Alasdair’s Engineering Pages © A. N. Beal 2018    

Alasdair’s Engineering Pages

The Structural Engineer, Vol. 70, No. 3, 4 February 1992

Designing connections simply and safely

A. N. Beal BSc CEng MIStructE MICE

Thomason Partnership

Alasdair Beal is an Associate with Thomason Partnership, consulting engineers. After graduating from Glasgow University in 1975, he joined Freeman Fox & Partners, working in long-span bridge design. Since then he has worked for W. G. Curtin & Partners, Fairclough Building Ltd. and Strucom Ltd. At Thomason Partnership he has designed retail, commercial and industrial structures and prepared reports on a wide variety of structural problems.


The basic forms for steel connections have been established for quite a long time (endplates, seating cleats and web cleats) and they are usually of fairly standard proportions. However, in recent years, the author has noted the appearance of a significant number of connections that are oddly proportioned or unorthodox in concept. There is no harm in novel ideas if they are well executed, but unfortunately the opposite is often the case. Some examples (obviously anonymous) are illustrated (Figs 1-3). Of course, they are not typical of the majority of current steelwork practice, but such poor connections display a serious lack of understanding of engineering principles, and their appearance in significant numbers suggests that present-day detailing standards are not all they might be.

Connection design has an influence on safety out of all proportion to the cost of the items involved, so problems in this area demand particular attention from the profession. Recent research confirms the view that all is not well. Davies & Morris [1] commissioned an experienced portal frame manufacturer to design and build a frame and this was then test loaded .The all-important haunch connection began to fail at only slightly above the design working load. The mode of failure indicated that certain basic aspects of the structure had simply not been checked and later examination showed extensive yielding in several areas not involved in the actual failure. Distress in these areas at a load well below the supposed capacity of the assembly suggests that the design was inadequate in more ways than one (Fig 4).

If detailing standards in commercial practice are not as good as they should be, this is something which needs to be borne in mind when introducing new types of connection and recommending appropriate design methods.

Fig. 1

Fig. 2

Fig. 3

Fig. 4 (from Davies & Morris)

What’s the problem?

Part of the problem is economic: both fabricators and consultants have been forced by competitive pressure into severe cost-cutting and many have responded by leaving design work to inexperienced staff working without adequate supervision. To make matters worse, the 80s steel boom led to many concrete designers trying their hands at steel. The concept of a connection sometimes appears to be the product of inspired guesswork, with the detailer then striving to make the design as ‘economical’ as possible by looking for the formula or computer program which appears to justify using the thinnest plate, the smallest welds and the fewest bolts.

Complex formulae may be used without understanding their purpose or validity, and over-optimistic assumptions may be made, with Code safety factors being relied on to act as a safety net to cover the inadequacy. Thus in the design of moment endplates it is common to assume the presence of large prying forces in the bolts when designing the endplate (where they help to justify a thinner plate), but then to ignore these forces when designing the bolts themselves. Some even claim that BS 5950, cl.6.3.6 [2], permits this violation of the laws of statics and they say that the design safety factor is sufficient to cover the consequences. It is simple to demonstrate that prying forces can easily add 50-100% to bolt tension - far too much to ignore. Logically, if prying forces are relied on to make an endplate work, they must be allowed for in the bolt design, particularly in critical connections.

Another failing is to check the easy, obvious parts of the load path through a connection but to ignore the less obvious ones, even if they are equally important. Thus in portal frame haunches it is common to see bolt forces calculated and then bolts, endplates and welds designed to transmit these forces to the rafter web, yet the chosen area of web is not checked to see whether it is actually strong enough to receive the force being delivered to it. This can lead to the nonsense of an endplate attached by two 10 mm welds to a rafter web which is only 8mm thick. (An inspection of Fig. 4 will reveal yielding of the rafter web near bolt positions.)

Much of the blame for bad practice in connection design clearly belongs to the practising engineers responsible but some also rests with researchers. For many years, much research has been concerned with subjects of limited relevance to practical designers such as semi-rigid design. However, some of the more practically-directed work has sought to make common connections more economical than standard design methods would allow. This has used tests and yield line analysis to produce formulae for use in design, with rules-of-thumb to limit their application. While the intention of economising on materials is not necessarily wrong, an unintended consequence has been that practising engineers have been drawn away from traditional design methods based on simple structural theory to a more empirical approach. Instead of reaching for a standard text such as Steel Designers’ Manual [3]  or Boston & Pask’s Structural fasteners and their application [4] and checking the essential aspects of the connection on the basis of bolt force resultants, bending, shearing and bearing stresses, today’s detailer increasingly reaches for a formula or a computer programme ‘proved’ by research and claiming to be more economical than a normal design.

Theory good, empiricism bad?

It may seem strange for a practising engineer to bemoan a drift towards empiricism in connection design and to advocate a return to a more theoretical approach, yet there is method in this apparent madness. The battle between empiricism and theory is an old one, and the strengths and weaknesses of the two have not changed over the years, although they are often forgotten or misunderstood.

The virtue of empirical design is (perhaps surprisingly) its economy - designs proved directly by tests will always have the edge over those based on calculations which must always be conservative in their assumptions. Its weakness is its restricted application - a test will establish only the strength of the particular assembly tested and gives no guidance on the effect any variations might have. Thus, for example, if the tensile resistance of an endplate is being tested, the preload (if any) of the bolts and the clearances and fit-up of the assembly can have a major effect on the ultimate strength. If the bolts are not very tight and in comfortable clearance holes, the assembly can deform considerably and develop a V-shape to resist load in direct tension. Applying a substantial pretension to the bolts will make the assembly much stiffer, the ‘V’ flatter and the ultimate tensile strength of the assembly lower. The problem with an empirical solution is that, although it tells you whether something works or not, it may shed little or no light on how, or why, it works.

The weakness of theory-based design is its conservatism. However, its great strengths, which make it essential to the practising engineer, are its flexibility in making sense of myriad different arrangements of plates, welds, and bolts and the insight it gives into how a connection works. It is a grasp of how connections work which enables the detailer to get the concept for a connection right, so that the forces can be transmitted efficiently and safely, without causing overstresses on the way. It also allows him/her to understand why, say, a thin endplate gives large bolt prying forces and vice versa and why generous edge distance and close fit-up are essential if the required prying force is to be reliably developed. To produce good designs, the detailer needs a good understanding of how, in simple terms, connections work.

Back to the future

It may seem retrograde to advocate rejection of the various refinements researchers have developed in favour of designs based on simple, safe theory. Some will worry about the resulting loss in economy from designs which will sometimes be much stronger than the bare minimum proven in tests. However, a sense of perspective is essential in this area. The difference between a portal frame haunch connection designed carefully using simple, safe theory, and cutting no corners, and one designed ‘to the bone’ by the most optimistic treatment is commonly only the difference between making the endplate 25 mm thick and making it 20 mm thick. As fabrication costs are unaffected, although the difference in strength is considerable, the difference in cost is trifling. If the concept of a connection is sound, the saving to be made by fiddling about with the calculations is very small; if the concept is not sound, this imposes extra cost which no amount of clever calculations can outweigh.

A return to simple theory does not limit the variety of connections we may use - the design of endplates, web cleats, seating cleats and finplates is not difficult. It may be frustrating to the researcher or Code drafter not to see designers pursuing the last degree of economy; it may also frustrate the commercial designer seeking to save the last £50 on the cost of a job and thus win the contract. However, it is a matter of priorities. The first priority in the design of connections is to ensure that they are safe. If the examples shown earlier are in any way typical of current practice, there is an urgent need to forego the frills and the quest for ultimate economy in favour of consistent promotion of good details and the basic principles which lead to designs that are reliable and safe.

In the case of finplate connections, this means keeping them short and avoiding the ‘long’ variety as far as possible, not making them too thin and wobbly and taking care to calculate proper resultant bolt forces, making full allowance for load eccentricities and out-of-balance loadings.


1. Davies, J. M., Morris, L. J.: ‘Realistic modelling of steel portal frame behaviour’, The Structural Engineer, 68, No. 1, 9 January 1990.

2. BS 5950 Structural use of steelwork in building: Part 1: Code of practice for design in simple and continuous construction: hot rolled sections, London, British Standards Institution, 1990.

3. Steel designers’ manual, London, Constrado, Crosby Lockwood Staples, 1979.

4. Boston, R. W., Pask, J. W.: Structural fasteners and their application, London, British Constructional Steelwork Association, 1978.

The original copy of this paper is available from