Surface repair: the costs of cutting corners

Screed flaws

30 November 2018

Rectifying unbonded screeds in industrial premises is a fraught process and due care needs to be taken, cautions Trevor Rushton 

Early in my career, I was involved with the rectification of a failed concrete topping that had been applied to an existing slab in a small, light-industrial unit. Having recently let the refurbished building the landlord was less than happy with the performance of the floor topping, which rapidly developed cracks and became uneven, particularly at bay joints. 

The floor had to be replaced at a cost much greater than the original installation a problem that could easily have been avoided had the designer followed the published guidance available.

Earlier this year, I was involved with the diagnosis of similar problems in a much larger warehouse, where cracking and delamination of a floor had defied all attempts at rectification work. Again, the designer had ignored common sense and relied heavily on a proprietary screed material, an approach that was doomed to failure and which will inevitably lead to a significant loss.

Perhaps it is uncharitable to say so, but the 1st case I mention was the product of ignorance combined with the effects of poor-quality work and a failure to apply a curing membrane once the topping had been laid. 

The 2nd was one of misplaced optimism in the claims of the screed manufacturer, coupled with a fundamental failure to appreciate that the selected product was not being used for its intended purpose. 

Both examples illustrate the difficulty of restoring heavily used industrial floors to meet the punishing demands of material-handling equipment (MHE), particularly of the type that uses hard composition wheels and requires flatness. 

If one decides to deal with the problem by applying a hard-wearing, cementitious topping then great caution is needed; the process is fraught with pitfalls, and unless scrupulous attention is paid to design and the quality of work then subsequent failure is a significant risk.

Concrete topping 

My 1st example involved the application of a concrete topping of around 75mm thickness to level the floor and deal with surface damage following a fire at the unit. In this case, the topping was intended to be bonded to the sub-floor using a proprietary bonding agent. 

To limit shrinkage cracking, the topping was divided into bays by cutting with a saw before the concrete was fully cured. However, it subsequently shrank nonetheless, causing it to curl, with the result that it delaminated from the base and created lips at bay joints. 

The raised edge then broke down under the action of the MHE and cracked. Subsequent investigation showed that the bonding agent had been allowed to go off before the topping was laid, while the mix proportions and aggregate size were poorly selected for the particular use in this case. 

The floor topping was thus broken up and removed to be replaced with an improved mix, and great attention was paid to bonding and subsequent curing; it has performed well since. 

Unbonded screed 

The 2nd example involved a different type of finish, namely an unbonded screed. The warehouse in question had previously formed part of an automotive workshop and was finished with quarry tiles that were deemed unsuitable for the new use.

The designer decided to provide an unbonded screed over the tiling and specified a proprietary hydraulic binder to try to reduce cracking. The manufacturer's data sheets claimed that the binder was suitable for screeds down to 35mm thickness, so at first sight it seemed a perfect solution. 

The product could be used in combination with polymer fibres to control cracking further, and it was claimed that its rapid setting would enable the laying of floor finishes between 1 and 4 days later. 

A specialist screeding company laid the product and achieved a high standard of surface finish. However, one can see a clue to the later failure in the wording of the product literature 'enable the laying of floor finishes'. This was a warehouse floor, and one does not normally expect to apply finishes to surfaces that would be exposed to punishing wheel loads.

Further clues are provided by the manufacturer 'the product is as defined within BS 8204-1 section 5.1.3 part f'. This is the relevant code for in-situ screeds that are designed to receive a wearing course; that is, they are normally covered with other forms of flooring such as tiling, timber and carpet.

As part of the laying process a warehouse floor is normally traversed with a power floating machine used to produce a smooth, dense and level surface finish to in-situ concrete. A dry shake powder may then be applied to improve wearing, but it is very unlikely that there will be further finishes unless of course epoxy or polyurethane coatings are used. 

Essentially, a product that is designed for use as a screed, and which will normally have mechanical protection, is not likely to perform well in circumstances where it is used as a wearing finish in its own right. 

In relation to unbonded screeds, BS 8204-1 recommends that a screed should be no less than 50mm thick at any point except when laid on a bonding layer such an epoxy resin. Since it is rare to find a perfectly flat or uniform surface it is safer to assume a design thickness of 70mm or greater. In my example, the screed had been laid on a polythene separating layer so there was no question of it being bonded. 

The code goes on to emphasise the high risk of the screed curling with unbonded and floating levelling screeds, which can lead to steps at joints. Where curling cannot be tolerated, a concrete overslab designed as a new floor at least 100mm thick will need to be specified. 

One can appreciate the difficulty faced by the designer: the warehouse was linked to an adjoining building via openings in the party wall, and a ramp that could accommodate the 100mm change in level would have intruded significantly into the useable floor space. 

Fundamental error

However, in selecting a product intended as a screed, the designer made a fundamental error.

The correct code would have been BS 8204-2, the single digit making all the difference. This part of the code gives recommendations for in-situ direct finished base slabs with concrete as a wearing surface as well as wearing screeds (concrete toppings). 

Of these recommendations, there are 2 basic forms: 

  1. monolithic construction: the wearing screed is applied while the base is still plastic, probably no more than 3 hours after laying;
  2. separate bonded construction: the wearing screed is laid onto a set concrete floor, which has been prepared in such a way as to obtain the maximum bond that is practically possible. 

Option 1 was clearly impossible in this case, while option 2 would probably have necessitated the removal of the old tile finish. The code suggests that in instances where laying a bonded screed is not possible then a new overslab needs to be provided. 

It will be apparent that a new overslab is something entirely different from the 35mm screed that had originally been selected as a quick fix. The minimum thickness of slab would be 100mm, but to reduce the risk of curling this would need to be increased to 150mm. 

Failure of the 35mm screed was probably inevitable, not only in terms of mechanical performance but also in terms of wear. Cracks formed in the screed parallel with the bay joints and about 600mm from them. These cracked areas were very hollow and had clearly curled. 

To try to deal with the problem, the contractor cut out and re-laid some of the affected areas and treated the whole slab with an additional coating. Although this may have served to reduce dusting it offered no mechanical strength, and shortly afterwards the repaired areas began to fail. 


Curling of slabs usually occurs as a result of differential drying shrinkage and can be exacerbated by poor curing practice. Allowing the surface to dry quickly – particularly in warm weather – can affect the degree of curling, with the top surface contracting at a faster rate than the bottom surface. 

As the curled screed was trafficked it effectively broke its back, leading to cracks. Over time and repeated use, such cracks will gradually interlink, creating myriad small areas of broken screed that can then be dislodged by the MHE. 

In conclusion, while the provision of a replacement screed may seemingly be a simple solution, such work is fraught with difficulty. Solutions involving a thin, unbonded cementitious screed are likely to prove unsatisfactory, so if you compromise on the design you can expect to pay more in the long term. 

Trevor Rushton is a partner at Watts

Further information