{This is a partial rewrite of report originally written in 2011 for a specific project, involving the testing of existing balustrade installations, which resulted in an uneasy feeling that few balustrades were BCA compliant, and that the testing was a hazardous exercise.  It has been put online so that it can be made into a live document and be regularly revised, and hyper-linked to explanations having greater clarity than I can write. More importantly it contains common recurrent issues which will be shifted to their own posts in the near future. }

Structural testing is to be carried out on existing installations, and as such it is desirable that the test does not damage the installation. Given that the design codes are based on probabilistic limit state methodologies, design by calculation is providing some allowance for the formation of plastic hinges (Z versus S) including the potential for fracture (fy versus fu). The testing of guardrails to AS1657 permits the guardrail to be considered acceptable if residual deformation is constrained to a limiting value when the unfactored nominal load from AS1170 is removed. The residual deformation permitted is relatively large and is not considered desirable for an installation which is in service. The testing requirements in the codes largely refer to prototype testing and proof testing, with no reference to retaining the tested product in an operational mode after such tests. For an existing installation non-destructive testing is preferable to destructive testing.

Given that the installations to be tested can be made from any material (steel, aluminium, timber, etc), it is considered that the 6 point force-displacement curve testing requirements of AS1720 be adopted. The commentary to AS1170.1 recommends a serviceability deflection limit (h/60+l/240) for handrails, for short term loading ψs = 1 from AS1170.0. When applied to the nominal load Q, this matches the unfactored test load in AS1657 for which residual deformation is permitted. To avoid permanent deformation therefore the test load should be kept below this value: this can be achieved by making the serviceability design load the seventh point, and dividing the serviceability load into 7 equal increments. If the 6 force-displacement points do not form a line passing through zero, then not in the linear elastic range of the material, and further investigation is required. If extrapolation of the best-fit line indicates that the deflection would be greater for the serviceability load permitted then the balustrade is non-compliant. If extrapolation of the best-fit line indicates that the deflection at the serviceability load will be less than permitted, then accept as compliant. Note, such compliance is based on the assumption that a larger structural section is typically required to control deflection than required to prevent collapse or fracture. There may be some sets of conditions under which this does not hold valid for example low loads with high deflections permitted. However AS1657 is not testing ultimate strength (eg. 1.5Q), therefore ultimate strength is not considered to be a critical characteristic for acceptance.

When this test indicates non-compliance, and it is safe to do so, then the full limit state serviceability (Q) and ultimate strength (1.5Q) tests can be carried out to confirm the initial findings. Such test may indicate that the strength of the balustrade is not an issue, but its serviceability deflection is. Since serviceability deflection limits are a recommendation in the commentary to AS1170, compliance is desirable but otherwise optional. Such tests are likely to render the structure in need of replacement, it is therefore preferable to approach the manufacturer if possible and test the product. Testing the product however does not confirm the adequacy of the installation: anchors and customised connections at site.

It should be noted that many balustrade systems are in the market on the basis of testing, since the prototype testing permits permanent deformation, stresses have entered into the plastic behaviour zone for the material and it is difficult to validate the structure by calculation. New codes have also introduced more statistical methods for prototype testing to allow for manufacturing variation. Further, design loads have been increased over those used in the past. The net result is that balustrade systems which may have been code compliant may no longer be so. Also irrespective of the compliance of the balustrade system the installation itself may be inadequate. Testing should therefore differentiate between the adequacy of the balustrade system and the adequacy of the installation, and also indicate compliance with prior codes versus non-compliance with current codes.

Structural Assessment

Determining the required loads for testing is not a simple matter because fitness-for-function is a matter of subjective judgement: different people will hold different views as to whether the product is fit-for-function. Regulations and codes of practice impose a lower constraint on that which can be considered acceptable level of performance. But subjective judgement is still present in terms of interpretation of the codes, the intentions behind the actual clauses, and the classification of objects against the codes. Codes of practice are typically an on going work in progress as ambiguities and contradictions are removed.

Whilst the magnitude of loads and the resistance of structural elements can be quantified, the assessment of structural failure is largely a qualitative and subjective judgement. A judgement which has been complicated by the change from permissible stress design philosophy to limit state design philosophy. The conversion of the Australian codes to limit state is considered a soft conversion, and has otherwise been calibrated against the permissible stress codes so that for the more common situations the two codes should give similar results.

This conversion can be illustrated by looking at the conversion for the assessment of bending. To the permissible stress codes the basic formula was:

M ≤ 0.6 fy.Ze


M = design bending moment

fy = yield strength of the materials

Ze = effective elastic section modulus

This formula indicates that the resistance has been reduced by the design factor 0.6, or on rearrangement of the formula that the load has been magnified by 1.67.

1.67M ≤ fy.Ze

This presentation tends to give the impression that the structure is designed to be stronger than it needs to be. But such view neglects variability and uncertainty in the magnitude of the yield strength (fy) and the similar variability in the magnitude of the elastic section modulus Ze. To allow for variability the Building Code of Australia [BCA:§BP1.2] requires resistance be determined from 5th percentile characteristic properties. Allowing for variability in the properties used to calculate resistance and the difficulty of directly obtaining 5th percentile values, capacity reduction factors Φ [phi] are used, whilst variability in load is allowed for by partial load factors ψ [psi].

ψM ≤ Φfy.Ze

Where typically to calibrate with the permissible stress codes the design factors are:

ψ = 1.5

Φ = 0.9

To give:

1.5M ≤ 0.9fy.Ze

This could otherwise be expressed more generally as:

95th percentile design action-effect ≤ 5th percentile resistance

Where the 95th percentile design action-effect has a 5% probability of exceedance over the expected life of the structure, and where there is a 5% probability that the 5th percentile resistance will be lower than used in design. Depending on circumstances these probabilities may be varied to reduce the risk that the resistance is less than expected, and reduce the risk that the load is greater than expected. Now with such a philosophy if the design action with the acceptable level of risk is known, then its magnitude can be used without need to magnify by the partial load factor ψ = 1.5, as is the case when using limit state wind loads.

Limit States

Limit state design materials codes are not based on pure linear elastic design philosophy, the section properties vary between elastic section modulus (Z) and plastic section modulus (S), whilst material properties vary between yield strength (fy) and ultimate strength (fu). {NB: For international readers, Z and S are not mixed up, that is the way round we use them.}

There are many limit states, but typically divided into 3 classes:

  1. Stability limit states
  2. Strength limit states
  3. Serviceability limit states

Between ultimate strength (fu) or fracture and the yield strength (fy) materials undergo plastic deformation, that is deformation which does not elastically recover when the load is removed. Below the yield strength elastic deformation occurs, which recovers after the load is removed.

Depending on the limit state and operating condition being considered the magnitude of the elastic deflection, the extent of permanent deformation, collapse or fracture may be considered a failure.

Probability concepts add a complication. The basic design concept is that the design action-effects should be strictly less than the fracture strength never equal to: since fracture would represent failure. The issue of how much less than the fracture strength the design load should be is a matter of subjective judgement: different materials codes and different countries have different values.

The limit state codes put the requirement in terms of probability due to the uncertainties of operating conditions and manufacturing variations. The actual fracture strength is uncertain and the actual operating load is also uncertain, and probability distributions typically deal with inequalities with no knowledge of equality: so always less than or equal to, or greater than or equal to. So if design action-effect exactly equals the design resistance at fracture, it doesn’t mean the structure will break, for there is a probability the structure is stronger than required. Also fracture itself does not mean the structure has failed, for the ultimate strength load is typically not the operating or service load.


This influences the nature of testing and acceptance. The permissible stress version of the aluminium structures code (AS1664.1), applies a load factor of 2 to the nominal design load (Q) used in testing to allow for variations in manufacture. However this code does not differentiate between prototype testing and proof testing. When looking at the limit state design codes, in general they all differentiate between prototype and proof testing. To allow for variability in performance of manufactured units from the tested prototype, different factors of variability are employed depending on the number of prototypes tested. The more prototypes tested, the closer the test load can be to the limit state design load. For proof testing the test load equals the limit state design load.

Destructive and Non-Destructive Testing

Missing from both the limit state and permissible stress codes is consideration of non-destructive and destructive testing. Whilst destructive testing can be used with statistical process control (SPC) to control the output of manufacturing, the destruction of the samples is wasteful of materials. Non-destructive testing is thus a preferable test method, more so if 100% testing is to be adopted. Non-destructive testing is also important for testing existing facilities for compliance. There is little point in testing something for compliance if the test is destructive and the item tested needs replacing. The purpose of the testing is, so that it is not necessary to replace the structure: so breaking the structure to prove its strength is of little value. Similarly leaving the structure permanently deformed is undesirable.

The test loads therefore need to remain within the elastic range of the material; most structural materials are linear elastic or if not so, then conservatively approximated to by a variety of methods. Most design for serviceability would tend to stay below the yield stress (fy) and within the linear elastic range. Whilst design for strength may permit exceeding the yield strength, allowing permanent deformation and keeping below the ultimate strength (fu) and if using probabilistic loading then fracture may be permitted. The platforms code AS1657, permits residual deformation at the serviceability loads (ψQ where ψ=1), therefore serviceability loads are unsuitable for testing of existing installations where no damage is required and the installation is to remain in operation.

Test Method: Linear Elastic

The serviceability load therefore becomes an upper limit on the test load, whilst the commentary to AS1170.1 provides recommended serviceability deflections (h/60+l/240) [AS1170.1§C3.6]. Taken together these provide a control point or reference point on a force-displacement curve. For a linear elastic material a straight line should pass through this reference point and the origin zero (0,0): no force then no displacement. If this range is divided into incremental parts, then a collection of force and displacement values can be obtained which will either lie along the reference line, or produce a line above or below the reference line.

Since the testing is applying a load and displacement is being measured, rather than plotting force against displacement, it is preferable to swap the axes and plot displacement (y-axis) against force (x-axis). If the test data produces a line below the reference line then the balustrade is stiffer than required and acceptable, if the test data produces a line above the reference line then it is weaker than required and unacceptable. If the test data does not produce a line, then testing not in the linear elastic range and something else needs to be done.

By testing in the linear elastic range an estimate of the stiffness of the installation can be determined without need to take the load to a magnitude that would permanently deform or fracture the structure. Noting that it is dangerous to fracture a balustrade meant to be protecting people from falling from the edge of a floor. By applying the load incrementally and plotting the test data immediately, it is possible to monitor the performance of the structure predict response to next load increment, and terminate the test if necessary. The first test point joined with a line through zero gives an initial indicator of future failure of the structure if testing continues. The second point when connected to the first point with a straight line, gives an indicator of whether the line passes through zero or not.

If it does not pass through zero, then not in the linear elastic range of the material or system, and collapse may occur suddenly without warning, therefore need to proceed with caution. Two points not producing a line through zero may just be a consequence of inherent variability in the test procedure, so the more points tested the more reliable the estimate of the best fit straight line.

Most of the materials codes do not require plotting a force-displacement curve, however the timber structures code (AS1720.1) does, given that many balustrade systems are of timber construction, adopting the timber structures procedures for all testing should not be a burden. The timber structures code requires a minimum of 6 test points not including zero[§D3.4]. It is here considered that 7 points be adopted, with the 7th point being the expected serviceability load and not tested: the last load tested being 6/7th of the serviceability load.

It should be noted that such a test is only checking the stiffness of the structure, there is still the possibility that the structure will collapse or fracture below the ultimate strength load (1.5Q), where Q is the nominal design load in the code. However, if the structure does not meet the serviceability check then it is highly unlikely to meet the strength requirements, and rejection of the balustrade as non-compliant is a reasonable conclusion. If the balustrade meets the serviceability check there is a small possibility that it will not meet the strength check, extrapolating the line to 1.5Q is not possible since we only have a known test load and no limit on displacement. Calculations could be used as a further check, but such would require material properties and section properties, information typically not reliably available for existing installations.

Not being able to fully confirm compliance is not considered to be a major issue, only structures approved as compliant should have been constructed in the first place. The reason for testing would primarily be to determine if the existing installation has experienced any deterioration in performance since original approval and construction. Additional reason for testing is to determine if an existing balustrade installation complies with the current loading requirements compatible with the current purpose of the building space.

The latter being important since by calculation, the designer is likely to stay within the linear elastic range of the material, whilst testing is likely to permit plastic deformation. Thus a design based on lower loading in previous codes may still be found acceptable to current codes using higher loads if plastic deformation is now permitted.

The deflection limit in the commentary to AS1170 is also only a recommendation, not a mandatory requirement. However given the recommendation is now available, it is likely that most new designs will comply. Given that the proposed testing is measuring deflection it is also considered preferable to adopt this deflection limit.

If the balustrade installation being tested fails the deflection limits, then the full strength tests can be conducted, with the portion of the balustrade damaged by the test being replaced if it passes the test. If the tested portion fails the test then the entire balustrade needs to be replaced with a balustrade system that is compliant.

Products, Systems and Installations

A balustrade is in some ways similar to scaffolding (AS1576), in that consideration of both systems and installations is necessary. The system comprises of the component parts which make up a balustrade, these parts can be arranged in different ways to produce different assemblies to suit specific purposes. The installation is a specific configuration of the system to suit a specific site and its interface with the site. A product in its most generic sense is the output of a production process. Here a more restrictive sense is adopted, whereby a product is a specific system configuration which is complete in its own right. A product only involves installation-design to interface with the site. With a system of components, design is required for the specific assembly as well as the interface with the site: typically system-design and installation-design are integral and cannot be separated.

Manufacturers of products can save their customers and agents a great many problems by having their product certified against the Australian Building Codes Boards (ABCB) CodeMark scheme. The components of a balustrade system have the potential to be CodeMarked but the system itself poses problems. The system needs designing and approving for a specific installation. With a product only the installation needs designing and approving.

A bolt is a simple structural/mechanical product, once upon a time every single nut and bolt had to be made to fit, and with standardisation nuts and bolts became interchangeable. Unless a special thread form is required for a mechanical application, no one sits down and designs a bolt. Selecting the number of bolts from the available sizes to suit the requirements of the assembly is system design if within the assembly and installation design if between the assembly and the site.

Similarly a guard railing post is a simple product that needs selecting to suit a specific application: the spacing of posts can be adjusted to suit the appropriate loading and available resistance of the post. Such design is representative of system design. Selecting a suitable connection type to fasten the post in place on site is representative of installation design.

An aluminium balustrade may form an acceptable system and installation if cast into a concrete slab. The same system however if welded to base plates and bolted to the slab may cease to be adequate: due to the reduced strength of the aluminium in the heat affected zone (HAZ) of the aluminium. Similarly an aluminium system specifically designed with base plates and anchor bolts, may still be inadequate as an installation due to the concrete slab not providing adequate depth for the anchor bolts, or adequate clearance from the slab edge.

Furthermore, whilst the installation design may adequately consider all the requirements for “suitability of purpose”, the actual physical installation may fail to comply with the design. Similarly, whilst the specification of the component parts of a system, and the system itself are properly compliant with the required codes of practice, there is still the possibility that the physical component parts do not comply.

It is therefore important that manufacturers and builders required to supply products, systems and installations compliant with technical specifications are able to validate such compliance. Materials going into manufacturing need to be validated, all designs need to be validated, and all installations need to be validated. Some system of traceability, documentary or otherwise needs to be implemented to assure the quality and compliance of the product (goods and services) supplied.

It is therefore recommended that manufacturers avoid simply getting calcs-for-council, but rather get a nationally recognised CodeMark for their manufactured structural products, and otherwise provide comprehensive technical specifications along with design manuals to cover system design and installation design. This will improve consumer confidence regarding a manufacturer’s competence to supply a given product.

The current interest in testing largely stems from a common industry belief that many of the systems do not comply with the codes. The approved specifications for the system/installations may have complied, but the product/system supplied and actually installed may have been a lower quality substitute. Once installed it is difficult to identify the grade of material used. One painted piece of timber looks just the same as any other, thus unknown whether it is grade F4 or F34, unless physically tested: similarly for steel and aluminium. It is therefore desirable to have some traceable evidence of the quality of materials and components supplied.

The proposed testing is also that for existing balustrade installations rather than for balustrade systems. There is the possibility that many of the installations will fail, and that the site itself is unsuitable for any of the available systems. For example as already suggested the installation may be too close to the edge of a concrete slab, or the slab not thick enough for required anchorage. In such situation a custom design will be required to suit the specific conditions of the site.

So manufacturers may have suitable systems, but otherwise their agents fail the product by providing poor installation design, and poor physical installation. So when testing installations, we need to be clear whether it is the installation or the system which failed before discrediting another manufacturers system.


There are products, systems and installations. The testing is going to be carried out on existing installations. For existing installations proof testing and prototype testing loading requirements are unsuitable. Destructive prototype testing is suitable for testing of systems and component parts. Destructive proof testing is suitable for testing output of manufacturing to control variation. Non-destructive proof testing is required for testing installations which are to be kept in operation after testing. Non-destructive proof testing is also preferable for control of manufacturing variation since it reduces waste by permitting the tested component if it passes the test to be made available for sale. Non-destructive proof testing also permits 100% testing if desired, whereas destructive testing can only be by statistical sampling of production.

Since the loads in the loading code AS1170 are largely concerned with ultimate strength, the use of such loads in testing would result in destructive testing: either resulting in fracture or undesirable permanent deformation. Whilst AS1657 permits residual deformation to remain after testing, this testing is for prototypes, there is no indication that the deformed post and railing, is permitted to be placed into use after testing.

Since the primary requirement of a balustrade is its deflection under operating or service loads, the ultimate strength test is not so important. The loads (Q) used in testing to AS1657 are the nominal unfactored loads, with reference to AS1170.0 this could be considered as being the short term serviceability loading with partial load factor ψs = 1. Given the residual permanent deformation permitted by AS1657, this is more a strength test than a service test. It is thus likely to be found that this is incompatible with the recommended serviceability deflection given in the commentary to AS1170.1 §C3.6, it is also likely to be incompatible with the typical ultimate strength requirement of 1.5Q. The plastic section modulus is 1.5 times the elastic section modulus (S=1.5Z) for a simple rectangle. So if a simple bar is adequate at load Q, it is likely to form a plastic hinge at 1.5Q, and collapse: for more complex shapes the ratio will be different. The ultimate collapse load of a balustrade is not so important as its performance or behaviour under its operating load. If a full plastic hinge forms in the post at service loading, then the post is no use, but if a hinge forms in the guard rail then it may still be capable of fulfilling its function. If a guardrail deflects too much under the loading of one or more persons, then a person may find themselves balanced precariously over open space and about to fall from height. The critical design issue is therefore constraining deflection and enabling people to maintain a solid footing. The maximum permitted deflection would therefore be less than half the width of a human foot: assuming a person could be side on to the guard railing. A span deflection ratio is therefore not a suitable criterion for acceptance checking. At present however that is all that is available.


The recommended test procedure therefore is to incrementally load the balustrade and measure the deflections and plot a displacement versus force curve: this should be a straight line. It is suggested that the load increments be 1/7th of the service load Q. The limit on the deflection at load Q is known. Without testing at the load Q, it is possible to extrapolate the straight line data, and estimate the deflection at Q, if this is within the acceptable limits then the balustrade is considered acceptable. If the extrapolated deflection exceeds that permitted at load Q, then the balustrade is unacceptable.

If the balustrade is found unacceptable by this method, and there is a suggestion that it needs to be replaced, then it is recommended that the rejection of the system/installation be confirmed by destructive testing. Destructive testing should only be carried out if it is safe to do so, if possible better to approach the manufacturer and conduct prototype testing at ground level.

The primary purpose of this report is to identify suitable loading criteria for testing of existing installations, for more details on testing requirements refer to the individual materials codes and AS1170.0:2002 appendix B.

The appendices of this report contain sample output from a spreadsheet to assist with testing and assessment. It should be noted that there is more than one design load involved and the maximum of these need to be taken into consideration for testing. The reactions from the handrail may produce the maximum loading at the post, whilst the uniformly distributed load (UDL) on the handrail needs transforming into a point load producing equal bending moment for the hand rail test. The loads in AS1170 cannot simply be extracted and applied as test loads.

There is little in the BCA and associated standards regarding the performance characteristics of guard railing and balustrades, there is thus scope to produce a more quality robust specification. In particular structural performance is not so much dependent on magnitude of loading but more on structural behaviour and mode of failure: as far as possible a structure should be designed to have a safe mode of failure.




[05/09/2013] : Original

Start Previous Barrier Design Next End