How MRT Works | It’s Mathematics

Wait—doesn’t an MRT simply follow the tracks and the timetable?

It does.

But that sentence hides almost everything.

A track does not tell a train how quickly to accelerate, where to begin braking, how closely another train may follow, how long the doors should remain open, whether the next platform can receive more passengers, how much electrical power is available, whether a route is temporarily constrained, or whether a small vibration is harmless noise or the beginning of asset deterioration.

A timetable does not move a train. It declares desired events. The railway still has to turn those events into safe physical trajectories while the world keeps changing around them.

Passengers do not arrive at perfectly even intervals. Walking speeds differ. Interchanges release sudden waves of people. One train may dwell ten seconds longer. A following train may then encounter a different movement constraint. The longer gap attracts more passengers. The busier platform slows boarding. A delay that began as ten seconds can therefore become a passenger-flow problem several stations later.

Meanwhile, motors draw power. Braking trains can return part of their kinetic energy. Wheels press against rails through tiny elastic contact patches. Track geometry changes gradually with wear. Sensors inspect equipment. Engineers decide what to maintain during a short overnight window. Operators prepare degraded modes for the day when ordinary service cannot continue exactly as planned.

The MRT looks simple because an enormous amount of complexity has been successfully compressed into “tap, wait, board, transfer, arrive.”

This page opens that compression.

It is the permanent whole-system hub for the eduKateSG MRT mathematics cloud. Each specialist article owns its deeper derivations. This page owns the connections among them.

The Reason for Existence

The weakest possible objective for an MRT would be:

move trains as fast as possible

A train racing through stations without stopping would perform badly as public transport.

Another weak objective would be:

fit as many people as possible into every carriage

A train filled beyond workable passenger exchange can lengthen dwell, leave people behind downstream and destabilise the service.

Another would be:

make every train exactly punctual

But holding one train to protect a clock time can damage headway, transfers or the journeys of many more passengers.

The MRT’s Reason for Existence is larger:

convert changing human travel demand
into safe, regular, accessible and efficient movement
through finite tracks, stations, trains, power and time
while preserving maintenance, recovery and future service

This page gives that operating class a reader-safe name:

MRT Runtime Class: Mass Mobility Continuity.

Mass, because the system must move large passenger flows rather than one person or one vehicle.

Mobility, because the output is not train motion by itself. It is the passenger’s ability to reach a destination.

Continuity, because one successful trip is not enough. The service must repeat, recover, maintain its assets and remain usable tomorrow.

An operational definition of how an MRT works

An MRT is a closed-loop mass-mobility system that observes the state of trains, tracks, stations, passengers and power; selects safe and feasible control actions; delivers repeated journeys through a timed network; measures the result; and maintains or reconfigures the system so the service can continue.

The definition has seven parts.

1. Physical field: trains, rails, stations, tunnels, depots, power and communication equipment exist in real space.
2. Observed state: position, speed, separation, passenger flow, electrical demand and asset condition are measured or estimated.
3. Constraints: safety, geometry, vehicle limits, platform flow, capacity, accessibility and operating rules define what actions are admissible.
4. Control: acceleration, braking, dwell, holding, routing, train insertion, withdrawal and maintenance are selected within authority.
5. Passenger function: the system must deliver reachable journeys, not merely moving vehicles.
6. Recovery: disturbances must remain bounded where possible, with a route back to useful service.
7. World Return: actual travel times, queues, energy, wear, faults and passenger outcomes return as evidence for correction.

This is why “It’s Mathematics” does not mean the railway is only mathematics.

Steel, concrete, electricity, institutions, budgets, law, skilled people, maintenance practice and passenger behaviour are real. Mathematics is the language used to represent their relationships precisely enough to test, coordinate and improve them.

Mathematics is not the train. It is the disciplined agreement that helps the train, track, timetable and passenger remain in the same reality.

The Singapore scale

As of September 2026, the Land Transport Authority describes Singapore’s operating MRT network as more than 140 stations across six lines, spanning more than 200 kilometres and carrying more than three million passenger journeys a day. LRT adds more than 40 stations and another layer of feeder movement.

At that scale, one operating day contains an immense number of coupled events:

• millions of passenger arrivals and route choices;
• thousands of station stops;
• repeated acceleration and braking cycles;
• continuous train-position and separation updates;
• changing queues, loads and transfer waves;
• large electrical power flows;
• millions of wheel revolutions;
• continuous equipment diagnostics and condition observations;
• and a narrow engineering window in which the system must be inspected and renewed.

No person manually calculates every event.

The system is decomposed into models, controllers, procedures and professional roles. Yet the pieces still have to fit.

What the passenger sees	What the railway must solve
A train arrives	Timetable, position, movement authority, braking, dwell, platform readiness and train regulation
The doors open	Stopping accuracy, platform interface, door state, passenger exchange and safe dwell
The next train is “2 min” away	Measured position, predicted running time, headway, current constraints and information latency
A crowded carriage	Origin–destination demand, train capacity, passenger distribution, downstream boarding and dwell risk
A transfer	Graph routing, walking time, next-train wait, accessibility and missed-connection probability
A smooth ride	Track geometry, suspension, wheel–rail contact, speed profile and maintenance condition
A quiet ordinary journey	Energy, reliability, inspection, staff capability and many failures prevented before service

The MRT state vector

A useful way to see the whole railway is to imagine its changing state collected into one large vector.

X(t) = [
train positions,
train speeds,
accelerations,
headways,
station dwell states,
platform queues,
train loads,
route availability,
power demand,
asset health
]

The exact internal representation differs among systems. The educational principle is that the railway is never described by one number.

State	Symbol	Why it matters	Specialist pillar
Train position	xi(t)	Determines where each train is and what movement remains possible	MRT timing
Velocity and acceleration	vi(t), ai(t)	Shape travel time, braking, energy and comfort	MRT braking
Train separation	hi(t)	Connects safety margin, frequency and capacity	MRT headway
Dwell	di,s	Allows passenger exchange but consumes line time	Station dwell
Platform queue	Qs(t)	Reveals whether passenger arrivals exceed boardable capacity	Passenger capacity
Train load	Li,s	Changes crowding, boarding, energy and downstream space	Passenger capacity
Network and route state	G(t)	Determines which paths and transfers are available now	Routes and transfers
Power	P(t)	Connects traction, regeneration, coasting and system demand	Energy optimisation
Asset health	Hj(t)	Indicates degradation, failure risk and maintenance need	Predictive maintenance
Track condition	T(x,t)	Maps geometry, defects and change along the rail	Track inspection
Wheel–rail interface	Cwr(t)	Transmits load, traction, braking and guidance	Wheel–rail contact
Network functionality	F(t)	Measures retained service and recovery during disruption	Network resilience

The railway observes part of this state through sensors and data:

Y(t) = observations from trains, tracks, stations,
passenger systems, power systems and maintenance records

It selects permitted controls:

U(t) = traction, braking, holding, dwell management,
routing, train insertion, withdrawal and maintenance actions

And it encounters disturbances:

W(t) = passenger surges, variable dwell, faults,
service constraints, weather, incidents and model error

A general control description is:

Ẋ(t) = f[X(t), U(t), W(t)]
Y(t)  = g[X(t)] + measurement noise

This is not an LTA or operator control formula. It is the standard systems idea: the state changes according to its present condition, chosen actions and external disturbances; measurements reveal the state imperfectly.

The MRT optimisation problem

A railway has multiple objectives:

• protect safety;
• carry passenger demand;
• reduce waiting and excessive crowding;
• keep headways regular;
• arrive within useful time windows;
• preserve accessibility;
• use energy efficiently;
• limit asset damage;
• retain recovery margin;
• and leave enough time and resources for maintenance.

Safety is not merely another preference to trade against convenience. It defines the feasible set.

Choose U(t) to minimise:

J = waiting + excess journey time + crowding
    + headway irregularity + energy
    + disruption loss + maintenance burden

subject to:

safe movement and braking constraints
train and station capacity
vehicle and power limits
route availability
accessibility requirements
maintenance and operating rules

The weights and exact constraints belong to authorised engineering and policy processes. The public lesson is that no single output owns the railway.

The MRT is a multi-objective system in which improving one number can damage another unless the whole field is checked.

The operational state sequence

The sequence below is a reader-safe synthesis. Real railways use more detailed operational states, procedures and authorities.

State	What is happening	Mathematical job
Planned service	Demand, timetable, fleet, power and engineering access are prepared before operation.	Forecast, schedule, simulate and test feasibility.
Service ready	Trains, staff, stations, routes and systems are available for the operating day.	Confirm initial conditions and resource readiness.
Viable operation	Trains and passenger flows remain inside normal working envelopes.	Regulate speed, headway, dwell and energy around the plan.
Stress	Demand rises, a dwell lengthens or a small constraint removes margin.	Absorb variation without losing safe, regular service.
Degradation	Headways distort, queues grow or an asset operates with reduced capability.	Prioritise function, limit propagation and preserve alternatives.
Incident or breach	A route, train, station or system cannot perform its ordinary role.	Enter an authorised safe state and define the remaining feasible network.
Stabilisation	Further loss is contained and a degraded service pattern is established.	Control queues, reroute passengers and stop cascading overload.
Recovery	Assets, trains and passenger flows are restored progressively.	Choose restoration order and rebuild regularity, not only hardware.
Return and learning	Actual outcomes are compared with plans and assumptions.	Update models, maintenance, training and future design.

The visible incident may last minutes or days. The mathematical preparation for it began years earlier in network design, redundancy, maintenance, staff training and data quality.

Prompt 1 — How does a human journey become a mathematical transport job?

The MRT should begin with neither the train nor the track.

It should begin with a person who needs to move.

At time t, passengers arrive at station s for direction d at rate:

λs,d(t) = passenger arrivals per unit time

But entry count alone is not enough. The system also needs to know where passengers are going.

An origin–destination matrix records demand:

Oij(t) = passengers travelling from station i to station j during time t

From this, the railway can estimate which track sections, trains, platforms and interchanges will carry the largest loads.

The same number of passengers can create very different operating problems.

Ten thousand passengers spread evenly across a day are not the same as ten thousand arriving within twenty minutes. Ten thousand short journeys across different sections are not the same as ten thousand journeys all crossing one central bottleneck. Ten thousand passengers with several route alternatives are not the same as ten thousand dependent on one interchange.

Demand therefore has at least four coordinates:

who or how many
× origin and destination
× direction
× time

Then the receiver enters.

A route that works for a fast walker with no luggage may not work equally for a wheelchair user, an older passenger, a parent with a stroller or a traveller carrying bags. The physical rail network can be the same while the feasible passenger network differs.

For passenger g, define a feasible set of paths:

Rg(t) = routes usable by passenger g at time t

A staircase edge may be finite for one passenger and effectively unavailable for another. A transfer may be theoretically possible but too uncertain for someone trying to reach an examination or flight.

The railway’s transport job is therefore not one average:

demand forecast
+ passenger capabilities
+ destination urgency
+ route availability
= service requirement

The specialist article How MRT Routes and Transfers Are Optimised Using Mathematics develops the graph and passenger-choice mathematics. The capacity article shows how origin–destination demand becomes section load.

The first mathematical object in an MRT is not a train. It is a journey that must remain possible for a receiver.

Prompt 2 — How does a timetable become safe moving separation?

A timetable is an event network.

For train i at station k:

Departure = arrival + dwell
Di,k = Ai,k + di,k

Next arrival = departure + running time
Ai,k+1 = Di,k + ri,k

Those two equations can generate a line of scheduled events.

They do not yet guarantee that two trains can use the same railway safely.

Successive trains must maintain permitted separation. At a reference point, scheduled headway is:

Hi = ti − ti−1

and must remain above the relevant safe and operational minimum under the current conditions.

LTA’s public description of the renewed North-South and East-West Line signalling says the CBTC system has technical capability for 100-second headways, compared with 120 seconds before 2018. That does not mean every train on every day should operate at exactly 100 seconds. Technical capability is one boundary inside a larger operating field containing stations, terminals, power, passenger exchange, fleet and recovery margin.

The train also needs a safe speed trajectory.

v(t) = dx/dt
a(t) = dv/dt

Position is obtained by integrating speed:

x(t) = x0 + ∫v(t)dt

A simplified constant-deceleration stopping distance is:

s ≈ v²/(2b)

Real braking envelopes include gradients, response times, adhesion, uncertainty, vehicle performance, safety margins and system-specific rules. The public equation is a starting intuition: braking distance grows with the square of speed.

The signalling and control system therefore does not merely ask:

Where is the train?

It also asks:

What movement is safely available now?
What braking must remain possible?
What is the train ahead doing?
What route is set?
What station and platform state is approaching?

This creates a closed loop:

measure position and speed
→ compare with permitted trajectory
→ command traction, coasting or braking
→ train moves
→ measure again

See How MRT Timing Works Using Mathematics, How MRT Headway Works Using Mathematics and How an MRT Timetable Is Built Using Mathematics.

The timetable says when the railway wants an event. The control system decides how the real train can still reach that event safely.

Prompt 3 — How does electricity become motion, braking and wheel–rail guidance?

A train moves because forces change its velocity.

A simplified longitudinal balance is:

meff dv/dt
= Ftraction − Fbrake − R(v) − mg sinθ

The traction motors must overcome resistance and gradient while producing the required acceleration.

Mechanical power is:

Pmechanical = Ftraction v

Electrical energy over a journey is the integral of power:

E = ∫P(t)dt

The train stores kinetic energy:

Ek = ½mv²

During regenerative braking, part of that energy can be converted back into electricity and used by other receptive loads, including another accelerating train where the power-system conditions permit.

Yet the motors do not push against empty space.

Motor torque becomes useful only when wheel–rail contact transmits tangential force.

|Ft| ≤ μN

This simplified adhesion condition says available tangential force is bounded by effective friction and normal load.

The steel wheel and rail deform elastically at a small contact patch. Average contact pressure is roughly:

pmean = N/Acontact

Inside the patch, microscopic relative motion called creepage helps produce traction and lateral guidance. Wheel tread geometry creates different effective rolling radii as a wheelset moves laterally, helping it negotiate curves. The same geometry also participates in lateral dynamics and wear.

So one acceleration event activates several mathematical layers:

electrical power
→ motor torque
→ wheel force
→ elastic contact
→ creepage and adhesion
→ acceleration
→ new train position
→ new timetable and headway state

Braking reverses much of the chain while maintaining stopping accuracy and passenger comfort. Jerk, the rate of change of acceleration, matters because abrupt changes can be uncomfortable even when peak acceleration remains acceptable:

j = da/dt

The deep dives are How MRT Braking Works Using Mathematics, How MRT Energy Use Is Optimised Using Mathematics and How MRT Wheel–Rail Contact Works Using Mathematics.

The train moves only when electricity, force, friction, geometry and time agree at the same instant.

Prompt 4 — How do stations turn moving trains into usable passenger capacity?

A train that never stops cannot serve stations.

A train that stops too long can reduce line capacity.

Dwell time therefore sits between two requirements:

enough time for safe passenger exchange
but
not so much uncontrolled time that headway and capacity collapse

The Ministry of Transport stated in November 2025 that typical MRT and LRT dwell ranges are around 35–60 seconds at interchange stations and 30–50 seconds at non-interchange stations, depending on operating conditions and passenger loading. Those seconds are a major part of the railway, not a pause outside it.

Let Q_k be the passenger queue after train k departs. During the next headway H_k, passengers arrive at rate λ. If B_k+1 passengers can board the next train:

Qk+1 = max(0, Qk + λHk − Bk+1)

If passenger arrivals repeatedly exceed boardable space, the queue grows from train to train.

Train capacity and frequency combine as:

Ctheoretical = Ptrain × 3600/H

where P_train is passenger places per train and H is headway in seconds.

But theoretical capacity is not necessarily delivered capacity.

If trains are irregular, one large gap collects a large passenger batch while the following short gap collects few. The first train becomes overloaded; the second may carry unused space. Under standard random-arrival assumptions, expected waiting for irregular headways is:

E[W] = E[H²] / (2E[H])

This contains headway variance inside E[H²]. Two services with the same average headway can therefore create different waiting times.

Crowding can also change dwell.

longer gap
→ more waiting passengers
→ heavier boarding
→ longer dwell
→ still longer following gap

This is positive feedback and one pathway into train bunching.

Doors create another bottleneck. If door j processes passengers at effective rate μ_j, exchange time is often controlled by the busiest door:

Texchange ≈ maxj(Nj/μj)

Several quiet doors cannot fully compensate for one congested doorway if the whole train must wait for that final flow to clear.

Accessibility belongs inside this calculation. A dwell that works only for fast, unencumbered passengers is not a successful public dwell. The station must allow safe exchange for elderly commuters and passengers using wheelchairs or mobility aids. The human requirement changes the feasible mathematical envelope.

Continue to How MRT Station Dwell Time Works Using Mathematics and How MRT Passenger Capacity Is Calculated Using Mathematics.

A station is where railway time becomes human access. If passenger exchange fails, the line can lose capacity even while every train remains mechanically healthy.

Prompt 5 — How does the network choose routes, transfers and alternative paths?

An MRT map can be represented as a graph:

G = (V,E)

Stations are nodes V. Rail connections and transfer walkways are edges E.

Each edge has a cost. But “cost” need not mean money.

Cr = βride Tride
   + βwait Twait
   + βwalk Twalk
   + βtransfer Ntransfers
   + βcrowd Xcrowd
   + βrisk Xrisk
   + βaccess Xaccess

The fastest in-train route may lose once transfer walking and waiting are included. The geographically shortest route may be crowded. The route with the lowest average time may have greater uncertainty. A path involving stairs may be unusable for one passenger.

The correct route is therefore receiver- and state-dependent:

best route = arg min Cr(passenger, time, network state)

Route costs also change when passengers respond to them. If everyone is sent to the same apparent shortcut, it can become crowded and lose its advantage.

recommend route
→ passengers move
→ loads change
→ waiting and crowding change
→ route cost changes

This is why passenger assignment and network operation interact.

Circle Line Stage 6 provides a current Singapore example of graph change. The three-station extension between HarbourFront and Marina Bay opened on 12 July 2026, closing the Circle Line loop and creating more direct and alternative routes. The completed line connects to all existing MRT lines, changing travel times, transfer counts, crowding patterns and disruption alternatives across the wider network.

A new edge has effects far beyond the stations it touches:

new connection
→ new shortest paths
→ fewer transfers for some journeys
→ different passenger assignment
→ different section loads
→ different network resilience

But an alternative path that exists on a map may still be weak. It may lack spare train capacity, require difficult transfers, or share the same dependency as the failed route.

The specialist route article develops generalised cost, time-dependent graphs and passenger choice. The resilience article asks whether the alternatives remain useful during failure.

A railway network is not resilient because another line can be drawn between two stations. It is resilient when enough passengers can actually use the alternative under the disrupted conditions.

Prompt 6 — How does the MRT use energy without simply slowing everything down?

The lowest-energy train is one that never moves.

That is why energy minimisation must remain attached to transport function.

A station-to-station journey can often be completed with several feasible speed profiles.

hard acceleration → high cruise → late braking

or

appropriate acceleration → shorter cruise → coasting → braking

If the timetable contains usable running-time margin, coasting can reduce positive traction energy with little or no passenger-visible lateness.

During regenerative braking, a train can return electrical energy. Useful recovery depends on conversion efficiency and whether the power network has a receptive load at that moment.

Let train power P_i(t) be positive during traction and negative during regeneration.

Pnet(t) = Σi Pi(t) + losses + other loads

If one train brakes while another accelerates on an electrically compatible part of the network, their power curves can partly offset. Timetable timing therefore changes electrical efficiency.

Singapore’s Green CBTC programme makes this connection visible. SMRT currently reports around 17 GWh in annual energy savings from Phase 1 and says the overall project is intended to achieve up to a 15 per cent reduction in traction energy on the North-South and East-West Lines.

The deeper lesson is multiplication:

small energy improvement per movement
× many interstation runs
× many trains
× many operating days
= large annual effect

But energy cannot be optimised alone.

• More coasting can use recovery margin.
• Faster running can consume more energy but recover delay.
• Longer journeys extend auxiliary energy use.
• Lower headway can require more active trains and change power peaks.
• Passenger loading changes vehicle mass and energy per passenger.
• Regenerative overlap depends on the timetable of other trains.

The energy article follows these relationships in detail.

Energy efficiency is not making the MRT do less. It is arranging time, motion and electrical flow so less energy is wasted while the same public job is performed.

Prompt 7 — How does the MRT know whether its own infrastructure is becoming unhealthy?

A railway cannot rely only on visible failure.

Many components produce evidence before they fail:

• temperature drift;
• changed vibration;
• longer door movement time;
• unusual current draw;
• track-geometry change;
• rail-surface irregularity;
• diagnostic-event patterns;
• and deviations from comparable assets.

A sensor gives a measurement:

observed signal = physical condition + measurement noise

An anomaly score asks whether the observation is unusual relative to a healthy baseline. A z-score is the simplest example:

z = (x − μ)/σ

But unusual is not identical to faulty. Context, persistence, multiple sensors and engineering inspection matter.

Predictive maintenance adds a future question:

Given the current evidence,
what is the probability of unacceptable degradation
before the next maintenance opportunity?

Bayesian updating gives the conceptual structure:

P(fault | data)
= P(data | fault)P(fault) / P(data)

Remaining Useful Life treats future failure or intervention time as uncertain, not exact:

RUL = predicted threshold time − current time

The output should include a range or probability, not merely a confident-looking date.

Singapore is extending this capability. The Rail Reliability Taskforce recommendations announced in February 2026 include a standardised network-wide approach to condition monitoring and stronger predictive-maintenance capability.

SMRT’s JARVIS platform integrates maintenance, condition-monitoring and operational data. In August 2026, SMRT reported that JARVIS had helped initiate more than 500 maintenance inspections in response to early signs of component degradation.

Track inspection gives the same logic a spatial form.

T(x,t) = measured track condition at location x and time t

If a train moves at speed v and a sensor samples at frequency f, average spatial sample spacing is:

Δx = v/f

Repeated passes turn one observation into a trend. Cameras, vibration and geometry can be fused. The North East Line Automatic Track Inspection system publicly demonstrates this approach: onboard cameras and sensors inspect track condition during normal operation, complementing physical inspection.

Wheel and rail profiles also evolve. Contact changes wear; wear changes geometry; geometry changes future contact. Inspection and maintenance close that loop.

measure
→ detect change
→ estimate significance
→ prioritise
→ engineer inspects
→ maintain
→ measure again

The two specialist pillars are Predictive Maintenance and Track Inspection.

The best fault is often the one that becomes an ordinary maintenance job while it is still only a change in the data.

Prompt 8 — How does the MRT remain useful when something goes wrong?

No railway can guarantee that no train, component, station or route will ever be disrupted.

Resilience begins with an honest premise:

failure probability can be reduced
but not made universally zero

The question is what happens next.

Suppose one train loses delay δ. The following train may encounter reduced running freedom. A longer gap ahead attracts more passengers. A shorter gap behind attracts fewer. Dwell and load diverge. Delay can therefore propagate through both operational constraints and passenger feedback.

small delay
→ headway distortion
→ uneven passenger accumulation
→ unequal dwell
→ bunching tendency
→ wider service impact

The delay article explains how regulation, holding, running-time flexibility and recovery margin can return the line towards a useful rhythm.

At network scale, resilience asks how much mobility remains after a loss and how quickly it returns.

Let normal functionality be F₀ and actual functionality during disruption be F(t).

Resilience over interval T
= (1/TF0) ∫0T F(t)dt

The area under the curve rewards both a shallow loss and fast recovery.

But the definition of functionality matters.

If F(t) counts stations open, it may miss severe passenger delay. If it counts average journey time, it may hide a smaller group whose journey became impossible. A passenger-centred measure should examine served demand, excess travel time, accessible alternatives, queue growth and unserved journeys.

Alternative capacity is crucial:

reserve on alternative route
= usable capacity − normal passenger load

A structurally connected network can still fail functionally if the remaining path cannot absorb rerouted passengers.

Recovery also continues after the physical fault is cleared. Trains may be displaced. Queues may remain. Headways may still be irregular. Passenger recovery time can exceed component repair time.

fault contained
→ safe movement restored
→ degraded service stabilised
→ train sequence rebuilt
→ queues cleared
→ normal service verified

Continue to How MRT Delays Propagate and Recover Using Mathematics and How MRT Network Resilience Is Measured Using Mathematics.

A resilient MRT is not one that never loses anything. It is one designed so that a local loss does not automatically become the loss of the whole passenger journey.

One passenger journey activates the whole railway

Follow one passenger from origin to destination and the entire cloud becomes visible.

Journey stage	Visible action	Hidden mathematical work
1. The need	A passenger decides to travel.	The journey joins an origin–destination demand field by place, direction and time.
2. Route selection	The passenger checks a map or app.	A time-dependent graph compares riding, walking, waiting, transfers, crowding and accessibility.
3. Station access	The passenger enters through an entrance and fare gate.	Pedestrian flow, lift/escalator capacity, fare validation and platform assignment interact.
4. Platform arrival	The passenger waits.	The queue changes with arrival rate, headway and the boardable space on the next train.
5. Train approach	The information display counts down.	Position and speed observations are converted into an estimated arrival under current constraints.
6. Braking	The train slows into the station.	A safe trajectory balances braking distance, jerk, adhesion, gradient and stopping accuracy.
7. Doors and dwell	Passengers alight and board.	Doorway flows, crowding, accessibility and timetable margin determine useful dwell.
8. Departure	The train accelerates.	Traction power, wheel–rail force, headway and movement authority must remain compatible.
9. Interstation run	The passenger rides.	Speed, coasting, energy, train separation and track condition are continuously managed.
10. Transfer	The passenger changes lines.	Walking geometry, next-train timing, connection probability and route cost are recalculated.
11. Disruption possibility	A service update appears.	The graph changes, alternatives are evaluated and passenger loads begin redistributing.
12. Arrival	The passenger reaches the destination station.	The delivered journey is compared implicitly with the promised service and predicted time.
13. Exit	The passenger leaves the network.	Aggregated journey data contribute to future demand and capacity models under applicable governance.
14. After the passenger	The train continues.	Energy, wear, diagnostic events and timetable performance return to maintenance and planning systems.

The passenger’s journey may take thirty minutes.

The capability required to make it ordinary was built over decades and must be renewed every night.

The ticket buys one journey. The railway supplies an inherited civilisation of standards, engineering, mathematics and maintained trust.

One train cycle activates a second cloud

The passenger sees one segment. The train must complete a repeating cycle.

depot or stabling
→ enter service
→ terminal departure
→ repeated run–brake–dwell sequence
→ terminal turnback
→ return journey
→ frequency transition or route change
→ leave service
→ inspect, clean, maintain and prepare again

If full cycle time is T_cycle and service headway is H, a simplified active train requirement is:

N = ceil(Tcycle/H)

This explains why a small headway reduction can require several more trains. It also explains why dwell, running time and terminal turnaround are fleet variables.

The ceiling function matters. A railway cannot operate 31.4 trainsets. Crossing an integer boundary changes the real fleet requirement.

One train saved from the cycle may result from many small improvements distributed across the line. One additional train required may result from a few seconds added repeatedly at every station.

The train fleet is continuous mathematics forced to purchase, maintain and staff whole vehicles.

The MRT Mixer — no variable changes alone

The central lesson of the hub is coupling.

Proposed change	Direct benefit	What else must be checked
Reduce headway	More trains and passenger places per hour	Signalling, dwell, terminals, fleet, power, margin and regularity
Increase dwell	More time for safe, accessible boarding and alighting	Headway, journey time, following trains and terminal cycle
Run faster	Shorter journey or more recovery ability	Energy, braking, comfort, wheel–rail force, noise and timetable interaction
Coast more	Lower traction energy	Arrival time, recovery margin and interaction with other trains
Fill trains more densely	More passengers per movement	Door flow, dwell, accessibility, comfort and downstream boarding
Add a route or interchange	More direct journeys and redundancy	Transfer geometry, new crowding patterns, operating complexity and capacity
Increase inspection sensitivity	Earlier detection of possible faults	False alarms, engineering workload, data quality and actionability
Defer maintenance	More short-term operating or engineering availability	Future reliability, accumulated risk, asset life and recovery capability
Optimise every train independently	Locally efficient trajectories	Network power peaks, regenerative overlap, headway and system optimum
Optimise average passenger time	Good aggregate score	Passengers with high transfer, accessibility or disruption cost

A local improvement becomes a system improvement only after the return path is checked.

proposed change
→ immediate effect
→ neighbouring systems
→ passenger receiver
→ delayed consequences
→ maintenance and recovery
→ measured world return

The mathematics stack inside one MRT

Mathematical field	MRT job	Example
Arithmetic	Counts trains, passengers, stations, minutes and resources	Trains per hour, capacity totals, fleet counts
Algebra	Connects variables and constraints	Headway, running time, dwell, force and load equations
Geometry	Represents track, curves, platforms and contact	Curve radius, wheel profile, transfer distance
Trigonometry	Handles gradients, angles and oscillatory signals	Incline force, phase and wave analysis
Calculus	Connects position, speed, acceleration, energy and accumulated loss	x′=v, v′=a, E=∫Pdt, resilience area
Probability	Represents uncertain demand, failures, connections and predictions	P(fault|data), missed-transfer probability
Statistics	Separates pattern from noise and evaluates performance	Headway variance, anomaly scores, confidence intervals
Graph theory	Represents stations, links, routes and redundancy	Shortest paths, centrality, alternative routes
Optimisation	Selects among competing feasible choices	Timetable, energy, maintenance and passenger assignment
Control theory	Regulates a changing physical system around desired states	Speed control, headway recovery, closed-loop operation
Queueing theory	Models passengers waiting for finite service	Platform accumulation and denied boarding
Reliability engineering	Models failure, availability and repair	Hazard rates, redundancy, remaining useful life
Signal processing	Extracts useful information from sensor streams	Vibration spectra and track irregularity
Computer vision	Turns images into located condition evidence	Rail, fastener and surface inspection
Behavioural modelling	Represents how passengers choose and respond	Transfer penalties, crowding sensitivity, route choice

The MRT does not keep these subjects in separate classrooms.

A platform queue becomes a dwell-time problem. Dwell becomes a headway problem. Headway becomes a passenger waiting problem. A faster recovery trajectory becomes an energy and wheel–rail problem. A new route becomes a crowding and resilience problem. A sensor anomaly becomes a maintenance scheduling problem.

The railway is what happens when the mathematics syllabus is forced to cooperate.

The complete MRT mathematics cloud

This hub owns the synthesis. The pages below own the specialist jobs.

Pillar	Canonical question	What it contributes to the whole
How MRT Timing Works Using Mathematics	How are speed, braking, dwell, headway and control fitted into seconds?	The first whole-line timing model and original series entry.
How MRT Braking Works Using Mathematics	How does a train know when and how to stop?	Stopping distance, braking curves, gradients, jerk, regeneration and platform accuracy.
How MRT Headway Works Using Mathematics	How closely can successive trains operate?	Separation, frequency, capacity, variance, passenger waiting and bunching.
How MRT Station Dwell Time Works Using Mathematics	Why can doors limit a whole railway?	Boarding, alighting, doorway flow, accessibility and dwell feedback.
How MRT Delays Propagate and Recover Using Mathematics	How does one lost minute move through the line?	Primary and secondary delay, headway distortion, regulation and recovery margin.
How an MRT Timetable Is Built Using Mathematics	How does a city become a sequence of events?	Demand, event networks, running and dwell times, terminals, fleet and engineering access.
How MRT Energy Use Is Optimised Using Mathematics	How can energy fall without simply slowing the railway?	Power curves, coasting, regenerative braking, receptivity and multi-train coordination.
How MRT Passenger Capacity Is Calculated Using Mathematics	Why is a full train not necessarily a full railway?	Passengers per train, pphpd, section loads, queues, stations and delivered capacity.
How MRT Routes and Transfers Are Optimised Using Mathematics	Why is the shortest path not always the best journey?	Graph theory, generalised cost, transfer probability, crowding and accessibility.
How MRT Network Resilience Is Measured Using Mathematics	What remains when part of the railway is lost?	Redundancy, functional alternatives, cascading overload and recovery curves.
How MRT Predictive Maintenance Works Using Mathematics	Can degradation be found before it becomes disruption?	Anomaly detection, hazard, remaining useful life, uncertainty and maintenance decisions.
How MRT Track Inspection Works Using Mathematics	How does a railway turn kilometres of rail into a health map?	Geometry, sampling, computer vision, sensor fusion and spatial degradation.
How MRT Wheel–Rail Contact Works Using Mathematics	How do two pieces of steel carry, guide and stop a train?	Contact pressure, adhesion, creepage, conicity, curves, wear and profiling.
How MRT Noise and Vibration Work Using Mathematics	How can microscopic roughness become sound and vibration?	Decibels, spectra, resonance, curve squeal, transfer paths and mitigation.
How MRT Power Supply Works Using Mathematics	How does electricity follow moving train loads?	Traction substations, voltage drop, regeneration, peak demand, protection and redundancy.
How MRT Signalling and Train Regulation Work Using Mathematics	How does the railway turn uncertain position into safe real-time movement?	CBTC, movement authority, ATP, ATO, headway regulation and safe recovery.
How MRT Platform Screen Doors Work Using Mathematics	Why do centimetres of stopping error matter at a station?	Alignment, synchronisation, obstruction detection, passenger flow and door reliability.
How MRT Tunnel Ventilation and Airflow Work Using Mathematics	Why does an underground train behave like a moving piston?	Pressure, airflow, heat, fan energy, platform boundaries and smoke control.
How MRT Depots and Fleet Operations Work Using Mathematics	How is tomorrow morning’s operating fleet built overnight?	Fleet availability, circulation, stabling, maintenance, launch sequencing and reserve.
How MRT Tunnels Are Built Using Mathematics	How do underground drives remain on the same three-dimensional railway map?	Alignment, surveying, TBM steering, geology, settlement, lining and as-built geometry.
How MRT Stations Are Built Using Mathematics	How does a deep excavation become a passenger-flow machine?	Retaining structures, groundwater, construction staging, interchange interfaces and evacuation.
How MRT Escalators and Lifts Work Using Mathematics	How can one unavailable lift disconnect a station for one passenger?	Vertical capacity, batch queues, accessibility graphs, energy, redundancy and reliability.
How MRT Drainage and Flood Protection Work Using Mathematics	How does the underground railway keep rainfall outside and remove water that enters?	Runoff, crest levels, barriers, sump storage, pumping, sensing, redundancy and flood resilience.
How MRT Communications and Passenger Information Work Using Mathematics	How does a changing railway state become a timely passenger decision?	Information age, ETA prediction, latency, consistency, multi-channel guidance and disruption feedback.
How MRT Fare Systems and Origin–Destination Data Work Using Mathematics	How does a tap become a fair fare and useful city-scale demand evidence?	Transaction pairing, distance fares, transfer rules, account-based ticketing, OD matrices and privacy-aware planning.
How MRT Rail Grinding and Wheel Profiling Work Using Mathematics	Why can removing wheel or rail material make the railway last longer?	Profile restoration, corrugation, material removal, maintenance timing, roughness and post-maintenance verification.

The similarly titled duplicate braking post is intentionally not promoted here. The canonical braking owner is the clean URL without the “-2” suffix.

Where this fits in the wider eduKateSG library

The failure shadows

Every successful mathematical function casts a failure shadow when it is lost.

Mathematical function preserved	Failure shadow when lost
Position and speed remain observable	The controller cannot place the train confidently inside the movement field.
Braking remains inside a validated envelope	Stopping accuracy, safe separation or service speed must be reduced.
Headway remains regular	Passenger loads become uneven, waiting increases and bunching can amplify.
Dwell remains sufficient and bounded	Passengers cannot exchange safely, or station delay propagates through the line.
Capacity is measured by section, direction and time	A comfortable average hides the true bottleneck.
Routes include transfers, crowding and accessibility	The “shortest” path becomes unusable or misleading for real passengers.
Energy is optimised with timetable and reliability	Local savings create lateness, weak recovery or harmful power peaks.
Track and asset health remain observable	Degradation reaches operation before maintenance knows where to act.
Alarms carry uncertainty and consequence	False confidence either floods engineers or misses meaningful precursors.
Alternative routes retain spare capacity	A map remains connected while displaced passengers cannot be absorbed.
Repair capacity is protected	A bounded fault becomes a long disruption because people, parts or access are missing.
World Return remains honest	Models improve on paper while passenger experience and asset condition deteriorate.

Failure geometry: how a local MRT problem becomes a system problem

Geometry	How it travels	MRT example pattern
Serial chain	One required input is lost and downstream functions stop.	Power constraint → train movement constraint → reduced frequency → passenger queues.
Headway feedback	A gap changes passenger accumulation, which changes dwell and enlarges the gap.	Delay → crowding → longer dwell → further delay.
Convergent overload	Passengers from several lost paths move onto one surviving route.	Disrupted line → alternative interchange → platform and train overload.
Common dependency	Apparently separate systems share one vulnerable input.	Several functions depend on the same power, communications or data layer.
Spatial coupling	Co-located assets experience the same physical event.	One corridor affects track, station access and nearby surface transport.
Maintenance cascade	Deferred work raises faults, which consume more emergency work and reduce planned maintenance.	Backlog → failures → reactive repairs → larger backlog.
Information lag	Passenger and operator decisions use a state that has already changed.	Outdated route guidance sends more passengers towards a constrained interchange.
Receiver isolation	The system reports service, but a passenger cannot access the remaining path.	An alternative exists but its transfer is inaccessible or impractical.
Recovery shadow	The fault clears but displaced trains and queues keep service degraded.	Component restored → headway still irregular → passenger recovery continues.

These patterns explain why “what failed first?” and “what caused the final passenger loss?” may have different answers.

Trigger, constraint, buffer, breach and recovery are different objects

Object	MRT question
Trigger	What event changed the service, asset or passenger field?
Constraint	Which safe movement, capacity, route or maintenance limit became binding?
Exposure	Which passengers, trains, stations and journeys were in the affected field?
Vulnerability	Why could the event create large loss here?
Buffer	Which timetable margin, spare capacity, alternative route or standby resource absorbed part of it?
Breach	Which necessary service function could no longer be sustained?
Propagation	How did delay, crowding, lost capacity or information travel?
Recovery resource	Which people, parts, procedures, vehicles and routes enabled restoration?
World Return	What evidence showed whether the response actually worked for passengers?

A train fault does not automatically explain a network-wide passenger impact. The mechanism between the first event and the received consequence must be traced.

Twelve false diagnoses this MRT hub must refuse

1. “The fastest train makes the best MRT.”

Passenger access, stopping, dwell, waiting, energy, comfort and reliability belong to the journey too.

2. “The shortest possible headway should always be used.”

A technical minimum may leave too little margin for stations, terminals, energy, fleet and recovery.

3. “A full train means the entire line is full.”

Capacity varies by carriage, section, direction, time and train spacing.

4. “The timetable is the service.”

The timetable is a plan. Actual trains, queues and passenger journeys determine delivery.

5. “Average headway tells us average waiting.”

Irregularity increases passenger-weighted waiting because long gaps collect more arrivals.

6. “The shortest route is the best route.”

Walking, waiting, transfers, crowding, reliability and accessibility can reverse the answer.

7. “Energy saving means slowing trains.”

Speed profiles, coasting, regeneration, power timing and recovery margin can improve energy without simply degrading service.

8. “More automation removes the need for people.”

Automation shifts human work towards design, supervision, maintenance, verification, incident response and accountable authority.

9. “More data automatically means better maintenance.”

Unstandardised, noisy, drifting or unlabelled data can create false confidence and alarm overload.

10. “An anomaly is a diagnosis.”

An anomaly is evidence requiring context and engineering verification.

11. “A connected map is a resilient network.”

The remaining paths need capacity, acceptable time, accessible transfers and independence from the same failure.

12. “A good reliability score proves every passenger journey worked.”

Aggregate performance can hide particular stations, periods, groups or high-consequence journeys.

A railway model becomes dangerous when one successful metric is allowed to impersonate the whole service.

The MRT deletion tests

Remove one component mentally and observe what the railway can no longer do.

• Remove passenger demand: the timetable may be feasible but no longer has a reason for its frequency pattern.
• Remove position measurement: safe automatic control loses its present state.
• Remove braking margin: speed can no longer be separated safely from uncertainty.
• Remove headway: one train can be planned without regard for the next.
• Remove dwell: the railway moves vehicles without serving passengers.
• Remove station flow: platforms and doors are assumed able to process unlimited people instantly.
• Remove route choice: every passenger is forced onto one path regardless of transfers or accessibility.
• Remove energy: a physically feasible timetable is assumed to have unlimited electrical consequence.
• Remove wheel–rail contact: motors have no mechanism for transmitting useful force to the track.
• Remove inspection: physical degradation remains invisible until operation reveals it.
• Remove maintenance windows: assets are expected to remain healthy without time to renew them.
• Remove human authority: models can recommend actions but nobody is accountable for safe decisions.
• Remove alternatives: one local closure becomes a complete passenger dead end.
• Remove World Return: the system can report success without learning whether trains, assets or passengers agreed.

The deletion test reveals why the MRT is not one machine.

It is a field of mutually necessary capabilities.

The MRT system audit

Use this audit when evaluating a railway proposal, service change, disruption or mathematical claim.

1. What passenger function must remain possible?
2. Which receiver is being modelled?
3. At what station, section, direction and time?
4. What is directly observed, estimated or assumed?
5. What is the relevant state vector?
6. Which safety constraints define the feasible set?
7. Which variable is being optimised?
8. What other variables may worsen?
9. Is the capacity theoretical, scheduled, delivered or used?
10. Does the model include headway variance, not only average headway?
11. Does dwell include real passenger and accessibility requirements?
12. Is the route feasible for the actual passenger?
13. Does an alternative route have spare capacity?
14. What timetable or fleet margin is being consumed?
15. What energy and power consequences follow?
16. What wheel, rail or asset consequences follow over time?
17. How are measurement noise and uncertainty represented?
18. What would create a false alarm or missed failure?
19. Which human authority verifies and acts?
20. What happens if one input is wrong?
21. How can a local problem propagate?
22. What degraded safe mode remains?
23. What people, parts and engineering access are required for recovery?
24. How will passenger recovery be measured after the physical fault clears?
25. What evidence from the world would make the model change?

The MRT gate

PASSENGER JOURNEY REQUIRED
          ↓
ORIGIN, DESTINATION, TIME AND RECEIVER KNOWN?
          ↓
FEASIBLE ROUTE EXISTS?
          ↓
DEMAND WITHIN DELIVERABLE NETWORK CAPACITY?
          ↓
TIMETABLE, FLEET, STATION AND POWER FIT?
          ↓
SAFE MOVEMENT AND BRAKING ENVELOPE AVAILABLE?
          ↓
RUN → OBSERVE → REGULATE
          ↓
HEADWAY, DWELL, QUEUE, ENERGY AND ASSET HEALTH STABLE?
   YES → CONTINUE SERVICE → MEASURE PASSENGER RETURN
   NO  → IDENTIFY BINDING CONSTRAINT
                     ↓
CAN MARGIN OR SAFE DEGRADED MODE ABSORB IT?
   YES → STABILISE → RECOVER → VERIFY
   NO  → RECONFIGURE ROUTES / CAPACITY / SERVICE
                     ↓
REPAIR CAPABILITY PRESENT?
   YES → RESTORE → CLEAR RESIDUAL QUEUES → LEARN
   NO  → WIDER MOBILITY FAILURE

The gate is not an operating procedure. It is a public reasoning map that prevents one number from running ahead of the whole system.

The World Return — how the MRT proves that it works

The railway begins with predictions.

• Predicted passenger demand
• Scheduled arrival and departure
• Expected running and dwell time
• Predicted energy use
• Expected track and asset condition
• Predicted route choice and crowding
• Expected recovery after disturbance

Then the world returns measurements.

prediction ŷ
observation y
error e = y − ŷ

The error is not automatically failure.

It is information.

If dwell is repeatedly longer than predicted at one interchange, demand or door-flow assumptions may be wrong. If energy exceeds prediction on one section, the speed profile, gradient, load or regenerative receptivity model may need revision. If passengers avoid the recommended route, the transfer penalty or crowding estimate may be wrong. If track condition worsens faster than expected, maintenance interval or degradation assumptions may be weak.

The return loop is:

observe city and railway
→ build model
→ plan service
→ operate
→ measure trains, assets and passenger outcomes
→ compare
→ diagnose mismatch
→ change model, maintenance, timetable or infrastructure
→ operate again

This is what makes the MRT a learning system rather than a fixed machine.

But the receiver must remain visible.

A train can be recorded as operated while passengers were unable to board. A route can be declared available while an essential lift path was unavailable. A fault can be repaired while accumulated queues remain. Delivered is not identical to received.

The final output of the MRT is not the movement of rolling stock. It is the passenger journey that actually arrived.

What this article does not prove

• It does not reproduce Singapore MRT internal control logic, safety parameters or restricted operating procedures.
• It does not claim one universal equation governs every rail line or train type.
• Its equations are educational abstractions unless explicitly tied to a public source.
• Technical capability is not the same as the timetable used in every operating period.
• Maximum train capacity is not the same as comfortable, scheduled or delivered capacity.
• Mathematical optimisation does not remove policy, ethical or accessibility choices.
• Automation does not eliminate human engineering judgement or authorised responsibility.
• Network connectivity does not guarantee sufficient alternative capacity.
• An anomaly does not prove a physical fault.
• Reliability statistics do not describe every individual passenger experience.
• More complexity is not automatically better; every dependency needs maintenance and recovery.
• No public explanation replaces professional railway engineering or operator instructions.

The model should bend when the railway disagrees. The railway should never be forced to fit the model.

Frequently asked questions

How does an MRT work in simple terms?

It coordinates trains, tracks, signalling, stations, power, passenger demand and maintenance so trains can move safely at useful frequencies and passengers can complete journeys through the network.

Why is mathematics necessary for an MRT?

The railway must calculate or estimate position, speed, braking, time, separation, passenger queues, capacity, routes, power, failure risk and maintenance. Mathematics makes these relationships testable and coordinatable.

What is headway?

Headway is the time separation between successive trains at a reference point. It helps determine frequency, capacity and passenger waiting, while remaining subject to safe and operational constraints.

Why can MRT doors delay an entire line?

Passenger exchange determines dwell. A longer dwell changes the following headway, which changes passenger accumulation and can propagate delay to later trains and stations.

How is MRT passenger capacity calculated?

A theoretical starting point multiplies passenger places per train by trains per hour. Practical capacity also depends on station flow, section and direction, regularity, passenger distribution, accessibility and delivered service.

Why is the shortest MRT route not always fastest?

Transfer walking, next-train waiting, crowding, reliability and accessibility can make a geographically shorter path take longer or impose greater journey burden.

How does regenerative braking save energy?

During braking, traction motors can act as generators and return part of the train’s kinetic energy to the electrical system for use by receptive loads, subject to conversion and network conditions.

How can an MRT predict maintenance needs?

Condition-monitoring data are compared with healthy patterns. Trends, anomaly models, failure probabilities and remaining-useful-life estimates help engineers decide when inspection or maintenance is justified.

Does automation mean the MRT has no human operators?

No. Automated movement still depends on people who design, test, supervise, maintain, regulate, respond to incidents, communicate with passengers and carry legal and professional responsibility.

What makes an MRT resilient?

Reliability, safe degraded modes, alternative routes, spare capacity, trained people, parts, clear authority, truthful information and effective recovery all matter. A route that exists but cannot absorb passengers is weak resilience.

Where the MRT cloud grows next

This hub is designed to expand without changing its canonical job. Future specialist pillars can deepen:

• the mathematics of future MRT network expansion — how new lines, stations and interchange edges change accessibility, demand, capacity, resilience and long-term system value.

Each new article should claim one precise job, link back to this hub and avoid repeating the established pillars.

Research and further reading

Each specialist pillar contains its own deeper bibliography. The sources below support the whole-system synthesis and current Singapore layer.

• Land Transport Authority — Singapore Rail Network: current operating scale, ridership, hours and typical service frequencies.
• Land Transport Authority — Rejuvenating the Railway: NSEWL signalling, power, track circuits, trains and condition monitoring.
• Ministry of Transport — Dwell Times for Trains: current public dwell-time ranges and accessibility rationale.
• Land Transport Authority — Circle Line Stage 6: completion of the loop and additional travel alternatives.
• Land Transport Authority — Circle Line: line characteristics, rolling stock and public capacity information.
• SMRT — Decarbonisation and Green CBTC: current energy-saving results and traction-energy objective.
• LTA, SMRT and SBS Transit — Rail Reliability Taskforce recommendations: renewal, condition monitoring, predictive maintenance and service recovery.
• Land Transport Authority — North East Line condition monitoring and Automatic Track Inspection.
• SMRT — JARVIS and AI-enabled maintenance: integrated asset data and early maintenance inspections.
• Li and colleagues — Joint optimal train regulation and passenger-flow control: coupling train operation with passenger management.
• Yin and colleagues — Dynamic passenger-demand-oriented metro train scheduling: time-dependent demand and timetable optimisation.
• Yin and colleagues — Energy-efficient metro train rescheduling under uncertainty: passenger delay, energy and stochastic operation.
• Mukherjee and colleagues — Resilience of urban metro rail networks globally: topology, disruption and recovery across 45 metro networks.
• Railway track-geometry degradation review: prediction, uncertainty, interpretability and maintenance value.
• Data-driven prognostics review: remaining useful life, robustness, uncertainty and interpretability.
• Montenegro and colleagues — Wheel–rail contact model for railway vehicle–structure interaction: contact modelling and vehicle dynamics.

Teaching and discussion guide

1. Draw a five-station line and calculate a simple timetable from running and dwell times.
2. Compare capacity at 180-second, 120-second and 100-second headways.
3. Create two headway sequences with the same average but different variance. Explain the passenger effect.
4. Model a platform queue when arrivals exceed boardable capacity for three trains.
5. Compare a direct route with a faster route requiring one transfer using generalised cost.
6. Explain why a 20 per cent speed reduction produces a larger reduction in v²/R lateral acceleration.
7. Draw two train power curves and identify where regenerative braking could overlap with acceleration.
8. Design a fictional anomaly detector and show how its threshold changes false positives and false negatives.
9. Map one track measurement by position and time, then explain why repeated passes are more valuable than one pass.
10. Remove one interchange from a small graph and measure how shortest paths change.
11. Distinguish physical repair time from passenger recovery time after a disruption.
12. Choose one proposed railway improvement and trace its effects through the MRT Mixer table.
13. Ask which passenger is hidden by an average journey-time measure.
14. Finish with the hardest question: what evidence would make your model admit it was wrong?

The shortest useful summary

• An MRT is a closed-loop mass-mobility system, not merely a train on rails.
• Passenger demand becomes an origin–destination, direction and time problem.
• A timetable is a network of events; control turns those events into safe trajectories.
• Headway connects train separation, frequency, capacity, waiting and delay propagation.
• Dwell converts railway movement into passenger access and can limit the entire line.
• Capacity is space delivered through time with sufficient regularity, not just people per train.
• Routes are weighted graphs whose costs include waiting, walking, transfers, crowding, reliability and accessibility.
• Energy depends on speed profiles, coasting, regenerative braking and the timing of multiple trains.
• Wheel–rail contact transmits load, traction, braking and guidance through tiny elastic patches.
• Inspection and predictive maintenance turn sensor evidence into justified engineering action.
• Resilience depends on alternatives that possess usable capacity and on recovery that reaches passengers.
• World Return compares the model with real trains, assets and journeys so the railway can learn.

The real operating condition

An MRT can have excellent trains and still fail if stations cannot process passengers.

It can have advanced signalling and still underperform if dwell becomes unstable.

It can have a fast timetable and still deliver slow journeys if transfers and waiting are ignored.

It can have alternative routes on the map and still lack resilience if those routes have no spare capacity.

It can have abundant sensor data and still miss degradation if the data are noisy, inconsistent or disconnected from maintenance action.

It can restore a component and still leave passengers in recovery.

The whole railway works only when the pieces return to one another.

Passenger need
→ demand
→ route
→ timetable
→ safe separation
→ motion and braking
→ station exchange
→ capacity
→ energy
→ inspection and maintenance
→ resilience
→ passenger return
→ learning

The passenger sees a train.

The timetable planner sees events.

The control engineer sees states and constraints.

The electrical engineer sees power flowing in both directions.

The station planner sees queues and doorway rates.

The operations researcher sees a network optimisation.

The maintenance engineer sees degradation and remaining life.

The passenger sees whether the next train came and whether the journey remained possible.

An MRT works when mathematics keeps thousands of changing states inside a safe, useful and recoverable operating envelope while people move through the city.

Keep the trains separated. Keep the passengers moving. Keep the assets observable. Keep the recovery path open. Let reality correct the timetable.