequences belong to How MRT Noise and Vibration Work Using Mathematics. Track sensing belongs to How MRT Track Inspection Works Using Mathematics.

The RFE — Why Remove Metal From a Rail or Wheel?

The weak answer is:

make the surface smooth

Smoothness matters, but the deeper job is to restore a controlled geometry before uncontrolled wear decides the geometry instead.

The Reason for Existence is:

remove the smallest justified amount of damaged or unwanted material at the right time so wheel–rail contact returns toward a healthier geometry without consuming useful component life unnecessarily.

Prompt 1 — What Geometry Is Being Restored?

A rail profile can be represented as height y over lateral coordinate x:

yrail(x)

A wheel profile can be represented similarly:

ywheel(x)

Let target profile be y*(x).

Profile error is:

e(x) = ymeasured(x) − y*(x)

A simple root-mean-square profile error is:

RMS = √[(1/L)∫ e(x)² dx]

But one scalar cannot describe everything.

Engineers care about local shape because small profile changes can alter contact position, equivalent rolling radius and local stress.

The correct maintenance question is therefore not:

How much metal can we remove?

It is:

What geometry must be restored,
where,
and with how little material removal as practical?

Prompt 2 — What Is Rail Corrugation Mathematically?

Rail corrugation is a periodic or quasi-periodic irregularity along the rail surface.

A simple sinusoidal representation is:

r(x) = A sin(2πx/λ)

where:

A is roughness amplitude,
λ is wavelength.

A train moving at speed v encounters spatial wavelength λ at temporal frequency:

f = v/λ

This directly connects rail geometry to noise and vibration.

Suppose corrugation wavelength is 0.05 m and train speed is 15 m/s:

f = 15/0.05
  = 300 Hz

The geometric roughness has become a 300 Hz forcing input.

Grinding removes part of the peaks and restores a controlled rail-head profile.

This is why surface geometry can be heard.

Prompt 3 — How Much Material Should Grinding Remove?

Too little removal leaves the defect.

Too much consumes rail life.

Let required defect-removal depth be d_defect.

Let grinding removal be d_grind.

Conceptually:

dgrind ≥ ddefect + process allowance

while:

dgrind should be as small as justified

For rail length L and effective removed cross-sectional area A_r:

Vremoved = Ar L

Mass removed:

mremoved = ρsteel Vremoved

Material removal is therefore a finite-life expenditure.

The maintenance objective is not maximum smoothness after one intervention.

It is long-term control of surface condition across many interventions.

Prompt 4 — How Does Wheel Profiling Differ From Rail Grinding?

Rail grinding treats a fixed infrastructure surface.

Wheel profiling treats the rotating geometry on the vehicle.

A wheel can develop:

out-of-roundness,
flats or local damage,
diameter differences,
profile wear,
flange shape changes.

Wheel radial error around angle θ can be represented:

er(θ) = rmeasured(θ) − rtarget

Out-of-roundness can be decomposed using Fourier components:

er(θ) = Σn [an cos(nθ)+bn sin(nθ)]

Different harmonic orders correspond to different repeated shape patterns around the wheel.

Profiling restores a target contour and diameter relationship.

But every cut reduces wheel diameter.

Again, maintenance spends material to buy improved geometry.

Prompt 5 — When Should Grinding or Profiling Happen?

Maintenance too early wastes useful material and engineering time.

Maintenance too late allows roughness, wear or fatigue damage to grow.

Let condition metric be D(t).

A simple degradation model is:

D(t) = D0 + kt

for roughly linear degradation.

Real wear can be nonlinear and load-dependent.

Maintenance at time τ creates a reset:

D(τ+) = Dreset

The optimisation is:

minimise
maintenance cost
+ material consumed
+ noise/vibration penalty
+ wear/fatigue risk
+ disruption risk

subject to
condition remaining inside acceptable engineering limits

SMRT’s public maintenance description shows the operational constraint clearly: rail grinding occurs during the short non-revenue engineering window rather than during passenger service.

The problem is therefore not only when does the rail need grinding?

It is also:

when can the railway safely give the track to the grinder?

Prompt 6 — How Does Grinding Change Noise and Vibration?

Roughness excites wheel and rail vibration.

For roughness amplitude spectrum R(f), a simplified system response is:

Y(f) = H(f)R(f)

where H(f) is the wheel–rail–track transfer function.

Grinding modifies R(f).

It does not change every structural resonance H(f).

This distinction is why rail grinding can reduce one noise source without being a universal cure for every noise problem.

LTA’s TEL example makes that public: curved alignment contributed to higher noise on some sections, and mitigation included speed reduction and rail grinding to smooth wheel–rail contact.

The Noise and Vibration pillar owns the acoustics.

Prompt 7 — How Does Inspection Decide What to Grind?

Grinding should follow evidence.

Track inspection can measure or infer:

rail profile,
roughness or corrugation,
surface defects,
geometry change,
noise and vibration trends.

LTA’s NEL Automatic Track Inspection system publicly lists rail corrugation among the defects that onboard cameras and sensors can detect.

Let measured condition at location x be D(x).

Maintenance priority can be conceptualised:

Priority(x)
= condition severity
× traffic exposure
× consequence
× degradation rate
÷ maintenance opportunity

A small defect in a high-use, high-consequence location can outrank a larger but stable defect elsewhere.

This prevents maintenance from becoming a simple “worst number first” list.

Prompt 8 — How Does the Railway Know Grinding Worked?

The maintenance action makes a prediction:

profile error falls
roughness falls
noise/vibration forcing changes
contact improves
future degradation slows

Then the railway measures again.

Before/after profile error:

ΔRMS = RMSafter − RMSbefore

Roughness reduction at frequency f:

ΔR(f)=Rafter(f)−Rbefore(f)

Noise change:

ΔL = Lafter − Lbefore

But immediate grinding marks can themselves create a temporary acoustic signature before the surface runs in, something SMRT notes publicly.

That means evaluation must use the correct time horizon.

immediately after maintenance
≠ steady post-run-in condition

Maintenance is not complete when the grinder leaves. It is complete when the measured wheel–rail system returns the intended geometry and behaviour.

A Complete Fictional Maintenance Example

Consider a fictional curved rail section with corrugation amplitude 0.08 mm at dominant wavelength 50 mm.

Train speed is 15 m/s.

f = 15/0.05
  = 300 Hz

Grinding reduces dominant amplitude to 0.025 mm.

amplitude reduction
= (0.08−0.025)/0.08
≈ 68.75%

If the relevant dynamic response remained approximately linear, vibration forcing at that wavelength would fall substantially.

Suppose 0.20 mm average material depth is removed over an effective 70 mm rail-head width and 500 m length.

Vremoved
= 0.00020×0.070×500
= 0.007 m³

With steel density about 7,850 kg/m³:

mremoved ≈ 55 kg

The intervention has spent real rail material to restore contact geometry.

The correct decision depends on whether the life gained through slower future damage and better contact exceeds the life consumed by the grinding removal.

The Grinding and Profiling Deletion Tests

Remove target profile: maintenance has no geometric definition of success.
Remove roughness wavelength: all surface defects become one undifferentiated number.
Remove material removal: grinding becomes free and cannot consume component life.
Remove inspection: maintenance timing becomes calendar-only regardless of real condition.
Remove traffic exposure: a low-use defect and high-use defect receive the same priority.
Remove engineering window: maintenance can occur whenever desired without service consequence.
Remove wheel–rail coupling: rail profile is optimised without considering the wheel profile it will meet.
Remove World Return: the railway never checks whether grinding actually improved geometry or behaviour.

The Maintenance Paradoxes

Removing metal can extend component life.
The smoothest possible rail after one intervention may not maximise whole-life rail value.
Grinding can temporarily sound worse before the surface runs in.
Wheel and rail maintenance can each fail if optimised without the other.
A small roughness wavelength can create a high-frequency vibration problem.
Maintenance too early and maintenance too late are both forms of waste.
The best time to grind is partly determined by when the railway can spare the track.

The Grinding/Profiling Audit

What target rail or wheel profile is required?
What measured profile error exists?
What roughness wavelengths dominate?
What speed converts those wavelengths into forcing frequencies?
What contact or noise consequence is being observed?
How fast is the condition degrading?
How much material must be removed to restore the target?
How much useful life does that removal consume?
What wheel profile will interact with the restored rail?
What traffic exposure does the location experience?
What engineering access exists?
What other maintenance work competes for the same window?
What inspection evidence justifies intervention now?
What post-maintenance measurement proves the intervention worked?
What new degradation rate appears after the reset?

How the Mathematics Grows from School to Research

Primary: circles, diameter, length and repeated patterns.

Secondary: graphs, periodic functions, ratios, rates and geometry.

Junior College: trigonometric functions, calculus, frequency, statistics and optimisation.

University and research: tribology, wheel–rail dynamics, spectral roughness analysis, fatigue, optimisation, reliability and condition-based maintenance.

The World Return — When the Restored Surface Meets the Train Again

Grinding predicts a reset.

Wheel profiling predicts a reset.

The next trains return the evidence.

measure before
→ grind/profile
→ measure after
→ run trains
→ measure contact, vibration, noise and wear trend
→ compare with expected benefit
→ update next maintenance interval

If roughness returns much faster than predicted, the root cause may not have been removed.

If noise remains high after roughness falls, another source or transfer path may dominate.

If wheel profiles wear unusually after a rail-profile change, the paired geometry must be reconsidered.

The grinder changes the surface. The returning train decides whether the maintenance changed the railway.

Conclusion — Spend a Little Metal to Protect the Larger Machine

Wheels and rails begin with designed shapes.

Contact changes those shapes.

Inspection measures the drift.

Maintenance chooses a moment.

The grinder or lathe removes material deliberately.

Geometry moves back toward target.

The next train returns the result.

contact
→ wear
→ inspection
→ maintenance decision
→ controlled material removal
→ restored profile
→ new contact
→ measured wear return

Rail grinding and wheel profiling work when the railway removes less material deliberately today than uncontrolled contact would destroy tomorrow.

Key Equations

e(x)=ymeasured−ytarget
Profile error

RMS=√[(1/L)∫e²dx]
Profile-error summary

r(x)=A sin(2πx/λ)
Periodic rail roughness

f=v/λ
Spatial wavelength to temporal frequency

dgrind≥ddefect+allowance
Conceptual removal condition

Vremoved=ArL
Removed material volume

mremoved=ρVremoved
Removed steel mass

er(θ)=Σ[an cos(nθ)+bn sin(nθ)]
Wheel out-of-roundness harmonics

D(t)=D0+kt
Simple degradation model

Y(f)=H(f)R(f)
Roughness-to-response relationship

ΔR(f)=Rafter−Rbefore
Post-maintenance roughness change

Reader-safety note: This article does not reproduce Singapore MRT grinding depths, maintenance thresholds, wheel condemning limits, exact profiles, machine settings or restricted maintenance procedures. All numerical examples are fictional teaching values unless directly attributed to a public source.

How MRT Rail Grinding and Wheel Profiling Work Using Mathematics: Why Removing Metal Can Make the Railway Last Longer