TR34 floor flatness surveys: how floors are measured, and when they fail

A warehouse floor can be strong, crack-free and well finished and still be the wrong floor. If it is not flat and level enough for the trucks that run on it, reach trucks lean at height, very narrow aisle (VNA) trucks have to slow down, and the clearances the racking was designed around start to disappear. In the UK the reference for how flat is flat enough is The Concrete Society's TR34, Concrete Industrial Ground Floors, 4th edition. Its surface regularity section, revised in this edition to reflect the European Standard EN 15620, sets out what a survey measures, how the results are analysed and when a floor fails. This article walks through it.

Flatness and levelness are not the same thing

TR34 treats them separately. Flatness describes variation over short distances, traditionally controlled over 600 mm. Levelness describes variation over longer distances, traditionally 3 m. A floor can be flat but not level: smooth underfoot, yet falling steadily across a bay. It can also be level but not flat, with the right heights at the grid points but undulations a truck wheel will feel in between. A survey has to control both. The methods also assume the floor is meant to be horizontal, not laid to falls.

The first question: free movement or defined movement?

Every TR34 survey starts with how the trucks will use the floor, because the two cases are specified and surveyed in completely different ways.

• Free movement (FM): trucks can travel in any direction. Typical uses are wide-aisle racking, factories, retail, low-level storage, marshalling zones and food distribution.
• Defined movement (DM): trucks follow fixed paths, most commonly in very narrow aisles, where the regularity of the floor is critical to how the truck performs.

A defined-movement specification only makes sense when the aisle positions are known when the floor is built. If they are not, TR34 says it is not appropriate to specify the aisles as defined movement. Areas outside the aisles, such as goods in, goods out and transfer areas, remain free-movement areas.

Free-movement surveys: Properties E and F

No survey can measure every point on a floor, so TR34 samples it in two ways.

• Property E controls levelness. It is the level difference between adjacent points 3 m apart on a grid, measured along the grid lines and not across the diagonals. A 3 m grid is set out over the whole floor and a reading taken at every point, using a precise level and staff or another method of appropriate accuracy. Areas within 1.5 m of a wall, column or other existing structure are not surveyed.
• Property F controls flatness. It is the change in level difference between two consecutive 300 mm steps, in effect a measure of curvature over 600 mm. It is usually measured with specialist digital equipment along a sample of the grid lines. The total length of these runs, in metres, must be at least the floor area in square metres divided by 10, spread uniformly across the floor with equal lengths in each direction.

On a 10,000 m² floor that means Property F runs totalling at least 1,000 m, over 3,000 individual readings at 300 mm spacing, plus a level at every point of the 3 m grid. The grid and the run positions must be recorded accurately so that the same points can be revisited later.

How a free-movement floor passes or fails

The results are judged statistically. For each property, the 95th percentile value is calculated: the value below which 95% of the readings fall. TR34 Table 3.1 sets the upper limits.

The floor is non-compliant if either 95th percentile limit is exceeded, or if any point on the Property E grid lies outside ±15 mm of a fixed datum. Where no original datum exists, the check is ±15 mm of the mean floor level.

Because the limits are percentiles, a compliant FM floor can contain individual readings above the limit; up to 5% of them can be. That is deliberate, because it judges the floor as a whole rather than on one bad reading. The ±15 mm datum check is the exception: it applies to every grid point.

Defined-movement surveys: every reading counts

In a VNA aisle the truck runs in the same tracks every time, so TR34 controls the floor along those tracks, using the truck's own geometry.

• Z is the distance between the centres of the front wheels, in metres. X is the distance between the front and rear axles, taken as a fixed 2 m.
• dZ is the level difference across the front wheels. d²Z is the change in dZ over 300 mm of forward travel.
• dX is the level difference between the front and rear axles. d²X is the change in dX over 300 mm of forward travel.

Aisles are surveyed over their full length along the wheel tracks, starting with the load axle at the first rack upright. The instrument is a profileograph, a specialist digital device taking readings at intervals of no more than 50 mm; d²Z and d²X are computed from the dZ and dX data. Unlike free movement, there is no percentile. The floor is non-compliant if any measurement in any aisle exceeds any limit.

The limits all follow from one figure, the permitted cross-aisle slope (ZSLOPE): 1.3, 2.0 and 2.5 mm per metre for DM1, DM2 and DM3. dZ is Z × ZSLOPE, d²Z is three-quarters of dZ, and dX is 2 × ZSLOPE × 1.1, which reflects the fixed 2 m wheelbase.

The reason for this strictness is the mast. TR34 explains that a level difference between the wheel tracks is magnified at the top of the mast in direct proportion to its height. Level variations also set up dynamic movement that can significantly magnify that static lean. Take a truck with front wheels 1.5 m apart on a DM2 floor. dZ can be up to 1.5 × 2.0 = 3.0 mm. With the mast raised to 12 m, that 3 mm becomes 3.0 × 12 ÷ 1.5 = 24 mm of static lean at the top, before any sway.

A free-movement floor is judged on its 95th percentile. A VNA aisle fails on a single reading.

Choosing the class

• If the racking layout is not known when the floor is built, TR34 suggests building to FM2, which limits the grinding needed if VNA racking is installed later.
• Higher classes have a cost. TR34 notes that tighter tolerances can bring a higher floor cost and construction methods with more formed joints, so specifying FM1 "to be safe" is not free.
• Specify defined movement only for aisles whose positions are fixed, with the class chosen from the racking top beam height.

When to survey, and why timing matters

TR34 recommends surveying within one month of completing the whole floor, or a major section of it, to check that the floor complies as built. Surveys during construction are also valid as a quality control check that the finished floor will meet the specification.

Timing matters because floors move. TR34 lists deflection of suspended slabs and settlement of the ground, piles or other supporting structures as having a significant effect on surface regularity. Edges and corners of panels change as the slab curls. Curling is caused by differential shrinkage, because the exposed top surface dries and shrinks more than the bottom. TR34 notes it can occur at any time up to about two years after construction and tends to be unpredictable.

So a survey taken eighteen months after handover measures today's floor, not the floor that was built. If there is ever a dispute over whether a floor met its specification, the survey carried out within that first month, on a recorded grid, is the evidence that counts.

Accuracy and what a report should contain

• All instruments should be calibrated to an accuracy of 0.1 mm, and all data reported to 0.1 mm.
• For free movement: the 3 m grid of levels and the Property E values shown on the building layout, with exceedances highlighted, and the Property F runs shown and linked to that grid, with values for each run and exceedances highlighted.
• For defined movement: results, typically graphical, related to the building layout, with a summary for each aisle and non-compliances highlighted.

TR34 notes there are no Standards covering the use of specialist floor survey equipment such as profileographs, and that third-party accreditation, such as UKAS, is available for these survey methods. If a report gives you only a pass or fail, or percentile values without the grid and run locations, it can't be checked and it can't be used to plan remedial work.

If the floor fails

For a free-movement floor, TR34 recommends examining the individual measurements in detail to judge whether the exceedances matter for how the floor will perform. In defined-movement aisles, it may be possible to grind the high areas or, in unusual circumstances, to fill the low ones. Where wheel tracks are ground or filled, the wheels must stay in full contact with the floor so that no transverse thrust or other stresses act on them. Any remedial work will affect the appearance of the floor, and the treated areas should be re-surveyed by the same method.

Working in Germany or elsewhere in the EU

On German projects, floor tolerances are usually specified to DIN 18202, Toleranzen im Hochbau. Its approach is different. Flatness is typically checked by measuring the gap under a straightedge (the Stichmaß) over set distances between measuring points, and its tolerances are limit values for individual measurements, not 95th percentiles over a sampled grid. For warehouses with guided VNA trucks, DIN 15185-1 sets the floor requirements. Results from one system cannot simply be converted into the other. If a specification calls up DIN 18202, EN 15620 or a client's own tolerances, the survey has to be designed around that document from the start.

Further reading

• The Concrete Society, TR34 Concrete Industrial Ground Floors: A guide to design and construction, 4th edition (2013, with later amendments), Section 3
• The Concrete Society, Concrete Advice Sheet 44, Curling of ground floor slabs
• BS EN 15620, Steel static storage systems. Adjustable pallet racking. Tolerances, deformations and clearances
• DIN 18202, Toleranzen im Hochbau – Bauwerke
• DIN 15185-1, Lagersysteme mit leitliniengeführten Flurförderzeugen – Anforderungen an Boden, Regal und sonstige Anforderungen