1
2
ABSTRACT:   This paper examines the impact of using steels that are graded by selection on the capacity factors ( φ ) of
AS 4100 and NZS 3404, the respective Steel Structures Standards for design, fabrication and erection of structural steelwork in Australia and New Zealand. It details how the proposed changes to the Product Conformity requirements in the Australian/New Zealand structural steel products Standards referenced by AS 4100 and NZS 3404, reduce the risk of non-compliant steels that negatively impact on the intended capacity factors/reliability indices in Engineers’ designs to the Steel Structures Standards.
KEYWORDS: Grading by selection, type testing, factory production control, test certificates, compliance, conformity
1
Anthony Ng, OneSteel Manufacturing, an Arrium Company. Email: NgA@OneSteel.com
2 Arun Syam, OneSteel Manufacturing, an Arrium Company. Email: SyamA@OneSteel.com

“Grading by selection” violates the statistical methods used to calibrate the capacity factors (Ø) in AS 4100 [1] and NZS 3404 [2], the Australian and New Zealand Steel
Structures Standards, and hence may lead to the design limit states being exceeded for a structure. The practice of grading by selection [3] is described as choosing product that has been manufactured to a particular grade and Standard and making claims that a batch of product meets a higher grade or Standard based on one or two tests from that batch. Using steel that has been graded by selection in structures designed in accordance with
AS 4100 and NZS 3404 may lead to a reduction in its reliability, overstress and in extreme cases collapse or failure of the structure.
Changes to the product conformity requirements in
Australian and New Zealand Standards covering structural steel products have been proposed to address the issue of grading by selection. At the time of writing this paper, the draft Australian and New Zealand structural steel product Standards AS/NZS 1163 [4];
AS/NZS 3678 [5]; AS/NZS3679 Parts 1 [6] and 2 [7]
(herein referred to as - the “Steel Product Standards”) incorporating these changes have been released for public review. The combination of three key requirements in these proposed Standards are designed to inhibit grading by selection. These requirements are:-
1.
Type Testing;
2.
Factory Production Control; and
3.
Test and Inspection Certificates
An additional measure that Engineers can adopt to provide further confidence that product supplied for a structure is compliant with the Steel Products Standards is by specifying that product manufacturers are to be third party certified. This provides independent assessment that the three key requirements are met.
[8] which include:
1.
Variations in material properties;
2.
Differences in behaviour of isolated members compared with members in a structure;
3.
Simplifications and inaccuracies in design models;
4.
The actual degree of ductility and stability of a member;
5.
Eccentricities due to product and building tolerances.
(a) Probability distribution of design actions
(b) Probability distribution of design resistance
(c) Probability distribution of design actions and resistance superimposed
Figure 1: Probability distributions
AS 4100 and NZS 3404 (herein referred to as - the
“Steel Structures Standards”) are limit state design
Standards. For a structure subjected to actions or loads, the structural steel elements and connections are designed to ensure the structure is within the limit states for strength, stability, serviceability, brittle fracture, fatigue, fire, earthquake and durability. Put simply, the design action effects must be less than or equal to the design resistance (S* ≤ ØR u
).
Uncertainties relating to both the actions or loads and the actual capacity of the resisting members complicate the design process. This is resolved by using a probabilistic approach in design. The design actions are considered as having a probability distribution as shown in Figure 1a.
The expected load is represented by S* on the curve, while the upper and lower limit represents the uncertainty which arises due to the lack of control or incomplete knowledge of the actions. The design resistance is also subject to variabilities and uncertainties
Therefore the design resistance is considered to have a log-normal distribution [9] as shown in Figure 1b with the expected value represented by ØR u
on the curve and the uncertainty represented by the upper and lower limits.
Figure 1c shows the superposition of the two probability distributions. The probability of the design actions exceeding the design resistance is represented by the shaded area where the two curves overlap.
The capacity factors in the Steel Structures Standards for strength limit state designs have been calibrated such that a probability of failure (shaded area in Figure 1c) is in the order of 0.001. The process referred to as “code calibration” in the commentary of AS 4100 [8] involved the calculation of safety indices for known successful

designs using previous Standards, such as AS 1250 [10].
These safety indices were adopted as the target safety indices for the limit states Steel Structures Standards.
Capacity factors were then selected so that they provided a safety index matching the target requirement. The target safety index (now known as the “reliability index”), is in the order of 3 to 3.5 which is equivalent to a 0.001 probability of failure. Australian Standard
AS5104 - 2005 [11] provides guidance on the principles to determine the reliability index.
The Steel Structures Standards therefore have a safety index calibrated on the basis that compliant steel product will result in a log-normal distribution design capacity as shown in Figure 1b. Products produced by a manufacturing process which does not ultimately produce log-normal probability distributions will not be consistent with the assumptions made in the code calibration and therefore will not necessarily achieve the intended safety index.
15]. It is hard to imagine the extreme example of redesignation of the weight and dimensions of a product which are easily checked, and therefore would rarely, if ever, be exploited. The product grade is difficult to check without sampling and destructive testing, and therefore may not have the same protection.
The production of structural steel sections and plate has a long history where manufacturers have processes and production procedures to ensure the variations in grade are such that their minimum values exceed the required nominal grade values. These procedures which are commonly referred to today as factory production control results in the manufacture of products with resistance distributions consistent with those shown in
Figure 1b. More importantly for designers this distribution is consistent with the code calibration of the
Steel Structures Standards. It is presumably based on this history, that the current Steel Product Standards [12 –
15] had no reason to include explicit requirements to manufacture product with a minimum yield strength and an approximately normal distribution.
The current requirements for demonstrating conformity in Appendix B of each Steel Product Standard [12 – 15] is at best limited. It references only batch testing to demonstrate that the mechanical properties conform to the grade specification within the Standard. For example, the current version of AS/NZS 3679.1: 2010 [14] requires just two tensile tests per batch of product to demonstrate product conformity. It requires that one test be taken for a batch size of under 50 tonnes and two tests taken if the batch exceeds 50 tonnes. It is a concern that
The intent of the current AS/NZS 3679.1:2010 Standard
[14] is that the two test results taken for each batch over
50 tonnes confirms that the factory production process remains in control. The important data is not the values of the two results in themselves, but how they fit the existing long term distribution of results. The test results being consistent with the long distribution of data and the stability of the average and coefficient of variation are confirmation that the manufacturing facility is continuing to produce compliant product. no other testing or inspection is specifically required.
Hypothetically, in the extreme case, a manufacturer of
Grade 235, 250 or 275 universal sections could, for a typical 120 tonne batch of 310UB40 beams or 3000 metres of beam, just test two sections of beam approximately 0.5m in length. If those two tensile tests return yield strength results of say, 350 and 355 MPa,
AS/NZS 3679.1:2010 [14] does not currently restrict that material from being marketed as a 310UB46
AS/NZS 3679.1 - Grade 350. That is, both a heavier metre weight and a higher grade.
The intent of the Product Conformity provisions in the
The global steel economy, driven by intense competition, has increased the pressure for some organisations to push past the boundaries and exploit the limits where the market or regulations are unable to ensure the intended product requirements are met.
Structural steel is seen by many as a commodity product; so long as it meets a minimum standard, it is comparable, only by price. Without clear and specific requirements, what is thought to be, and is by definition - compliant product, may in reality not be compliant for its intended purpose. current Steel Product Standards [12 – 15] is clearly not to be determined by only using the provisions of
Appendix B in AS/NZS 3679.1 [14] and this needed to be addressed in the revision to that Standard [6]. Almost all manufacturers recognise the implied requirements in the body of the current Steel Product Standards [12 – 15] are to be complied with to demonstrate product conformity even though they are not specifically referenced in Appendix B of those Standards. These include the manufacturing requirements; chemical limits; dimensional tolerances, product straightness, other mechanical properties, traceability and minimum information required for supporting documentation.
Fortunately most responsible manufacturers, who want to protect their reputation and brand, will not exploit the deficiencies in the current Steel Product Standards [12 –
Grading by selection is a practice which has the potential to exploit the limits of the current Steel Product
Standards [12 – 15] and as such, is a cause for concern because it would contravene the assumptions made in the design code calibrations and the resulting capacity factors. Grading by selection at its very best, results in a left truncated normal distribution curve and at its worst offers a pair of tensile test results that provide no meaningful statistical significance for design. The concern arises because it is clearly not the intent of the
Steel Product Standards [4 – 7, 12 – 15] to produce compliant products in this manner and it creates a potential public safety issue for designs using the Steel
Structures Standards.

To illustrate grading by selection, an example of how a mill or distributor may market product using this process is described. The long term histogram of yield strength for product manufactured by a mill is shown in Figure 2.
Product is manufactured to a generic standard targeting a grade 235 MPa which is a common grade in Asia. Each batch of product is then batch tested in accordance with
AS/NZS 3679.1 [14]. As with all steel manufacturing processes, there will be an almost normal distribution as shown in Figure 3a. This distribution shows that some of the batch samples tested will exceed 300 MPa. The process of grading by selection will mean that these batches will be deemed in this grading process to be
Grade 300 MPa as shown in Figure 3b. It is important to note that this ignores any variation of yield strength within that batch.
Figure 2: Typical yield strengths for 235MPa steel
60
50
40
30
20
10
0
300 MPa
30
20
10
0
300 MPa
240 250 260 270 280 290 300 310 320 330
Yield Strength (MPa)
300 310 320 330
Yield Strength MPa
(a) (b)
Figure 3: Distributions of 235MPa grade product re-graded to 300MPa product, post-manufacture
To appreciate the consequences of product graded by selection, it is prudent to compare it with product that is manufactured in the traditional way that produces a normal distribution of yield strengths. This will be referred to as compliant product. Grading by selection uses testing to select the minimum yield strength and thus produces a left truncated distribution curve as shown in Figure 3b. This is compared with the intended normal distribution curve shown in Figure 4a. The difference between the two curve comparisons is plainly evident.
If these curves are then considered in respect of design code calibration the differences are even more alarming.
In section 2, it was described that the log-normal distribution of design resistance capacity were due to a number of uncertainties. To simplify this particular comparison we can consider only the uncertainty of the yield strength which is normally distributed and assume all other uncertainties will be the same for both cases.
Figure 4a shows the probability distributions of the design actions and resistances for compliant product using this simplification. Figure 4b shows the
(a) Probability distribution of compliant product
(b) Probability distribution of graded by selection product
Figure 4: Probability of failure represented by the overlapping curves

probability distributions of the same design actions and resistance of products derived from grading by selection.
If we recall that the probability of failure is the area where the probability distribution curves intersect it is clear that grading by selection will have a significantly higher area and hence a significantly higher probability of failure.
For various reasons, including legal, the transgressions as noted above are typically not documented. However, significant anecdotal evidence through the industry supply chain indicates that this is wide-spread practice.
Coupled with this is the fact that the current Steel
Products Standards [12 – 15] are not explicit on the situation. Though the above example considered
Universal Beams [14], the same situation occurs for steel materials in all four Steel Product Standards [12 – 15].
Concerns were raised recently in Australia for a significant project specifying AS/NZS 1163-C450L0
[12] Rectangular Hollow Sections (RHS). Issues with sourcing some sizes of this product resulted in the supply of overseas sourced RHS manufactured to AS/NZS
1163-C350L0 RHS [12]. Test results indicated that the product had a yield strength of 460MP based on a single tensile test. This was incorrectly considered acceptable by some project participants, however, this is statistically unreasonable and, as described above, technically untenable with the intent of the Steel Structures
Standards and Steel Product Standards.
Consequently, downstream (or post-mill) testing and inspection of manufactured products should not be used to (re)grade or categorise the product. It is clear that grading by selection makes the code calibration to produce the required capacity factors in the Steel
Structures Standards invalid.
Furthermore, variations within a batch are not considered in the grading by selection process. Any variation, which is inevitable, will further reduce the reliability factor of a structure designed to the Steel Structures Standards and constructed with products that are graded by selection.
The authors are currently collecting data on variations within a batch to determine the magnitude of this variation and the significance of any further reduction in reliability.
The initial type testing is designed to prove the capability of a facility to manufacture products to the required Standard. The requirements in the draft Steel
Products Standards [4 - 7] are not limited to grade. It covers all aspects of conformity including, the manufacturing process, chemical composition, tolerances, freedom from defects, mechanical properties and labelling for identification. The minimum testing frequency required will generally provide sufficient data to generate a normal distribution curve. Typically this is thirty samples, a figure most statistician use as a rule of thumb for statistical significance and is consistent with the requirements in EN 10025 and EN 10219.
The requirements for factory production control ensure that the facility has processes in place that allows it to consistently meet those requirements. The production tests and inspection processes which are used by manufacturers to monitor their facilities’ continuing production of compliant products is now specifically referenced in the draft Standards.
The requirements that a test certificate is available for products compliant to the Steel Product Standards and mandatory labelling provide greater confidence that compliant material is supplied. The product identified by the manufacturer’s certificate provides the link between the manufacturing facility’s type testing to prove capability and production testing to validate continuing factory production control and hence product conformity to the required Standard. The certificate must also be directly linked to the individual product length being assessed.
Specifying third party certification gives further confidence that compliant steel products consistent with the Steel Structures Standards are supplied. This certification usually involves annual assessment of the facility and the products manufactured along with ongoing surveillance of the product. This assessment ensures that type testing has been carried out and that factory production control processes are in place thus providing independent assessment of the product’s conformity to the Standard. Third party certified products are available in Australia and New Zealand [18] for products produced to all four of the Steel Product
Standards.
The proposed changes in the Steel Product Standards [12
– 15] include requirements to address concerns around product conformity. To address the issue of grading by selection, initial type testing and factory production control have been proposed. It is noted that EN 10025
[16] and EN 10219 [17], published a decade ago, includes theses same requirements which effectively inhibits grading by selection when coupled with the requirement that test certificates are to be made available for all products claiming compliance with these
Standards.
AS 4100 and NZS 3404 have capacity (Ø) factors that are derived from a process known as code calibration to give a required level of reliability for a structure.
Grading by selection produces steel products which are not consistent with this code calibration process. At the time of writing this paper, proposed changes to the four
Steel Product Standards [4 – 7] are designed to inhibit the practice of grading by selection.

The authors would like to acknowledge Stephanie
Johnston for producing the diagrams in this paper.
[1] Standards Australia, AS 4100-1998 Steel structures
[2] Standards New Zealand, NZS 3404 Part 1:2009, Part
2:1997 Steel structures Standard
[3] Ng, A., Syam, A. and Taylor, C., ‘Product conformity and conformity assessment to get the steel specified’, Steel innovations conference 2013,
Christchurch New Zealand.
[4] Standards Australia & Standards New Zealand, DR
AS/NZS 1163 (Project ID: 101291), Draft for Public
Comment Australian/New Zealand Standard, 18 Sept-20
Nov 2013, Cold-formed structural steel hollow sections
(Revision of AS/NZS 1163:2009)
[5] Standards Australia & Standards New Zealand, DR
AS/NZS 3678 (Project ID: 101292), Draft for Public
Comment Australian/New Zealand Standard, 18 Sept-20
Nov 2013, Structural steel – Hot-rolled plates, floorplates and slabs (Revision of AS/NZS 3678:2011)
[6] Standards Australia & Standards New Zealand, DR
AS/NZS 3679.1 (Project ID: 101293), Draft for Public
Comment Australian/New Zealand Standard, 18 Sept-20
Nov 2013, Structural steel – Part 1: Hot-rolled bars and sections (Revision of AS/NZS 3679:2010)
[7] Standards Australia & Standards New Zealand, DR
AS/NZS 3679.2 (Project ID: 101294), Draft for Public
Comment Australian/New Zealand Standard, 18 Sept-20
Nov 2013, Structural steel – Part 2: Welded I sections
(Revision of AS/NZS 3679.2:1996)
[8] Standards Australia, AS 4100 Supplement 1 -1999
Steel structures commentary, second edition, March
1999, pg. 13
[9] Pham, L. and Bridge, R.Q. and Bradford, M.,
‘Calibration of the proposed limit states design rules for steel beams and columns’, Civil Eng Trans, I.E. Aust.,
Vol. CE28, No. 3, July 1986, pp268-274
[10] Standards Australia, AS 1250 Steel structures code,
1981
[11] Standards Australia, AS 5104: 2005 General principles on reliability for structures
[12] Standards Australia & Standards New Zealand,
AS/NZS 1163:2009 Cold-formed structural steel hollow sections
[13] Standards Australia & Standards New Zealand,
AS/NZS 3678:2011 Structural steel – Hot-rolled plates, floorplates and slabs
[14] Standards Australia & Standards New Zealand,
AS/NZS 3679.1:2010 Structural steel – Part 1: Hotrolled bars and sections
[15] Standards Australia & Standards New Zealand,
AS/NZS 3679.2:2010 Structural steel – Part 2: Welded I sections
[16] CEN (European Committee for Standardization) EN
10025 :2004 Hot rolled products of structural steels.
[17] CEN (European Committee for Standardization).
EN 10219:2006 Cold formed welded structural hollow sections of non-alloy and fine grain steels
[18] Australasian Certification Authority for Reinforcing and Structural Steels (ACRS). www.steelcertification.com