How MRT Depots and Fleet Operations Work Using Mathematics

The first MRT train of the morning is not the beginning of the railway day. It is the result of the railway night.

Before passengers arrive, trainsets have to be inspected, maintained, cleaned, assigned, positioned, released and launched in the correct sequence. The operating fleet has to be manufactured from a larger physical fleet every single day.

At 7.30am, an MRT platform can make the railway look simple.

Train arrives.

Doors open.

Passengers board.

Two minutes later, another train arrives.

But those trains did not simply exist in the correct places at the correct time.

The night before, every physical trainset belonged to one of several competing states.

Ready for service.
Awaiting inspection.
Under scheduled maintenance.
Under fault investigation.
Being cleaned or washed.
Waiting for parts.
Stabled in a location that may or may not fit tomorrow’s launch sequence.
Needed as a spare rather than planned for normal service.
Accumulating mileage that must remain balanced against the rest of the fleet.

A depot therefore does not merely store trains.

A depot converts a collection of individual trainsets into a timed, maintainable and recoverable operating fleet.

This job becomes particularly visible in Singapore’s East Coast Integrated Depot. LTA describes the depot complex as the world’s first four-in-one depot, combining three independently operated train depots for the East-West, Thomson-East Coast and Downtown Lines with a bus depot. The train depots have a combined stabling capacity of around 220 trains while saving about 44 hectares of land compared with separate facilities.

From 9 December 2025, East-West Line trains could be launched and withdrawn from ECID. LTA explicitly said the new depot would increase train launch and withdrawal capacity, giving greater flexibility to adjust service patterns and respond more quickly to incidents.

That sentence reveals the mathematics.

depot capacity
→ launch capacity
→ fleet availability
→ timetable feasibility
→ incident recovery
→ passenger service

The depot is therefore part of the live railway even when passengers never enter it.

This article continues the eduKateSG MRT mathematics cloud. The permanent synthesis page is How MRT Works | It’s Mathematics. Closely connected pillars include How an MRT Timetable Is Built Using Mathematics, How MRT Predictive Maintenance Works Using Mathematics, How MRT Signalling and Train Regulation Work Using Mathematics and How MRT Network Resilience Is Measured Using Mathematics.

The RFE — Why Does an MRT Depot Exist?

The weakest answer is:

to park trains at night

Parking is real.

But a depot that merely stores trains and cannot inspect, maintain, clean, launch, withdraw or recover them is not doing the job the railway needs.

The Reason for Existence is:

turn the physical train fleet
into tomorrow’s reliable operating fleet
by allocating space, maintenance, people, parts and time
so enough safe trainsets are available
at the correct launch points
in the correct sequence
with enough spare capacity for failure and recovery

The depot must satisfy several objectives simultaneously.

Maximise service-ready trains.
Complete required maintenance before deadlines.
Balance mileage and wear across the fleet.
Minimise unnecessary shunting and dead running.
Fit trains into finite stabling and maintenance space.
Keep enough reserve for faults and disruption.
Use limited technicians, equipment and spare parts efficiently.
Release trains in the right order for the morning timetable.

The optimum is not:

put every available train into service

Doing that could remove reserve capacity and leave no room for maintenance.

The RFE is fleet continuity: enough trains working now, enough trains being repaired for later, and enough margin that one bad train does not become a bad morning for the whole line.

Prompt 1 — How Many Trains Does the Railway Need Tomorrow?

The timetable creates the first fleet requirement.

If a complete round-trip cycle takes T_cycle seconds and service headway is H seconds, a simple minimum number of simultaneously circulating trains is:

Nservice = ceil(Tcycle/H)

Suppose fictional cycle time is 6,600 seconds and target peak headway is 120 seconds.

Nservice = ceil(6600/120)
         = ceil(55)
         = 55 trains

That is not yet the fleet requirement.

The operator also needs trains for:

planned maintenance,
corrective maintenance,
inspection,
reserve or standby,
testing and commissioning where relevant.

A simplified fleet balance is:

Nfleet
= Nservice
+ Nreserve
+ Nmaintenance
+ Nother

Suppose:

Nservice = 55
Nreserve = 4
Nmaintenance = 8
Nother = 1

Then:

Nfleet = 68

The ratio:

rreserve = Nreserve/Nservice

is one simple measure of operational reserve.

But there is no universal correct reserve ratio independent of fleet age, maintenance regime, line design, reliability and recovery strategy.

Availability changes the answer

Suppose each train has probability A of being available for service on a particular day.

If availability were independent across N trainsets, number available K could be approximated:

K ~ Binomial(N,A)

Expected available trains:

E[K] = NA

If N=68 and A=0.95:

E[K] = 64.6

That expected value does not guarantee 55 service trains every day.

The railway cares about:

P(K ≥ Nservice + required reserve)

This converts maintenance reliability into service-risk probability.

The independence assumption is also fragile. A common fault can affect several similar trainsets at once. Supply-chain problems can make many repairs wait for the same part. Fleet availability therefore has correlated failure risk.

The timetable asks for a number of trains. The depot asks whether that number still exists after maintenance, faults, common dependencies and reserve are counted honestly.

Prompt 2 — How Does One Physical Train Become a Sequence of Train Services?

Passengers see train services.

The depot sees physical trainsets.

Those two identities must be connected.

Suppose the timetable contains service trips:

Trip 1: terminal A → terminal B
Trip 2: terminal B → terminal A
Trip 3: terminal A → terminal B
...

A physical trainset can operate a chain of compatible trips.

Trainset 041
→ Trip 1
→ turnback
→ Trip 2
→ layover
→ Trip 3
→ withdraw to depot

This is rolling-stock circulation.

Compatibility edges

Represent each scheduled trip as a node.

Create an edge from trip i to trip j if one trainset can physically finish i and then operate j.

edge i→j exists if
end location of i = start location of j
and
finish time i + minimum turnaround ≤ start time j

A full operating day becomes a path-cover problem.

The objective is to cover every required train trip with feasible physical trainset paths while respecting maintenance, depot and vehicle constraints.

Define binary variable:

xij = 1 if the same trainset operates trip j after trip i
      0 otherwise

Then each trip needs one predecessor and one successor, except depot launch and withdrawal boundaries.

This is naturally expressed through integer programming.

Circulation and timetable should talk to each other

A timetable can be passenger-efficient but fleet-inefficient.

For example, a pattern of departures can create awkward terminal gaps requiring extra trainsets even though passenger waiting barely changes.

Research therefore often jointly optimises timetable and rolling-stock circulation rather than solving one first and treating the other as an afterthought.

Recent 2026 research goes further, modelling train schedules and rolling-stock circulation under multiple depots, shared depots and dynamic demand.

The whole-system objective can look like:

minimise
passenger waiting
+ number of trainsets used
+ empty movement
+ depot shunting
+ maintenance conflicts

subject to
headway
capacity
turnback
fleet availability
depot capacity
maintenance requirements

The Timetable pillar owns the event-network side.

A timetable schedules journeys. Rolling-stock circulation proves that enough real trains can physically perform them.

Prompt 3 — How Does a Depot Fit Hundreds of Trains into Finite Space?

A depot is a spatial optimisation problem.

It contains different functional spaces:

stabling tracks,
inspection roads,
maintenance bays,
washing facilities,
workshops,
test areas,
launch and reception tracks,
internal movement paths.

One train cannot occupy two places at once.

One track cannot host incompatible movements at the same time.

One maintenance bay can process only a finite number of trainsets.

Stabling capacity

Let depot have S stabling positions.

At night, if N trains need stabling:

N ≤ S

sounds sufficient.

But physical arrangement matters.

Suppose several trains share one track in sequence.

The first train required in the morning should ideally not be trapped behind three later trains.

This creates a stack-like sequencing problem.

night stabling order
must be compatible with
morning extraction order

Otherwise unnecessary shunting is required.

Shunting cost

Let y_i be number of internal depot moves for train i.

A simple objective is:

minimise Σ yi

But movement count alone is incomplete.

A better cost can include distance and conflict:

Cshunt
= Σi [α moves_i + β distance_i + γ conflict_risk_i]

Again this is a conceptual public model, not an operating procedure.

Why ECID is mathematically interesting

ECID places three MRT depots on different vertical levels within one integrated complex, yet LTA says the three train depots are designed to operate independently.

That is a powerful land-use optimisation:

share scarce land volume
without forcing operational dependency among lines

The design saves about 44 hectares while maintaining separate train-depot functions.

This is civilisation mathematics.

Land is finite.

Rail capacity needs more fleet.

More fleet needs more depot capacity.

The solution is not simply “find more land”.

The depot solves a packing problem in both space and time: where can every train fit tonight, and can the right one still escape first tomorrow morning?

Prompt 4 — How Does the Morning Launch Sequence Work?

Before peak service begins, trains must enter the line in a controlled sequence.

For train i, define:

li = planned depot launch time

The train must be ready before that time.

maintenance complete_i ≤ li
cleaning complete_i ≤ li
inspection complete_i ≤ li

The train must also reach its first passenger-service event in time.

li + deadhead_i + setup_i ≤ first_service_i

where deadhead is non-passenger movement from depot to service position.

Several trains may need to use the same depot throat, reception track or access path.

This creates sequence constraints:

launch B after launch A + required movement separation

The actual safety rules and depot movement parameters are system-specific and not reproduced here.

Launch capacity

If one launch path can process μ trains per hour and peak build-up requires λ trains per hour:

λ < μ

must hold with sufficient margin.

If demand approaches capacity:

ρ = λ/μ → 1

small delays can create rapidly growing queues of trains waiting to launch.

This is why LTA’s statement that ECID increases launch and withdrawal capacity matters operationally. Depot access capacity can directly affect how flexibly the fleet can enter and leave service.

One failed launch can move into the timetable

Suppose Train 12 is scheduled to launch at 05:12 but is not ready until 05:20.

The depot can:

substitute another service-ready train if a suitable reserve exists,
alter the launch sequence under authorised operations,
or accept a service gap if no replacement can be made safely.

Again, specific operating procedures are not reproduced.

The mathematical consequence is:

depot readiness delay
→ launch delay
→ first-service delay
→ headway gap
→ passenger accumulation

The first headway error of the morning can be created before the train ever reaches a passenger platform.

Prompt 5 — How Does Maintenance Fit into a Fleet That Is Needed Every Day?

Maintenance creates a fundamental contradiction.

The railway needs trains in service.

Maintenance needs trains out of service.

Both are necessary.

A train that never enters maintenance can become unreliable.

A train that spends too much time in maintenance reduces fleet availability.

The depot therefore solves a scheduling problem.

For train i and maintenance task k:

start Sik
completion Cik = Sik + duration dik

Maintenance must respect deadlines based on time, mileage, condition or authorised maintenance rules.

Cik ≤ deadlineik

Resources are finite.

number of trains using bay b at time t
≤ capacityb

Technicians are finite.

Σ labour_required(task at t)
≤ workforce_available(t)

Parts are finite.

parts consumed
≤ inventory on hand + arrivals

This turns maintenance into a resource-constrained project-scheduling problem.

Maintenance queue

If faults and scheduled work arrive faster than maintenance capacity can complete them, a backlog forms.

Let λ_m be incoming maintenance workload and μ_m be completion capacity.

if λm > μm for sustained periods
→ backlog grows

Backlog then reduces available fleet, which can reduce spare ratio and make future faults more disruptive.

maintenance backlog
→ fewer service-ready trains
→ lower reserve
→ less ability to absorb faults
→ more reactive work
→ larger backlog

This is a dangerous positive feedback.

Engineering hours are scarce

Not all rail maintenance occurs in depots. Track, power and signalling teams also require the railway when passenger service stops.

SMRT’s current public maintenance schedule repeatedly shows overnight work windows around 1.00am to 4.30am on many sections. The exact schedule varies, but it illustrates the narrow nightly production window available to the railway.

The 2026 Rail Reliability Taskforce has therefore recommended setting aside more engineering hours, including longer service closures when necessary for major renewal.

Time is maintenance capacity.

The railway loses availability if it maintains too little—and also loses availability if maintenance itself consumes too much fleet or time. Depot mathematics lives between those two failures.

Prompt 6 — Why Must the Fleet Balance Mileage and Wear?

If one train runs every peak while another remains stabled, their physical ages diverge even if they were delivered on the same day.

Train i accumulates mileage m_i.

Fleet mean mileage is:

m̄ = (1/N)Σ mi

Mileage variance is:

σm² = (1/N)Σ(mi−m̄)²

A fleet planner may try to limit this variance so wear and maintenance milestones remain more evenly distributed.

But perfect mileage equality is not always optimal.

Some trainsets may need specific routes, tests or maintenance.

Some may have recent component replacements.

Some may be deliberately held as reserve.

So the objective can be:

minimise
mileage imbalance
+ maintenance bunching
+ empty movement
+ unnecessary shunting

subject to
service and maintenance constraints

Why maintenance bunching matters

If many trainsets reach the same overhaul threshold at once, maintenance demand spikes.

Suppose trains require a major task every M kilometres.

If ten trains have nearly identical mileage, all ten may become due within one short period.

Balanced use should therefore consider not only equal mileage but the future distribution of maintenance deadlines.

today's assignment
→ tomorrow's mileage
→ future maintenance due date
→ future depot workload

The fleet scheduler is creating tomorrow’s maintenance queue every time it chooses today’s train assignment.

A 2026 study on integrated weekly urban-rail vehicle scheduling specifically optimises task assignment, maintenance, washing and parking together, including mileage balance and shunting. That research captures the same principle: depot tasks should not be optimised independently because one decision changes the feasibility of the others.

Mileage is not merely distance travelled. It is future maintenance workload being accumulated one service trip at a time.

Prompt 7 — How Do Spare Parts and People Become Fleet Availability?

A train can be mechanically repairable and still unavailable if the required part is not present.

It can have the part and still be unavailable if the correct technician, test equipment or maintenance bay is unavailable.

Fleet availability therefore depends on a supply network.

fault diagnosed
→ part identified
→ part available
→ technician available
→ maintenance bay available
→ repair completed
→ test completed
→ train released

Remove any one of those and the train may remain out of service.

Spare-parts inventory

Suppose part demand during lead time L has mean μ_L and standard deviation σ_L.

A classic reorder point is:

ROP = μL + zσL

The second term is safety stock.

More safety stock reduces stockout probability but consumes money and storage.

Too little stock:

fault
→ no part
→ long train downtime

Too much stock:

capital tied up
+ obsolete inventory
+ storage burden

The 2026 Rail Reliability Taskforce specifically addresses this problem. LTA and the operators will strengthen spares forecasting through better data collection and analysis, stockpile additional parts where needed, procure extended OEM support, localise production and repair capabilities and diversify sources of supply.

This is supply-chain mathematics becoming train availability.

Workforce capacity

Suppose maintenance task k requires h_k technician-hours.

Daily depot labour capacity is:

Hcapacity = Σworkers hours_available

But skills are typed.

An electrical specialist is not automatically interchangeable with every mechanical or signalling task.

For skill s:

Σ task_hours_requiring_s
≤ skilled_hours_available_s

This is why workforce capability is part of asset reliability.

The same 2026 Taskforce recommends expanded training, digital skills, certification and structured staff development across rail operations and maintenance.

A spare train is not created by owning extra rolling stock. It is created by having the train, the part, the person, the bay, the time and the authority all available at once.

Prompt 8 — How Does a Depot Help the Railway Recover from Disruption?

A depot is a recovery resource.

During a disruption, the railway may need to:

withdraw a faulty train,
launch a replacement where operationally possible,
resequence train circulation,
inspect a train after an incident,
prepare recovery equipment,
repair damaged or failed rolling stock.

Specific rescue and depot procedures are not reproduced here.

The public systems geometry is:

faulty train
→ service capacity lost
→ withdraw or move train safely
→ reserve fleet and depot access become valuable
→ replacement or revised service pattern
→ repair
→ return train to fleet

LTA’s ECID statement is directly relevant: greater launch and withdrawal capacity creates greater flexibility to adjust service patterns and respond more quickly to incidents.

Recovery time

Let:

Tdetect = time to identify train problem
Tsecure = time to make movement state safe
Twithdraw = time until faulty train no longer blocks service
Treplace = time until replacement capacity is available
Trepair = time until faulty train becomes service-ready again

Passenger recovery may depend mostly on:

Tservice-recovery
≈ Tdetect + Tsecure + Twithdraw + Tresequence

while train asset recovery continues until T_repair.

The two clocks differ.

A railway can recover passenger service before the failed train itself is repaired.

This is the same distinction developed in the Network Resilience pillar:

asset repair
≠
passenger recovery

Reserve allocation

Suppose only three service-ready reserve trains exist.

Using one reserve now reduces reserve for the next incident.

Reserve allocation therefore has option value.

value of reserve
= immediate service benefit
+ preserved ability to respond to future uncertainty

The depot should not merely ask:

Can I launch a spare?

It should ask:

Is this the best justified use of the remaining reserve under current and future risk?

A depot is part warehouse, part hospital, part parking system and part emergency reserve. Its mathematics decides which role must dominate at each moment.

A Complete Fictional Night-to-Morning Fleet Example

Consider a fictional metro line with 70 trainsets.

All numbers are invented and do not represent a Singapore line’s fleet, depot allocation or maintenance rules.

Step 1 — Tomorrow’s peak requirement

Required in passenger service = 56
Required operational reserve = 4
Target ready fleet at morning launch = 60

Step 2 — Fleet state at midnight

Service-ready: 52
Scheduled maintenance: 8
Corrective repair: 4
Inspection/cleaning pending: 6
Total: 70

The depot needs eight additional trains to become ready before launch.

Step 3 — Maintenance resources

Suppose available overnight resources are:

4 maintenance bays
3 inspection roads
1 wash plant
80 technician-hours

Four scheduled maintenance jobs each require 10 technician-hours.

Two corrective repairs require 12 hours each.

Six inspections require 2 hours each.

Total required labour if all are attempted:

4×10 + 2×12 + 6×2
= 40 + 24 + 12
= 76 technician-hours

Labour fits inside 80 hours.

But bay capacity and task duration still determine whether the jobs fit in time.

Step 4 — One part is missing

One corrective repair cannot finish because a required component has not arrived.

Expected ready additions become:

4 scheduled-maintenance completions
+ 1 corrective repair
+ 6 inspection/cleaning completions
= 11 potential additions

But some of those trains may overlap the same initial ready count categories. Suppose net ready fleet reaches 62.

Morning requirement was 60.

ready margin = 62−60 = 2 trains

The railway is viable, but thinner than planned.

Step 5 — Stabling order problem

Train 031 must launch at 04:52.

But it is stabled behind Train 044, which is scheduled to launch at 05:20.

Without correction, an extra shunting move is required.

The night scheduler swaps the positions earlier while the depot movement network is clear.

One planned move at 02:30 prevents a launch conflict at 04:50.

Step 6 — Launch sequence

04:50 Train 031
04:54 Train 017
04:58 Train 065
05:02 Train 008
...

Each train has a first-service deadline downstream.

Depot launch, dead running and timetable arrival must all fit.

Step 7 — Morning fault

At 05:05, one planned train fails a final readiness check.

One of the two spare ready trains is substituted under the fictional scenario.

Ready margin drops:

remaining ready reserve = 1

The timetable is preserved, but resilience has fallen.

Step 8 — World Return

After morning peak, depot planners compare:

predicted ready fleet = 62
actual ready fleet at launch = 61

predicted launch delay = 0
actual maximum launch deviation = 3 min

predicted overnight maintenance completions = 11
actual = 10

The missing part caused one repair miss.

The launch deviation came from one unexpectedly long inspection.

Those are different failure families.

The next night’s schedule and inventory forecast should change accordingly.

The Depot and Fleet Deletion Tests

Remove the timetable

The depot knows trains are healthy but not which ones need to be where or when.

Remove rolling-stock circulation

Scheduled trips are assumed to have a physical train automatically available for each one.

Remove stabling order

Every train fits somewhere but morning extraction becomes a shunting mess.

Remove maintenance deadlines

The scheduler maximises service availability by indefinitely postponing required work.

Remove fleet reserve

One train fault immediately reduces planned passenger service.

Remove spare parts

A repairable train becomes unavailable because the supply chain is incomplete.

Remove skilled labour

Parts, bays and time exist but no authorised capability can complete the task.

Remove mileage balance

A subset of trains accumulates disproportionate wear and future maintenance demand bunches.

Remove launch capacity

Enough trains are ready but the depot cannot put them onto the railway quickly enough.

Remove World Return

The depot continues planning around predicted completion times after real maintenance and launch performance have changed.

The Depot Paradoxes

Paradox 1 — A train not carrying passengers can protect passenger capacity

A reserve train has value precisely because it is not already committed.

Paradox 2 — Taking trains out of service can improve future service

Maintenance consumes present availability to preserve future reliability.

Paradox 3 — More trains can make a depot less useful if launch and maintenance capacity do not grow

Fleet size without supporting depot throughput creates congestion rather than service.

Paradox 4 — Perfect mileage equality can be inefficient

Forcing identical mileage can create unnecessary empty movements, maintenance conflicts or reserve loss.

Paradox 5 — A spare part rarely used can be operationally priceless

Inventory value is determined partly by avoided downtime, not usage frequency.

Paradox 6 — A larger depot can save land

ECID integrates multiple depots vertically within one site, increasing total capacity while reducing total land footprint versus separate facilities.

Paradox 7 — The morning timetable can fail because of a decision made at midnight

Stabling, unfinished maintenance and train readiness create tomorrow’s first headways before passenger service starts.

Paradox 8 — Depot efficiency and railway resilience can conflict

Using every train and bay at maximum utilisation can look efficient while leaving no margin for a fault.

The Depot and Fleet Audit

How many trains does tomorrow’s timetable require?
How many service-ready reserve trains are required?
How many trainsets are physically in the fleet?
Which trains are unavailable and why?
Which maintenance deadlines are time-based, mileage-based or condition-based?
Which trains must be released before first service?
What maintenance bays, roads and equipment are required?
What skilled labour is required for each task?
Are the required spare parts on hand?
What is the expected completion time of each task?
What uncertainty surrounds those completion times?
Where will every train be stabled?
Does stabling order match morning launch order?
How many shunting moves are required?
What mileage will each assignment add?
Will today’s assignment bunch future maintenance deadlines?
What launch capacity is available?
Which depot access path becomes the bottleneck?
Can the first-service timetable still be reached after dead running?
What happens if one planned train fails readiness inspection?
What reserve remains after substitution?
What common-cause fault could affect several trainsets?
What supply-chain dependency can create long downtime?
How fast can a faulty train be withdrawn and the service resequenced?
What observed depot performance would force tomorrow night’s plan to change?

How the Mathematics Grows from School to Research

Primary Mathematics

counting trains,
time and schedules,
capacity,
simple averages.

Secondary Mathematics

ratios and percentages,
probability,
variance,
graphs and networks,
linear inequalities.

Junior College

probability distributions,
optimisation,
queueing,
statistics,
discrete decision variables.

University and Research

integer and mixed-integer programming,
graph path covering,
rolling-stock circulation,
vehicle-scheduling optimisation,
maintenance planning,
queueing theory,
inventory optimisation,
reliability engineering,
stochastic programming,
reinforcement learning and real-time rescheduling.

A student learns that 60 trains minus 5 broken trains leaves 55.

A railway fleet planner asks which 55, after what maintenance, parked in what order, released through which path, with which mileage, using which parts and workers, while still preserving enough reserve for the fault that has not happened yet.

The World Return — When the Morning Railway Answers the Night Plan

The depot creates predictions.

Predicted maintenance completion
Predicted ready fleet
Predicted spare ratio
Predicted launch sequence
Predicted mileage distribution
Predicted parts consumption
Predicted technician workload

The morning returns reality.

eready = Nready,actual − Nready,predicted

elaunch,i = tlaunch,actual − tlaunch,predicted

emaint,k = Cactual,k − Cpredicted,k

If maintenance repeatedly takes longer than predicted, task-duration estimates need correction.

If one component repeatedly drives stockouts, inventory forecasting needs correction.

If one stabling track repeatedly creates extra shunting, the parking algorithm needs correction.

If one trainset repeatedly returns with more defects, condition monitoring or assignment strategy may need review.

If the fleet launched on time but passengers still experienced poor headway, the cause belongs downstream in operations rather than the depot.

That distinction matters.

night plan
→ maintenance
→ cleaning and inspection
→ stabling
→ morning launch
→ actual train service
→ faults and mileage return
→ depot data
→ next night plan

The railway learns daily.

The depot’s final examination happens after sunrise: did the trains it promised actually become the railway passengers received?

RFE Return — What Does Good Fleet Operations Owe the Passenger?

The passenger should not care which trainset number arrived.

They should not need to know which maintenance bay it used last night.

They should not need to know which spare component allowed it to return to service.

They should experience:

enough trains
+
clean and service-ready trains
+
regular morning launch
+
reliable peak service
+
reserve for faults
+
maintenance that prevents tomorrow’s disruption

The depot therefore owes the passenger something they almost never see:

a functioning tomorrow.

The RFE of the depot is to make tomorrow morning’s fleet exist before tomorrow morning needs it.

Conclusion — Every Morning’s MRT Is Manufactured the Night Before

At the end of passenger service, trains return.

Some are healthy.

Some carry fault reports.

Some are due for inspection.

Some need cleaning.

Some have reached mileage thresholds.

The timetable for tomorrow is already waiting.

The depot must now solve the inverse problem.

given tomorrow’s timetable
which physical trains
must be maintained
where must they be parked
what must be repaired
which parts and workers are needed
what sequence must they launch in
and how much reserve must remain?

At ECID, that logic is now attached to one of the world’s most unusual depot structures: three independently operating MRT depots stacked and integrated with a bus depot, with around 220-train stabling capacity and increased launch/withdraw flexibility.

The fleet planner sees vehicle paths.

The depot controller sees track occupancy and launch sequence.

The maintenance engineer sees work orders and deadlines.

The inventory planner sees lead times and safety stock.

The workforce planner sees skill-hours.

The reliability engineer sees reserve and common-cause risk.

The passenger sees a train arriving at 7.30am.

An MRT depot works when finite trains, finite space, finite maintenance, finite people and finite time are rearranged overnight into the illusion that the morning railway was always ready.

Key Equations

Nservice = ceil(Tcycle/H)
Service train requirement

Nfleet = Nservice + Nreserve + Nmaintenance + Nother
Simplified fleet balance

rreserve = Nreserve/Nservice
Reserve ratio

K ~ Binomial(N,A)
Simplified independent fleet-availability model

E[K] = NA
Expected available trainsets

xij ∈ {0,1}
Rolling-stock trip-connection decision

Cshunt = Σ[αmoves + βdistance + γconflict]
Conceptual depot-movement cost

li + deadhead_i + setup_i ≤ first_service_i
Morning launch feasibility

ρlaunch = λlaunch/μlaunch
Depot launch utilisation

Cik = Sik + dik
Maintenance completion time

Cik ≤ deadlineik
Maintenance deadline condition

λm > μm → maintenance backlog grows
Maintenance queue condition

m̄ = (1/N)Σmi
Fleet mean mileage

σm² = (1/N)Σ(mi−m̄)²
Mileage variance

ROP = μL + zσL
Classic spare-parts reorder point

Σ task_hours(skill s) ≤ skilled_hours_available(s)
Skill-capacity constraint

A ≈ MTBF/(MTBF+MTTR)
Repairable asset availability

Tservice-recovery ≈ Tdetect+Tsecure+Twithdraw+Tresequence
Conceptual service-recovery time

er eady = Nactual−Npredicted
elaunch = tactual−tpredicted
emaint = Cactual−Cpredicted
World Return residuals

Reader-safety note: All fictional fleet sizes, reserve ratios, launch sequences, depot capacities, maintenance times and operating examples in this article are educational abstractions unless explicitly attributed to a public source. This article does not reproduce Singapore MRT depot layouts, security-sensitive access paths, train launch procedures, rescue procedures, internal spare ratios, maintenance thresholds, staffing rosters or restricted operating rules.

Continue the MRT Mathematics Cloud

How MRT Works | It’s Mathematics — the permanent whole-system hub.
How MRT Timing Works Using Mathematics — system timing.
How MRT Braking Works Using Mathematics — stopping trajectories.
How MRT Headway Works Using Mathematics — spacing and throughput.
How MRT Station Dwell Time Works Using Mathematics — passenger exchange.
How MRT Delays Propagate and Recover Using Mathematics — delay and recovery.
How an MRT Timetable Is Built Using Mathematics — planned service.
How MRT Energy Use Is Optimised Using Mathematics — traction energy.
How MRT Passenger Capacity Is Calculated Using Mathematics — train and line capacity.
How MRT Routes and Transfers Are Optimised Using Mathematics — network routing.
How MRT Network Resilience Is Measured Using Mathematics — resilience.
How MRT Predictive Maintenance Works Using Mathematics — maintenance prediction.
How MRT Track Inspection Works Using Mathematics — track health.
How MRT Wheel–Rail Contact Works Using Mathematics — wheel–rail physics.
How MRT Noise and Vibration Work Using Mathematics — acoustics and vibration.
How MRT Power Supply Works Using Mathematics — traction power.
How MRT Signalling and Train Regulation Work Using Mathematics — CBTC and regulation.
How MRT Platform Screen Doors Work Using Mathematics — precision station interface.
How MRT Tunnel Ventilation and Airflow Work Using Mathematics — piston airflow and ventilation.

The next natural pillar is How MRT Tunnels Are Built Using Mathematics: alignment geometry, horizontal and vertical curves, surveying, tunnel-boring-machine guidance, settlement, structural loads, construction tolerances and how two underground tunnels started kilometres apart still have to meet the railway map drawn above them.

How MRT Depots and Fleet Operations Work Using Mathematics: How Tomorrow Morning’s Railway Is Built Overnight