Summary: A stockyard is the one site where the largest, the fastest changing and the most asymmetric load on the foundation of a machine is the material the machine exists to store. A coal pile reaches 210 kilopascals and an ore pile more than 340, which is 1.75 to 2.83 times the 120 kilopascal dead plus live allowance for the stacker reclaimer berm itself. The ground under an improved yard may still be allowed 150 millimetres of settlement and 0.3 percent of tilt, while the rail on it is held to cross level differences of 5 to 10 millimetres and a warp of 10 millimetres in 10 metres, and the gap between the two tables is spent on maintenance: one published design allowed three re-levelling episodes in a 50 year life. A tilt channel on the rail reads the differential family only, and its record is a sawtooth whose rate exists only between the resets.

1. The Load That Moves the Track Is the Material Being Stored

The design loads of a rail mounted stacker or reclaimer are usually read from the machine side. ISO 5049-1, the design standard for mobile continuous handling equipment for bulk materials, confirmed current as recently as 2026, gives the machine its own load groups: a handling wind of 20 metres per second, a filling factor of 1.5 on the bucket wheel and a dynamic factor of 1.1 on conveying loads. An operator manual for a bucket wheel machine on an 8 metre gauge quotes a maximum wheel load of 250 kilonewtons and a working wind below 25 metres per second. A 750 tonne machine at 250 kilonewtons per wheel carries its weight on about thirty wheels, and the design allowance for dead plus live machine loads on the berm of a published Australian coal stockyard was 120 kilopascals.

None of those numbers is the governing load on the rail foundation. The governing load is the pile. The same Australian project, a coal export terminal stockyard treated by dynamic replacement, designed its coal pads for 21 metre piles at 210 kilopascals. A Brazilian steel plant stockyard on very soft clay, published through the ISSMGE, carried piles to 13 metres with a surcharge changing quickly from zero to more than 100 kilopascals for coal and coke and more than 340 for ore. The implied bulk unit weights are 10.0 kilonewtons per cubic metre for the coal and 26.2 for the ore, and the pile exceeds the machine berm allowance by 1.75 times for coal and 2.83 times for ore. On the ground under the rails, the machine is the light party.

Two features separate this surcharge from every other load in this series. It has no fixed level: a yard fills and empties, so the load cycles from nothing to full within days, and the Brazilian design had to run its stability and serviceability checks for different shapes and positions of the piles because the load moves with the yard plan. And it is an area load: the machine loads two narrow strips through about thirty wheels, while the pile loads the whole yard at once, which is why the settlements of untreated yards are published in metres, 700 millimetres at one Australian site, more than 3 metres under 10 metres of iron ore at Beira in Mozambique, and up to 4 metres calculated for the Brazilian plant.

There is also a structural gap between the two design documents that the monitoring plan has to fill. ISO 5049-1 governs the steel of the machine, its load groups, its stability against overturning and its fatigue checks, and it says nothing about the ground under the rails, because the foundation is a geotechnical deliverable. The rail support industry makes the same point from the other end: stackers and reclaimers run on rails supported by concrete and impart dynamic loads through the wheels, the wheel loads are not evenly distributed along the machine so the two rails are often different sections, and uncontrolled support forces produce misalignment, rail breaks, soleplate bending and premature grout failure. The machine standard stops at the wheel, the ground criteria stop at the berm, and the rail in between belongs to whoever maintains it, which is why every published remediation of this family is a rail project rather than a machine project.

2. Two Tables, One Track: The Machine Tolerance and the Ground Criterion

The machine side publishes installation tolerances, the ground side publishes performance criteria, and the two never meet. GB 51062, the Chinese construction norm for coal mine equipment installation, gives the arm type bucket wheel stacker reclaimer its rail table: gauge within plus or minus 4 millimetres up to 3.2 metres and 6 above, rail centre line position within 6 millimetres, straightness at 1 in 1000 with full length totals of 15 and 25 millimetres, and the cross level difference at any one section within 1/1000 of the gauge and never more than 5 millimetres. An industry installation table for a 7 metre gauge gives the span at plus or minus 5 millimetres, cross level at gauge over 1000, warp at 10 millimetres in 10 metres, total gradient at 1 in 1000, joint steps at 0.5 millimetres and sleeper tilt at 1 in 50. An operator manual for an 8 metre gauge allows plus or minus 10 millimetres on both gauge and cross level.

The ground side is a different currency. The improved ground criteria published for the Australian stockyard allow the stacker reclaimer berm 150 millimetres of vertical settlement, 25 millimetres of horizontal displacement, a vertical differential of 1 in 500 over a 25 metre chord, a horizontal differential of 1 in 1500 over an 8 metre chord, and a tilt of 0.3 percent. The coal pad, at 210 kilopascals, is allowed 200 millimetres. These are criteria for ground that was improved; on the untreated references in the first section the same quantities are measured in metres.

Converted into one unit the two tables sit three times apart at the tilt. The ground criterion of 0.3 percent is 3 millimetres per metre, 0.172 deg. The rail warp of 10 millimetres in 10 metres is 1 millimetre per metre, 0.0573 deg. The ground is permitted to rotate three times more than the rail, and nothing structural absorbs the difference: the rail is a re-levelled surface riding on settling ground, and the Brazilian design chose an adjustable ballast bed for exactly that reason, recording that the runways remained very sensitive to all types of deformation. One further reading is needed before using the machine table: the total gradient of 1 in 1000 is a design grade, not an error budget, and over a 600 metre runway it is 600 millimetres of intentional fall, while the deviation rows beside it are the actual tolerances.

The ground criteria only exist because the ground was treated, and the treatment menu is what sets the numbers. The Australian terminal used dynamic replacement, compacted stone columns pounded through the soft clay, and published the criteria in the same paper as the method. The Kooragang design used a two stage preload of 9 metres of fill, about 165 kilopascals, held long enough to take the clay through its primary consolidation before the yard was built, with the monitoring array of settlement plates, inclinometers and piezometers that a staged preload requires. The Beira design used wick drains with staged construction, checking stability and settlement at each stage as the strength grew. A Malaysian iron ore distribution terminal treated a 19 metre ore stockpile with vibro stone columns and put the stacker reclaimer berms on bored piles, adopting a staged loading programme with instrumentation, because the berms could not be permitted to follow the pad. The published limits in the ground table are therefore not properties of soft ground; they are what each treatment was bought to deliver, and the treatment, not the soil, is what the monitoring is watching.

The installation side is tighter than the operating side, which is the pattern the crane runway article found when the operational tolerances doubled the construction ones. A published Chinese erection procedure for a stacker reclaimer requires the rail top elevation within plus or minus 1 millimetre and the gauge within 4, and it requires the rails to be observed for settlement every five days during erection, with any settlement re-levelled before construction continues. The in-service tolerances cited by a later remediation record on the same family of machines run to plus or minus 10 millimetres on side, top and cross difference and plus or minus 60 millimetres of straightness. Within one published table the rungs already span a factor of about twenty eight, from the 0.041 deg cross criterion to the 1 in 50 sleeper tilt at 1.15 deg, and the loosest rung governs a component one level below the rail.

3. The Gauge Is the Denominator, and the Cap Decides the Number

The cross level allowance is the one row of the machine table written as a fraction of a length, and the length is the gauge. GB 51062 prints the same-section difference as gauge over 1000, capped at 5 millimetres. The cap binds as soon as the gauge exceeds 5 metres, and the gauges of the two machines cited here are 7 and 8 metres, so within this family the fraction never governs and the effective limit is the absolute 5 millimetres. The other two documents print 7 and 10 millimetres for the same quantity, a factor of two across three sources, which at their own gauges is 0.71, 1.00 and 1.25 milliradians of machine tilt.

This is the third garment one rule has worn in this series. A tolerance written as a fraction of the object itself is a taper, as the pile verticality article showed, because the permitted deviation grows without bound in the object size. A limit written as a rate against a measurement base is a gradient whose cap deletes the length term where everyone measures, as the railway twist article showed. Here the fraction is of a distance, but the cap is set below the whole family, so the honest reading of the table is that the absolute number is the tolerance and the fraction is decoration, the same verdict the American track surface table earned for printing one allowance over two different bases.

What the cross level buys is not rail geometry but machine geometry, and the machine is tall. A portal standing on an 8 metre gauge with 10 millimetres of cross difference rotates by 1.25 milliradians, 0.0716 deg, and at a 25 metre boom tip that is 31 millimetres of lateral displacement of the bucket wheel, on a machine whose function is to place material along a line. The same arithmetic at 5 and 7 millimetres gives 18 and 25 millimetres. The cross criterion is therefore a positioning criterion wearing a levelling criterion’s clothes, and it is the row of the table that the differential settlement of the ground attacks first, because a berm tilting toward a pile lowers one rail against the other directly.

The machine table carries its own half of the same clearance, and the two budgets are spent against each other. GB 51062 gives the travel machinery of the same machines its tolerances: all wheels on one rail in one plane within 5 millimetres, wheel vertical skew within 1 in 400 of the measured length, horizontal skew within 1 in 1000, and the diagonal across the four legs within 15 millimetres, while the erection procedure for the same family asks the wheels alone for verticality within D over 1000 of the wheel diameter, leaning only inward at the lower flange, and co-position of all wheels on one side within 2 millimetres. A rail cross error and a wheel skew error point the same flange the same way, so the clearance between flange and rail is protected by two documents neither of which knows the other exists. The remediation record closes the loop in the other direction: for the span it prints the tolerance as 3 millimetres up to 10 metres and 3 plus 0.25 times the span over 10 metres above, capped at 5, which is the same graded form the crane runway tolerance standards use for span, a constant plus a quarter of the span beyond a threshold with a cap, and it is published evidence that a stockyard track is maintained with the same span practice the crane standards codify.

4. The Settlement Budget Is Spent in Re-levelling Episodes

The design documents do not treat stockyard settlement as something to be prevented. They treat it as something to be scheduled. The instrumented preload design for a stockyard at Kooragang Island, published by an Australian ground engineering consultancy, computed settlements of up to 700 millimetres without treatment, set a target of 200 millimetres beneath the reclaimer berm and 300 beneath the coal pad over a nominated 17 year design period, and recorded that considerable costs are incurred when re-levelling of stacker and reclaimer rails is required, allowing for three re-levelling episodes over a 50 year design life. The 200 millimetre target is 11.8 millimetres per year, or 32 microns per day, a rate a maintenance department can plan around and that no movement limit could hold.

The remediation record is consistent with that philosophy and shows what the episodes look like. A Western Australian stacker reclaimer rail subsided beyond safe tolerance after unusually heavy rainfall and was re-levelled in 15 days, with levels monitored daily and reported metre by metre, and with the densification of the subgrade checked by running the extremely heavy machine back and forth over the treated section while monitoring deflection. A New South Wales reclaimer rail showed general subsidence with a notable differential between the two rails nearing operational tolerance, aggravated by fire hydrant network water, sprayer operation, historic pipe failures and localised flooding, and was corrected by injection during a shutdown. A Pilbara iron ore stacker rail reached misalignment beyond the requirements of the standards, with the stated consequences of increased structural stress and fatigue on the stacker and potentially derailment, and went to foundation replacement and drainage redesign.

A published Chinese remediation of a bucket wheel stacker reclaimer track is the fullest record of one episode. The survey found lateral offsets running from 4 to 52 millimetres westward on one stretch and from 1 to 234 millimetres eastward on another. The crew lifted the track on jacks, removed one sleeper plate every 12 metres to allow lateral adjustment, replaced the rubber rail pads, re-straightened the track to a 0.6 per mille grade computed at 5 metre points off a new benchmark at 8.030 metres, ground cross differences above 20 millimetres down to 15 and shimmed with taper plates, and fitted wedge boots against derailment while the work was in progress. The whole intervention is a reset of both planes at once, done against the belt conveyor centre line so that the discharge point returns to it.

That record is the monitoring question in one image. The rail is lifted, shimmed, ground and re-graded on a schedule driven by how fast the ground moves, and every intervention resets the geometry. A tilt record on such a rail is not a trend line, it is a sawtooth: between episodes it accumulates at whatever the local rate is, and at the episode it steps back by the amount the maintenance corrected. An unsegmented trend through that record reads the maintenance as recovery and the recovery as stability, and the rate only exists between the resets. The design’s published quantity, a total settlement over a period, belongs to the ground and to the survey; the channel’s quantity is the residual drift between interventions, which is exactly what times the next one.

The economics of the episode are published often enough to see the shape. The Kooragang design records that considerable costs are incurred when re-levelling of stacker and reclaimer rails is required, which is the sentence a design document usually refuses to write, and it is the reason the design targets were set at all. The Western Australian episode bought 15 days of re-levelling and a daily level report to the metre on an operating machine, the New South Wales episode was timed into a shutdown so the conveyor and the machines kept running, and the Pilbara case went the other way once the geometry passed the standards, into foundation replacement and drainage redesign, because the episode arithmetic stopped paying. The trigger in two of the three is water rather than the design rate: rainfall, hydrant network leakage, sprayers and localised flooding, on yards whose coal is watered and sprayed by design. A stockyard is a site where the ground is wetted on purpose, and the maintenance calendar is therefore weather driven at least as much as it is settlement driven.

5. The Yard Remembers Where the Piles Have Stood

The pile is not only heavier than the machine, it moves. A yard stacks against one runway, reclaims, and restacks against the other, so the berm tilt follows the live pile toward whichever side the material currently sits. The Brazilian design had to check its limit states for different shapes and positions of the piles for exactly this reason, and it names the horizontal component, the spreading of the stockpiles outwards toward the runways, as critical to the operation of the machines. The distinction that matters for monitoring is mechanical. In a warehouse rack the goods load the structure elastically, an unloaded rack returns to plumb, and the return position is the diagnostic. A stockyard is not elastic in the large: primary consolidation is irreversible, so an emptied yard does not go back to level, and the current tilt of a berm is a sum over where the piles have stood rather than a function of where the one pile is standing now.

That has one difficult consequence and one convenient one. The difficult one is that the record cannot be interpreted against the current yard state alone: a berm that has carried ore on its north side for two years still leans north after that side is emptied. The convenient one is an inversion of the monument case. On a historic building the environment moves the structure more than the damage does, and the separation is done against temperature. On a stockyard the inventory moves the ground more than the weather does, and the load channel is free, because the operator already keeps yard tallies: pile position and pile height over time exist as a record without a single additional instrument. The dam monitoring rule, that where the driving load is measured the reversible response can be regressed out, transfers here with the yard log standing in for the reservoir level gauge.

The practical reading order for a tilt record from a berm or a rail is therefore pile position and height first, then the yard’s loading history, then weather, then the sensor. A tilt that follows the pile geometry is the yard working as designed. A tilt that keeps accumulating with the yard empty is the consolidation trend, and that is the quantity the re-levelling budget was written against. A step that no yard event explains is either an intervention that was not logged, a water event of the New South Wales kind, or a channel fault, and the first two are more common than the third.

The size of the reversible part is worth putting in millimetres before anyone specifies against it. The ground criterion of 0.3 percent is 24 millimetres across an 8 metre gauge and 15 across a 5 metre berm, so a berm tracking the live pile at the full allowance is moving its rail by tens of millimetres, which is the whole cross level budget and more. The mechanical split is that the immediate elastic component of a load step recovers when the pile moves away, the primary consolidation component never does, and the secondary creep continues at a decaying rate under a constant load, so the record divides into a part that follows the yard plan, a part that follows the loading history, and a part that follows only time. The 32 microns per day of the Kooragang target is the average of the last two over 17 years, and it is the rate, not the excursion, that an alarm on this family should carry.

6. The Horizontal Family: Spreading, Gauge and Plan Straightness

The pile pushes the runways outward as well as down. The published horizontal figures are small numbers with large consequences: 25 millimetres of horizontal displacement on the improved berm, a horizontal differential of 1 in 1500 over an 8 metre chord, which is 5.3 millimetres, and a horizontal wheel load guide of 30 kilonewtons against 250 vertical on the 7 metre gauge machine. The Chinese remediation case measured up to 234 millimetres of lateral offset, which is 9.4 times the 25 millimetre criterion and 44 times the local differential figure, on a machine held against derailment during the works by wedge boots. Gauge narrowing, plan straightness, skew and offset all live in the horizontal plane, and the failure family they feed, flange climb, gauge widening and rail rollover, is the one the vertical criteria never touch.

An angle channel is structurally blind to the whole horizontal family. A displacement in plan, and a rotation about a vertical axis, leave every tilt reading unchanged, which is the sorting rule the crane runway article stated and the conveyor article refined to the axis of rotation. On a stockyard the blind family is not a residue: the largest measured deviation in the remediation record, the 234 millimetre offset, is in exactly the plane the instrument cannot see, and the derailment modes that the cited rectification projects name as the end state are fed by it. The horizontal work therefore belongs to the plan survey, and the Chinese case shows the datum: the track was re-straightened against the belt conveyor centre line measured at the head and tail pulleys, because the stacker discharge point has to sit on that line, and a plan datum is recovered with a total station rather than inferred from tilt.

The consequences of the horizontal errors are consumables rather than events, in the sense the crane runway article gave that word. A gauge error or a plan skew runs the wheels at an angle, the flange work and the guide forces rise, and the horizontal wheel load guide of 30 kilonewtons against 250 vertical is the number that grows, so the wear bill and the energy bill arrive years before the derailment risk does. The Chinese remediation shows the same consumable on the hardware side: the anti shift concrete beside the sleepers was cut away, one sleeper plate in twelve was removed to let the track move, the fishplate and clip bolts that could not be freed were cut off and scrapped, and the rubber pads were replaced wholesale, all of which is regular expenditure on a track whose ground is still moving. The horizontal family is where the money goes even when nothing fails, and it is the family no angle channel will ever price.

7. What the Rail Channel Reads, and What It Cannot

What the channel reads on a stockyard is the differential family in the vertical plane: the gradient of each rail, the warp between stations along a run, the cross slope where a transverse track element carries a sensor across the gauge, and the tilt of the berm itself. The criterion for the along-rail family is 10 millimetres in 10 metres, 0.0573 deg, and the berm criterion is 0.172 deg, so the measured quantities and the criteria are in the same currency and nothing has to be converted beyond the small-angle constant.

What it cannot read is everything else that matters. The total settlement, which is the quantity the design budget is written on, is invisible, because a tilt sensor reports differences and a yard that settles uniformly settles flat; the uniform part is the large part, which is why the ground table is a survey criterion and not a sensor criterion. The blindness is not marginal: a berm that settles its full 150 millimetre allowance without tilting reads zero on every angle channel on it, having consumed the whole vertical budget, while a berm that tilts to its full 0.3 percent across an 8 metre gauge has consumed 24 millimetres of that budget and is 16 percent into it. The two readings rank the same yard in opposite orders, and the survey is the only instrument that holds both. The plan family is invisible for the reason the last section gave. And the rail’s absolute level is invisible after re-levelling, because the maintenance deletes from the rail exactly the accumulated quantity that the ground still holds: a channel installed after the third episode reads its own baseline zero on a berm that has already taken a large part of its design settlement. The monitoring plan needs the survey for the absolutes and the channel for the continuous differential, and the two meet in a benchmark that must survive the interventions.

Between surveys the channel does a specific job that neither the survey nor the machine can: it dates and locates the change. A survey cadence set against an 11.8 millimetre per year average is a year-scale interval, but the published remediation cases were triggered in weeks, by rainfall and by water system failures, and a yearly survey misses that event entirely. The channel’s contribution is therefore not precision, which the ranking in the next section shows it does not have at this tolerance, but continuous presence: the step, the rate change and the station where it happened, in time to bring the survey forward rather than after the machine has been running on a geometry the last survey no longer describes.

The mounting rule follows the same logic as the conveyor article’s stringer rule: instrument the fixed structure, not the machine. The sensor belongs on the rail beam, the sleeper or the berm, because the machine is the element that moves and is itself adjusted. Station spacing follows the criterion, at one station per 10 metres of warp, which makes a 600 metre runway about 60 stations a side before redundancy is considered. Cross level is the one quantity that pairing across the gauge would buy, and it is expensive: the two zero deviations add in quadrature to 0.071 deg typical and 0.1 deg worst, against cross criteria of 0.041 to 0.072 deg, so the pair’s zero budget alone can exceed the criterion. Where a transverse track frame or crosstie exists, one sensor on it reads the cross slope directly and the pairing cost disappears; where it does not, cross level is better left to the periodic survey and the pair used only as change detection.

8. The Datasheet Against a One Millimetre per Metre Criterion

The warp criterion and the cross criterion are of the order of one milliradian, the same class as the crane runway tolerances, and the datasheet behaves against them the way it behaved there. Against 10 millimetres in 10 metres, the resolution of 0.001 deg is 1.7 percent of the criterion and the accuracy of 0.005 deg is 8.7 percent, and then the zero terms arrive: the zero deviation of 0.05 deg is 87.3 percent of the criterion, the typical zero temperature drift of 0.08 deg over a 40 degree Celsius swing is 140 percent, and the published maximum of 0.13 deg is 227 percent. An absolute compliance statement at this tolerance is not available from this instrument family on a permanent mount in an outdoor thermal environment, and the criterion, not the datasheet, is what decides that.

The conclusion here is constructive rather than negative, and this family is where it becomes so. A rail that is re-levelled on a schedule does not need an absolute geometry instrument, because the absolute geometry is a maintenance product rather than a measurement. It needs a comparator in time: what has changed since the last episode, at what rate, and whether the residual drift is approaching the rail tolerance or the ground criterion first. That is the question the zero terms allow the instrument to answer, because a stable zero error cancels in a difference against the instrument’s own baseline, which is the argument the calibration and crane runway articles made for reading against a baseline rather than against gravity, and it lands here for the fourth time in the series, this time on an object whose reference is reset by people.

The deployment shape follows the phenomenon. Consolidation and creep are slow, so a daily heartbeat is generous and the published duty cycle arithmetic of the wireless family, a 19,000 milliampere hour cell with a six year life at the default heartbeat and an alarm delay adjustable from 0.3 to 25.5 seconds, reads here as: the event of interest is a rate and a step, not a sample, and the interval should be chosen for the regression rather than for the peak. The company’s honest position is the same one the pile and conveyor articles took: no stockyard or stacker deployment is published, and this article claims none. The instruments named are the models with live datasheets, the ZCT2xxM-LBS-Ax-H5-460x family for the figures in the table and the ZCT-IOTH-WN-2x-JLA8 and ZCT330Mx-SWP wireless families for the deployment shape.

9. What to Specify

Specify on the differential, not the total. The contract quantities that matter are the ones the published tables already carry: a warp and gradient criterion of the order of 10 millimetres in 10 metres, a cross criterion of 5 to 10 millimetres with the gauge and the governing document named, a berm tilt criterion, and a horizontal displacement criterion with its chord. Name the document before the number, because the same cross level quantity spans a factor of two across the three sources quoted here, and the cap on the gauge fraction makes the absolute figure the one that will actually be enforced.

Specify the record, not just the channel. Re-levelling episodes, grinding, shimming and ballast work must be logged into the monitoring record as annotated events with dates and magnitudes, and the yard tally must be available to whoever interprets the data. A tilt channel on a rail without the episode log is worse than no channel, because it will read maintenance as recovery, and a berm tilt interpreted without the yard log will attribute to the ground what the pile schedule explains.

Specify the division of labour. The tilt channel is continuous on the differential family, the survey is periodic on the absolute and plan families, and the two meet in the benchmark. The alarm on the channel is a maintenance scheduler, a rate or a residual against tolerance, and not a safety threshold, because the safety family, derailment through gauge and offset, is in the plane the channel cannot see. Water belongs in the specification too: rainfall, hydrant leakage, sprayer operation and flooding triggered or aggravated the published episodes, so a tilt acceleration following a wet event is a ground water event until proven otherwise.

Monitoring of this kind does not replace the machine supplier’s rail survey, the geotechnical assessment of the yard, the structural inspection of the machine, or the drainage investigation that several of the cited remediation projects identified as the root cause. It tells the maintenance department when the next re-levelling episode is due, which is a scheduling answer, and it does that job well.

Frequently Asked Questions

1. How much rail deviation can a stacker or reclaimer tolerate? The published installation figures run to cross level differences of 5 to 10 millimetres, warp of 10 millimetres in 10 metres, gauge of 3 to 6 millimetres at installation and 10 in service, and full length straightness of 30 millimetres new and 60 in service. A 10 millimetre cross difference on an 8 metre gauge already moves a 25 metre boom tip 31 millimetres, so the operating manual of the specific machine governs, and the looser in-service figures assume a maintenance regime that keeps them true.
2. Can tilt sensors replace the periodic rail survey? No. The channel sees the differential family in the vertical plane continuously, and the survey sees the absolute level, the gauge, the plan straightness and the benchmark. Cross level by paired sensors carries a zero budget of 0.071 deg typical in quadrature that can exceed the criterion, so it is best treated as change detection between surveys, with the cross level itself left to the level instrument or read from a transverse track element where one exists.
3. Where should the sensors be mounted? On the fixed structure: the rail beam, the sleeper or the berm, not on the machine, which moves and is itself adjusted. Station spacing follows the warp criterion at roughly one station per 10 metres, and two rails need two chains. A sensor on a transverse element that spans the gauge reads the cross slope directly, which avoids the cost of pairing zeros.
4. What rate should trigger action? Work from the budget, not from a fixed rate. An improved yard design published 150 millimetres of berm settlement over 17 years, about 12 millimetres per year, with three re-levelling episodes in a 50 year life, so the action question is how fast the residual drift is consuming the distance between the rail tolerance and the last reset, and how fast the differential is consuming the 1 in 500 over 25 metres that the ground criterion allows. An acceleration after rainfall, hydrant leakage or flooding is a water event in the ground, and it is the pattern the published remediation records share.