Structural integrity is the study of the safe design and assessment of components and structures under load, and has become increasingly important in engineering design. It integrates aspects of stress analysis, materials behaviour and the mechanics of failure into the engineering design process.
This OpenLearn course is an adapted extract from the Open University course T357 Structural integrity: designing against failure.
After studying this course, you should be able to:
differentiate between and describe dissolution, degredation and corrosion as they affect the deterioration of structural materials
predict electrochemical behaviour between dissimilar metals
explain galvanic corrosion in terms of the electrochemical series
distinguish between the hoop and longitudinal stresses in a pressure-vessel wall, and specify them in terms of the pressure, wall thickness and diameter of the vessel
describe the loads in the various parts of a structure and the most likely load path.
This unit is about the concepts and theories that underpin the field of engineering known as Structural integrity – that is, the safe design and assessment of load-bearing structures in their entirety, including any individual components from which they may have been constructed. Aspects of structural integrity are implemented in almost every engineering design process, even if the engineer or designer does not necessarily think of it in that way. In this unit, we have separated the skills and knowledge associated with expertise in structural integrity under two headings: Stress analysis, which is the study of how applied forces lead to internal stresses in structures; and Fracture mechanics, which is the study of components and structures containing cracks.
In thinking about part (b) you should have come to the conclusion that virtually nothing is entirely load free. At the very least, any component or structure has to bear its own weight, irrespective of any external loads. A designer might make an intuitive judgement that the loads on a product do not need to be considered, but hopefully that would be underpinned by an educated estimate of what the forces are likely to be and what intensity of load, or stress, the assembly can support. And this is why the concept of stress is important: the limiting material property we are dealing with is ‘strength’, and for safe operation the stresses experienced during use need to be well below the material strength.
Engineering failures can be spectacular and highly publicised, especially when they result in death and destruction; but the failure of a household product can be more immediately annoying, and just as indicative of a poor design.
Make a list of three structural integrity failures of which you have experience. In other words, list three single items, assemblies or structures that have snapped, collapsed, fractured or just plain fallen to pieces in your home, car or workplace, for example. Try to make them as different as possible – so don't choose three smashed pieces of china. Try to think why these items might have failed, paying particular attention to how they were loaded during use and whether this loading was different when failure occurred.
You have obviously made a personal list and I can't really help you with that, but the process is useful because it has allowed you to begin to engage with how to assess structural failure.
You may be feeling pleased with yourself, or you may be thinking how little you know. In any case, I would like you to keep your list safe, along with your interpretation of each failure.
We have all experienced component failures in one form or another. In many cases this is because something has reached the end of its working life due to a slow-acting failure mechanism: car tyres wear slowly and will eventually burst if not replaced; the filament in a light bulb slowly loses material until it cannot sustain the applied voltage and melts. Failures where something has been so badly designed that it cannot withstand its intended loading during normal use are rarer, but they do occur nonetheless. Take a look at Figure 1, which shows the broken handle of a decorative cake knife, the sort that gets used only on ‘special’ occasions. In fact, this example of failure was caused by poor design. Note that a metal ‘tang’ extends from the blade into the handle as a means of reinforcement. In this case the tang was simply too short to strengthen the ceramic handle sufficiently against the bending loads that arose during cutting. The failure occurred while the knife was being used at a wedding reception and resulted in blood-soaked icing.
The elbow connector shown in Figure 2 is another case of poor design having disastrous consequences. The elbow was part of pressurised pipework used in a hydraulically powered waste compactor that ruptured during use, causing severe injury to the operator. Subsequent analysis indicated that the wall thickness and material properties of the connector were just not adequate for the job of containing pressures over 17 MPa (17 × 106 N m−2). You may have had similar experiences of structural failure, although hopefully more mundane. (In this unit we will use the shorthand ‘Pa’ or ‘pascal’ for the units of stress and pressure, N m−2 (newtons per square metre).)
The way the load intensity, or stress, varies within a material is also important. If we can understand and quantify the internal stress distribution then we are some way towards figuring out why the failure occurred. But not only that: we can also redesign structures so that they are less likely to fail next time round. Or better still, we can avoid it ever happening in the first place. In other words, stress analysis is a very useful tool. Often the calculation of stress is relatively straightforward, but for complicated components and systems of components, with multiple loads and varying material properties, more complex analyses are required.
It is often not possible to design something to withstand any foreseeable load. The laptop on which I am writing this is reasonably robust, but I would not expect it to keep working properly after I had reversed my car over it. Figure 3 shows a couple of examples of ‘overload’ failures that occurred because the structures were subjected to a loading for which they were not designed. A bicycle wheel is perfectly capable of carrying the load of two or three merry students provided that the loading direction is approximately in the plane of the wheel itself, where it is strongest. But a sideways impact can easily cause buckling (Figure 3a). Similarly, a stepladder leg is designed to bear heavy vertical loads, but a significant lateral force will cause it to bend (Figure 3b). More extreme, earthquake-induced, overload failures are shown in Figures 4 and 5. Of course, it is extremely difficult to design a building that will withstand any magnitude of earthquake, but there are design philosophies that can make them more ‘resistant’, including methods of damping vibrations and trying to ensure that damage, when it does occur, does not lead to total collapse of the structure. Nevertheless, under such circumstances engineers and designers are often faced with decisions about balancing effective design against cost and the likelihood of catastrophic failure.
Even a single column collapse in a multi-storey building places greater load on underlying storeys, which then collapse in turn leading to multiple floors stacked upon each other, like pancakes. This is due primarily to the use of low-cost structural columns of inadequate strength.
I indicated earlier that many failures occur after a product has been in service for some time: such as the wear of a car tyre, or corrosion of the car body itself. It is also possible for components to fail because of a combination of a manufacturing defect with the applied loading or with the environmental conditions during use. Figure 6 illustrates the link from mechanisms such as corrosion, fatigue (repeated loading) and creep (continuous deformation under load) to failure in some form.
So in addition to knowing the stresses in a material arising from the applied loading, depending on the environment in which the component is used it may be necessary to consider the effects of corrosion, wear, creep and fatigue. The effects of any of these mechanisms can weaken a structure to the point where it can no longer bear the loads for which it was originally designed, as shown in Figure 7.
The study of structures that contain cracks from the day they are made is sufficiently important for us to dedicate a large proportion of this course to it. Fracture mechanics allows us to assess whether cracks will be safe under the applied loads. But even in cases where we can be reasonably sure that there are no cracks of any significant size (say more than a millimetre), there is still the possibility that fatigue loading or another mechanism can cause them to grow, as illustrated in Figures 8 and 9.
There are also complex resonant loading cases that can cause failure. You may well remember the publicity surrounding the ‘wobbling’ of the Millennium footbridge in London when it was opened. More spectacular was the actual collapse of the Tacoma Narrows suspension bridge in America in 1940 (see Figure 10). The structure was perceived in its time as the pinnacle of structural lightness, grace and flexibility in bridge design. However, it met its end within four months of construction because of a woeful inability to cope with even moderate winds – only 45 mph on the day of collapse.
All the examples I have used here are of things that have failed, which is perhaps not the best illustration of the design process. However, learning from failures can give insight into why the failure occurred and how to avoid it in the future. Successful designs that simply do the job rarely make the headlines.
Have another look at that list of failures you made for Exercise 2. Are you any the wiser as to the origin of the failure, in terms of the loading on the component when it failed?
Keep that list tucked away safely, I'm going to ask you to have a look at it again later.
Structures are not always doomed to fail, but they do usually have a limited useful life. Exceptions include many of the monuments that have survived from the ancient world, such as the Great Pyramid in Egypt (Figure 11a), the Pont du Gard in southern France (Figure 11b) and the Pantheon in Rome (Figure 11c). These are very stable structures. The Great Pyramid has a very low centre of gravity and few potential failure modes, while the other two structures are based on the principle of the arch and the dome respectively, where most of the loads in the stonework are compressive by nature, thereby exploiting the high compressive strength of stone. Stone is also very resistant to deterioration, particularly in relatively dry environments (which is why these structures have lasted so long).
In this section we will examine some of the mechanisms of deterioration of structures, the effects of the way that structures are loaded on the process of degradation, and the ways structures can be protected against environmental attack. To help you to tackle the problems encountered by real structures in different environments, some background in the chemistry of materials will be provided to enable you to pinpoint specific mechanisms of deterioration.
The next section will introduce you to some of the language and mechanisms of corrosion and degradation. The remainder of the unit uses a case study to illustrate the various factors that can lead to failure in practice.
A variety of common terms are used to describe the ways in which structural materials can be attacked by environments and although they do have specific connotations, they are frequently used as blanket terms for material deterioration. I shall attempt to define them in a more specific way, namely:
Suggest appropriate terms for the following phenomena:
When real products are examined in detail, one is forced to examine the many specific mechanisms by which they can deteriorate. Rusting, rotting and dissolution are very common in practical experience simply because of the widespread use of steel, limestone and timber in structures.
However, in order to study these (and other) mechanisms, we need to apply more rigorous analyses. The point of study is to design ways of eliminating deterioration, or at least (if attack is inevitable, as it often is in practice) ways of controlling and reducing the rate of attack. Most structures need to have a protracted life, not only to justify the expense of their erection, but also to protect the users. One of the unfortunate features of structural deterioration is the insidious way in which attack can occur, often hidden from view, and proceeding at a rate that can result in sudden and catastrophic failure of a safety-critical component. We shall be examining some examples later in this unit.
For many materials, degradation processes are simply one or a series of chemical reactions that act to erode or deteriorate the material. The deterioration of metals is a little more complex than that of non-metals because metals are electrical conductors. Local electrochemical cells frequently form in the exposed surfaces of metals, leading to corrosion of the metal in one part of the cell. Electron movement is an essential part of the process: as electrons are lost, metal ions are formed, and these soluble metal ions then pass into the aqueous environment, resulting in a net loss of metal. Electrochemical cells were actually used by Volta to produce electricity (in the first batteries), so you can see that corrosion can be turned to advantage to make portable power sources.
The most familiar type of electrochemical cell to most people is the common battery. Batteries harness the energy released by corrosion of metal components, which is why they are usually heavy. An electrochemical cell contains two electrodes made from differing metals. When the terminals are connected, each electrode reacts with a current-carrying solution known as an electrolyte and the cell provides a current.
The two electrodes in any cell are known as the cathode and the anode. At the anode, the metal reacts and releases electrons. These electrons then flow through the connection to the cathode. We can write the anodic reaction in chemical shorthand as:
M → M+ + e−
where M is a metal, and M+ is a positively charged ion formed when the metal atom loses an electron e−. Different metals will lose different numbers of electrons in a corrosion process. The resulting metal ion may be lost into a solution, or form part of a corrosion product such as rust.
At the cathode, the reverse reaction occurs. Electrons are ‘absorbed’ by ions, causing a different reaction, which might be the plating out of a metal from solution.
‘Primary cells’ are non-rechargeable: an example is the zinc-carbon battery, with a zinc anode and a carbon cathode that are separated by an electrolyte gel containing a salt (ammonium chloride). The electrolyte effectively attacks the zinc to produce electrons that can be tapped off at will. A Daniell cell uses copper and zinc as the electrodes, the copper being the cathode and the zinc the anode (Figure 13). Here the ‘salt bridge’ allows ions to travel between the two solutions, thus completing the circuit.
The life of every cell is limited by the amount of anode present, because this is attacked and effectively disappears, corroding away until little metal is left. So here is a possible tool to assess corrosive activity. Some metals are clearly more reactive than others; in other words, they have a greater electrical potential. It is possible to create a table that allows the electrical potentials of different metals to be compared. This may be done by putting two metals into a cell in order to determine which will corrode in preference to the other. Each specific cell has a characteristic electromotive force (emf), also known as the electrical potential difference, which is measured in volts. This shows how much more reactive one metal is than the other.
In order for all metals to be comparable, they must be measured against a standard point. Thus a ‘hydrogen electrode’ provides an arbitrary zero against which the other corrosion reactions are measured, to produce a list of standard electrode potentials (E0) for different metals. This is shown in Table 1 and is known as the electrochemical series. The least reactive metals are at the top of the list; the most reactive are at the bottom. So, the more positive the standard electrode potential, the less likely a material is to corrode; the more negative the value, the more likely the material is to corrode. When two dissimilar metals are in contact, it will always be the metal with lower potential that corrodes.
When two dissimilar metals are in contact, or in close proximity with a conducting fluid in between, an electrochemical cell can be formed that leads to the more reactive metal becoming an anode and the less reactive metal a cathode.
This kind of corrosion is known as galvanic corrosion. It is not uncommon, since metals are often coated with others of different E0, and different metals are often in close contact with a common electrolyte.
One of the earliest examples of galvanic corrosion was recorded in the eighteenth century. The wooden hull of the Royal Navy frigate HMS Alarm (Figure 14) had been covered by copper sheathing, which was attached to the hull by iron nails.
One of the purposes of the copper sheath was to limit marine biofouling, which is known to plague many materials that are immersed in sea water. The growth of molluscs such as barnacles on the hulls of ships, which can then trap trailing seaweed, results in reduced speed and manoeuvrability. Copper limits fouling by inhibiting the attachment of molluscs. (Other organisms, such as bacteria, can also actually cause corrosion, as discussed in Box 2: Bacterial corrosion).
The hull was covered in 1761, and the copper sheath was found to be detached two years after fitting, during which time the Alarm had visited the Caribbean and elsewhere. The iron nails were found mostly to have corroded. Some nails remained intact, however, where their brown paper wrapping had remained in place between the copper and the iron, a fortuitous event that prevented total detachment of the sheath. The iron nails in contact with the copper were subject to rapid galvanic corrosion that led to detachment of the sheathing. The small anode (iron nails) to cathode (copper sheet) area ratio favoured the loss of the iron, as the rate of corrosion is directly proportional to the current density (a measure of electron flow). In a sense, the nails acted as local electron concentrators, so attack was rapid. Where it was present, the brown paper insulated the nails and so there was insufficient electron flow to cause corrosion.
The reason why marine environments are especially pernicious is the salt content of sea water. The presence of sodium and chloride ions increases the electrical conductivity compared with pure water, so galvanic or other cells formed between dissimilar metals react much faster.
An unusual and perhaps unexpected corrosion problem can be caused by bacteria. As one of the oldest groups of organisms on the planet, bacteria have evolved to survive even in extreme environments. Bacterial corrosion can occur in fuel tanks, for example (Figure 15): fuel oil contaminates bilge water on tankers, and bacteria then grow profusely in the mixture.
The bacteria feed on the organic oil, releasing mild organic acids and depleting the oxygen content of the water. The acids will accelerate corrosion of the steel container, but a more serious stage can develop when certain species known as sulphate-reducing bacteria take over. These reduce the oxygen content of the sulphates commonly present in dirty fuel oils to produce hydrogen sulphide, or H2S. This compound is potent at corroding steel and can also enhance hydrogen embrittlement (which is a form of stress corrosion cracking), attack usually occurring as pits in the metal close to or under the bacterial colonies. Such colonies are perhaps better known for the ‘rusticles’ they produce – as were present on the wreck of the Titanic (Figure 16). The colonies of bacteria live on the rust, and promote further rusting through chemical attack of the underlying steel.
Such bacterial attack can also cause disasters directly, as in the gas explosion near Carlsbad in New Mexico, USA on 19 August 2000. The natural gas was carried in a 760 mm diameter steel pipe across a river via a suspension bridge. The pipe fractured suddenly, releasing gas that ignited into a fireball, engulfing the bridge and killing 12 people. It left a large crater, at the base of which were found the ends of the pipe; the missing pieces were ejected by the explosion (Figure 17).
Analysis of sludge found in the pipe showed evidence of extensive microbial attack in the form of deep pits in the pipe wall, and the presence of various contaminants including chlorides, hydrogen sulphide and sulphates. The fracture had occurred at a deeply corroded section of the 7.6 mm thick wall, where the wall thickness had been reduced to less than 2.5 mm. The rupture took the form of a 525 mm long crack along the axis of the pipe, which was under an internal pressure of 4.65 MPa. Better inspection procedures were recommended after the accident, including the use of cleaning ‘pigs’, which travel within pipes, both monitoring internal problems and cleaning debris away.
Calculate the approximate hoop stress in the pipe assuming no wall thinning, and then the effect of microbial corrosion on the hoop stress when the wall thickness has reduced to 2.5 mm.
You should recall that the hoop stress in a cylinder is given by:
where p is the internal pressure, r the radius of the cylinder and tc the wall thickness.The hoop stress acts such that the pipe will fail by a lengthways crack.
With the data provided, assuming no wall thinning, then:
With wall thinning to 2.5 mm due to the effects of corrosion, however, at failure:
Another example of a structure that was damaged by galvanic corrosion was the Statue of Liberty in New York harbour (Figure 18). Built in 1886 by Gustav Eiffel and Frederic Bartholdi, it was composed of an inner wrought-iron framework, with an outer cladding of copper attached by saddles of copper.
The risk of galvanic corrosion had been anticipated and so the two metals were separated by asbestos and shellac insulation. (Shellac is a natural resin that was widely used in the Victorian period as a lacquer or protective coating.) However, the shellac had degraded, and acidic rainwater had soaked the insulation, providing electrolytic conduction between the metals. The corrosion of the iron framework (Figure 19) was so extensive that there was concern it might collapse, and so in 1986 the statue was renovated. The wrought-iron framework was replaced by stainless steel, which will not corrode in the presence of copper, coated in a layer of PTFE insulation.
Highly localised attack, such as that found on the Statue, is also known as crevice corrosion, because attack is concentrated in the contact zone at the junction of the two dissimilar metals. A crevice forms and further attack occurs there, making the hole deeper (Figure 20). It is a common feature of corrosion, and can be contrasted with general overall attack. It is that much more dangerous since the damage is usually hidden from external inspection, until the strength of a product is reduced to a critical level and it fractures through the crevice. The loss of material lowers the section area, and there may also be a stress concentration within the crevice to magnify the stress further. Where the load levels are low, as in a galvanised water tank, nothing will happen until it leaks and alerts the owner to the problem. However, where a pressurised tank such as a boiler suffers the same problem, the effects may be much more dramatic.
Suggest why the rate of corrosion was lower on the Statue of Liberty than on HMS Alarm.
The Statue had corroded seriously in 100 years while HMS Alarm had corroded in only two years. The ship was subjected to continuous immersion in sea water, a good conductor owing to its high salt content, while the statue was subjected to only intermittent rain-water percolation through leaks in the outer copper skin. On the Alarm, the iron nails had a very small area, which meant that they corroded very quickly.
Explain the following observations of corrosion in terms of the electrochemical series:
A similar concept to the electrochemical series that has been used by engineers for many years is the galvanic series (one example of which is shown in Table 2: here the list should be read down the columns rather than across the rows). It ranks metals and alloys in order of reactivity or electrical potential, just like the electrochemical series. It also has the same properties: the greater the difference in position between two metals or alloys, the greater the likelihood that corrosion will occur. The series differs from the electrochemical series in showing alloys, which are of course of direct practical interest. Closely related alloys such as the brasses and bronzes are grouped together. Again, the most reactive materials are towards the bottom of the list.
However, such lists must be used with caution because they are highly dependent on the actual conditions. Also, the numerical values associated with the electrochemical series can allow more accurate information to be gathered about likely corrosion rates.
|Titanium||‘Active’ stainless steel (unstable oxide film)|
|‘Passive’ stainless steel (stable oxide film)||Cast iron and ‘mild’ steel|
|Bronze (Cu—Sn)||Magnesium and magnesium alloys|
Using first the galvanic series, then checking with the electrochemical series, suggest which pair of alloys below will show the greater tendency to corrode in a marine environment, if the exposed areas of the two components are roughly equal:
Reading from Table 2, the separation of steel and bronze is greater than that of steel and magnesium, so one might suggest that corrosion would be greatest for the steel/bronze couple.
However, looking at electrode potentials (Table 1), the standard E0 values are:
Mg/Mg2+: E0 = −2.37 V
Fe/Fe2+: E0 = −0.44 V
Cu/Cu2+: E0 = +0.34 V
The difference between magnesium and iron is much larger than that between iron and copper, so the electrochemical series contradicts the galvanic series in this case. The electrochemical series is a more accurate predictor of corrosion behaviour than the galvanic series alone.
If a stress exists in a product exposed to a corrosive environment, the rate of corrosion can then increase and be extremely localised, such as at the tip of a growing crack. Furthermore, some specific chemicals are so aggressive that corrosion will occur at relatively low stress levels, such as those created during manufacture, for example. The residual stress in a component can then be enough to trigger crack growth and failure.
On 9 May 1985 the roof of a swimming pool at Uster near Zurich collapsed, killing 12 and injuring several others.
The concrete roof had been held up by a set of stainless-steel tie bars, which were found after the accident to have cracked transversely (Figure 21). Chlorine is added at quite high levels to swimming-pool water supplies in order to control bacterial contamination from swimmers. It is a very powerful oxidising agent, and can attack a very wide range of materials, typically by forming hairline cracks in components that are in tension. Traces of chlorine gas in the general atmosphere of the building were found to be the cause, having attacked the material chemically. Stainless steel and many other types of steel are known to be susceptible to chlorine attack, as are certain plastic materials.
After that tragedy, yet further failures occurred in swimming pools. In the Netherlands, for example, a ceiling collapsed during the night of 8/9 June 2001 at a pool in Stenwijk. It was discovered the next morning by a party of visiting swimmers; fortunately there were no casualties. The pool had been open only nine years, and chlorine was again used to disinfect the water of the pool. The design of this swimming pool was slightly different, however. The ceiling above the pool was reasonably well supported, but there were air ducts above weighing several hundred kilograms. These were attached to the outer roof above by stainless-steel threaded bars. Upon examination, the threads were found to be deeply cracked. When the cracks reached a critical size, the air ducts fell onto the ceiling, which in turn failed owing to the extra load and the shock, or impact, loading.
Stainless steel is used extensively in swimming-pool fixtures such as handrails, ladders leading from the pool and diving boards. It has been found to perform well, without cracking, for many years, even with constant immersion in warm water dosed with low levels of chlorine. The metal has a polished surface and so can be cleaned easily, and would normally have a long and uneventful life. But the stresses applied in these fixtures are only intermittent, not continuous. Except perhaps for diving boards, the fittings are over-designed, so the imposed loads from the users are relatively low.
However, when similar grades of stainless steel are used as tie bars, they are under continuous tension, which is when stress corrosion cracking becomes a relevant failure mode. The air circulating in the ducts above the Stenwijk pool will have been saturated with water vapour as well as carrying traces of free chlorine gas, and will have attacked the protective oxide film on the steel. This attack will have produced brittle cracks, which then grew slowly under the tensile load, with new metal being exposed as the crack advanced. The high stress concentration at the advancing crack tip, together with the imposed tensile load, will have encouraged further crack growth until criticality.
Falling objects will impose much greater loads than their nominal weight owing to the momentum gained by the drop. The exact force of the impact will depend on the distance through which they fall.
Explain why the cracks were initiated at the roots of the threads.
The root of the thread of any screwed joint represents a stress concentrator, where the applied stress can be magnified many times at the corner. The exact value of Kt will depend on the radius of curvature at the root, sharper roots being more severe than shallow roots.
Even low tensile loads will have been enough to stimulate crack initiation, owing to the importance of stress concentrations in magnifying the imposed load. The critical threshold for the initiation of SCC is normally very low, but must be a tensile stress. The other interesting feature of these failures is the continued attack by very low levels of gaseous chlorine. The chlorine levels are lower than in the pool water, so the extent of the attack probably relates to the much greater ease with which gas molecules can penetrate a corrosion film.
The incident in the Netherlands led to a survey of several thousand swimming pools in the country. It emerged that 14 pools were at immediate risk of sudden failure owing to deeply cracked suspension bolts of a similar design to those at the original pool. They were closed immediately for repair. A further 18 pools were considered dangerous and needed extra support of the ducting. Most of the pools were very new, one being only a year old. The stainless-steel parts were replaced with galvanized steel equivalents or with stainless steel with a high molybdenum content (~6%), both of which are much more resistant to SCC.
The accident led to other surveys in Germany and the UK, where a warning notice was circulated by the Health and Safety Executive. It highlights the dangers arising from using what appear to be corrosion-resistant materials in situations where brittle cracks can develop quickly in safety-critical components.
Stainless steel is also sensitive to chloride ions (such as those present in sea water and brine), and especial care is needed in designing the material for use in ships and boats where exposure can occur.
Stainless steel is not the only metal to fall victim to SCC. One of the first discoveries of SCC occurred in India in the early part of the nineteenth century, when that country was still part of the British Empire. There was a large standing army that was always in need of live ammunition. The brass cartridge cases would occasionally split, and often at the worst possible time (when being fired), frequently causing injury to the marksman.
So what caused such failures? The two factors needed for stress corrosion cracks are, first, a tensile stress in the outer layers of the brass and, second, an active chemical that will attack brass or copper. The stress could be caused by the manufacturing forces used to shape the cartridges, since the cases were made from cold-deformed brass (70% copper, 30% zinc). The process involved successive stages of deformation of flat discs punched out from 3.25 mm thick sheet (Figure 22). After each stage, the product was annealed in order to recrystallise the metal, and pickled with sulphuric acid to remove oxide at the surface. The annealing process was intended to relieve residual stresses set up in the cases, but the process was not always successful in completely removing these stresses.
After some detective work, an association was seen between the rate of cracking and the season of the year. Cracking tended to occur during the monsoon season when humidity and temperatures were high, rather than during the cooler months. Yet although the rate of most chemical reactions increases with temperature, controlled experiments showed that this could not have been the only cause of the problem. Then the Woolwich Arsenal undertook a series of trials with many different chemicals. They exposed bent strips of brass to the chemicals and observed the metal surfaces at the most highly stressed zones. They found that ammonia gas and water vapour were, in combination, the two most potent agents needed to initiate brittle cracks. Bearing in mind the experience of stainless steel in chlorine-doped water, it is interesting that failure times for many samples dipped into ammonia solutions were longer than for exposure to ammonia gas and water vapour.
The mystery was therefore solved, because it was realised that ammonia is produced by manure and dung, so would have been present in the stables of the army horses, for example. If ammunition had been stored near the stables, it is most likely that trace amounts of ammonia in humid air could have cracked the brass cases extremely quickly. Hairline or microscopic cracks would have been formed, and then grown to a critical size by the time the ammunition was needed.
So why does cracking or highly localised attack occur in such a case, rather than general corrosion? The active agents attack at stress raisers, at the upper edge where the case makes contact with the bullet (Figure 23). The formation of a galvanic cell is unlikely, because a thin film of water on the surface is insufficient to provide the electrolyte. However, the final stage of manufacture, when the bullet is put in the explosive-filled case, will put the lip under a radial or hoop stress. The edge is unlikely to be totally level, and small degrees of roughness there will be attacked by the ammonia. Once a crack has formed, it will grow under the influence of the hoop stress, with the corrosive solution seeping away to leave a fresh crack tip ready for further attack.
The problem of chemical attack on brass and other copper alloys is not uncommon, as the example described in Box 3: Pump failures demonstrates.
Brittle ceramic products are frequently impregnated with softer and tougher materials to strengthen them. Ceramics usually have an open pore structure, and filling the pores with a crack-resistant material toughens the final product. Such a process is used to improve the toughness of anodes used for the electrolytic production of aluminium, and involves applying a vacuum to a chamber in which the anodes are placed. Liquid pitch is then pumped into the chamber to fill the pores, before the anodes are removed for baking so as to solidify the pitch.
After six months’ operation, the vacuum in one such chamber deteriorated and investigation pointed to failure of an impellor used to apply the vacuum. The impellor was made from brass and had suffered severe corrosion, with the formation of a green deposit over all the surfaces (Figure 24). The impellor was replaced, but the vacuum again began to deteriorate. The time had come to perform a serious investigation, especially as the impellors were rather expensive.
Suspicion fell on the liquid pitch, as it had a high sulphur content, but an alternative explanation quickly became apparent. Several operators had smelled ammonia in the pitch, but it took an alert manager to recognise the cause of the corrosion. As an OU student studying forensic engineering, he correctly identified that the ammonia had attacked the copper component of the brass to form cuprammonium salts, attack being most severe at the tips of the flight vanes and corners of the design where the local stresses were highest.
At these points, stress corrosion cracking had further weakened the impellor. The net result was loss of material leading to loss of evacuation power, with a lower vacuum for impregnation.
So what was the solution to the problem? One possibility was to replace the rotor with stainless steel, but this was an expensive option. A lower-cost solution would be to apply a resistant coating to the surface of existing brass impellors, which would prevent the ammonia contacting the brass surfaces. Several different polymer coatings could be used, such as sintering powdered PVC or polypropylene onto the part, but the final solution involved using epoxy resin. The pumps have, since this innovation, proved entirely trouble-free.
Stress corrosion cracks can also build up in other structures. These were a particular problem in locomotive boilers in the early days of the railways in Britain. All such boilers were made from wrought-iron sheet, riveted together to form a cylinder. In the earliest engines, the boilers were constructed using a single line of rivets, thus forming two corners, one inside and the other outside (Figure 25). Initially they apparently performed well, but a number of catastrophic explosions were experienced through the 1840s.
Some of the earliest explosions were caused by failures of the safety valves fitted to the boilers. For example, some of the first safety valves were simply a stopcock weighed down by a steelyard. The weight and its distance along the arm controlled the pressure at which steam would activate the cock, and so blow off harmlessly. It was tempting for engine drivers to increase boiler pressure by adjusting to the highest possible pressure – and if that didn't give enough driving power to the wheels, they would wedge the valve down further. Such manipulation of a safety device was asking for trouble. Another problem that also caused some explosions was that if the water level dropped too far, the structure would overheat and fail.
Many such boiler explosions were investigated by the Railway Inspectors appointed by the Board of Trade (one of the predecessors of the Department of Trade and Industry). They found that they could not always explain why the explosions had occurred, having excluded both human negligence in the use of the safety valves and the water level. When the inspectors examined the failed remains in some detail, they found a pattern revealing that failure almost always occurred from the horizontal line of rivets in the boiler, and there appeared to be a deep groove running alongside the joint that was filled with rust.
Explain why failure tended to occur along the longitudinal axis of the boiler.
Why would a riveted joint represent a line of weakness? How could such a joint be strengthened?
There are two stress components in the wall of a cylindrical boiler, the hoop stress and the longitudinal stress. The hoop stress is twice the longitudinal stress and is given by the equation:
where p is the internal pressure, r the radius of the cylinder and t the wall thickness.The hoop stress acts such that the cylinder will fail by a lengthways crack (rather than a radial crack).
A riveted joint will always be weaker than continuous material, simply because it is a break in the uniformity of the wall. The rivet holes themselves are stress concentrators, and a line of rivets is a line of such defects, so the line of rivets is the weakest part of the structure.
The problem could be tackled by replacing the lap joint with a butt joint (Figure 26), reinforced by extra layers of riveted metal.
The root cause of the problem lay in corrosive attack of the wall to one side of the joint, owing to the nature of the joint itself. The wall on either side would, when the pressure was being raised first thing in the morning, experience higher stress than the double wall thickness at the joint itself, owing to the existence of a corner where the plates met acting as a stress raiser. Corrosion of the iron by the boiler water tended to start here in the form of a slowly growing crack, a process repeated every morning the locomotive was worked. The cycling of hoop stress caused whenever steam was raised and the boiler pressurised could also have given rise to fatigue cracking, although the phenomenon was not recognised as such at the time.
The purity of the boiler water would be important in such a case, because if any dissolved salts were present they would increase its electrical conductivity and hence the likelihood of corrosion cells being set up near the joint.
Such ‘groove cracking’ was a design fault, which could be corrected only by developing a double-riveted butt joint (as shown in the answer to SAQ 2) in place of the lap joint, rigorous inspection and maintenance of locomotives and, ultimately, use of stronger steel in place of the wrought iron.
Stress corrosion cracking can produce devastating damage in large structures, as the examples of swimming-pool ceilings and roofs in Section 2 showed. But even larger structures can also be attacked, as was revealed by the events at the Silver Bridge in 1967.
The Silver Bridge spanned the Ohio river between West Virginia and Ohio. It was a long bridge (680 m), owing to the breadth of the river at Point Pleasant, the small settlement on the east bank where one end was sited. It was nearly 12 metres wide in total, carrying a concrete road two lanes in width as well as a footpath. It had been built in 1928 alongside an older railway bridge, but to a strikingly different design (Figure 27).
It was effectively a suspension bridge, but one that used eye bars and rigid hangers to support the deck, rather than the familiar steel cables of modern suspension bridges. The main chain system was not unlike a bicycle chain, each structural unit being a set of steel links held together laterally on a pin (Figure 28). Such a design had been used previously on the Brighton chain pier, Telford's Menai Straits bridge and Brunel's Clifton bridge in the Victorian period, although in those cases the links were generally made of three or more bars, rather than just two. The Clifton bridge possessed no fewer than six such bars strapped together.
There had previously been problems with suspension bridge roadways that were too flexible, and several bridges had been severely damaged, if not destroyed, by lateral winds: see Box 4: Historic suspension bridge failures.
Failures of suspension bridges (see Table 3) have most frequently involved failure of the suspended deck rather than of the support structure. A famous early failure occurred in the Brighton chain pier in 1836, when a storm destroyed the deck (Figure 29). Built in 1822, the pier was 352 m long and 3.9 m wide. Five cast-iron towers, spaced 78 m apart, supported the decks. The pier, which served for embarkation of new ferry services to France after the Napoleonic wars, was damaged many times by storms and then rebuilt, until it was finally demolished in 1896.
|Bridge||Location||Year of failure||Main span/m||Width of deck/m||Designer|
|Dryburgh Abbey||Scotland||1818||79||1.2||J. and W. Smith|
|Nassau||Germany||1834||75||—||Lossen and Wolf|
|Brighton Chain Pier||England||1836||78||3.9||Samuel Brown|
|Menai Straits||Wales||1839||177||7.3||Thomas Telford|
|Roche-Bernard||Scotland||1852||79||1.2||J. and W. Smith|
|Tacoma Narrows||USA||1940||854||11.9||Leon Moisseiff|
Perhaps more famous was the bridge built by Thomas Telford over the Menai Straits in 1826 (the deck of which was also damaged severely on several occasions, especially in 1839). The Angers disaster in France was the most serious failure, since 226 soldiers were killed when the bridge fell. Corrosion of the anchors of the main cables was one cause of the disaster.
The Clifton suspension bridge was built to a design by Isambard Kingdom Brunel, but was not completed until 1864, after his death. Despite having been damaged in the past, the Menai Straits and Clifton bridges still stand today. Both use many eye-bar chains to support the deck, giving high redundancy if one fails. However, many other suspension bridges failed as designers produced longer and longer decks, culminating in the famous collapse of the bridge over the Tacoma Narrows.
The Tacoma bridge was completely wrecked in 1940 (although the towers and main cables remained intact). The failure occurred through dynamic stimulus of the very long deck by a steady wind of approximately 40 mph blowing at right angles to the axis of the deck. But rather than oscillating from side to side, it started rolling up and down as it resonated. The amplitude rose steadily until failure of the deck occurred, and then other parts followed (Figure 30). The accident was the culmination of a sequence of similar incidents in which sub-critical oscillations occurred (which gave the bridge the nickname ‘Galloping Gertie’). The failure highlighted the problem of the effect of winds on very large structures. Large buildings, such as skyscrapers, routinely have giant pistons fitted within to damp movements caused by high winds.
In the Silver Bridge there was a small clearance of about 3 mm between the pins and eye-bar holes, to allow easy fitting of the parts together when on site. The links were made from a high-strength steel that had been developed in the 1920s by the American Bridge Company. The eye bars had been cast to shape and then heat treated to develop the strength of the steel; then the holes were drilled out.
The roadway of the bridge was trussed with steel girders to improve its rigidity. There were two separate towers supporting the main suspension chains, each 40 m high; each tower consisted of four tiers that were cross-braced except for the road gap (Figure 31). The towers were supported on two massive masonry piers anchored in the river bed, and the two sides supported by two further piers, one in the river bed and the other on land on the Ohio side of the river.
The main suspension chains were anchored in concrete troughs on each bank, each trough being pinned by 405 concrete piles to resist the tension exerted by the main chains. Each of the two towers was designed to move with temperature fluctuations in the chains by having curved rocker joints at its base on the masonry piers.
The bridge was inspected in 1954 and remedial work on decaying concrete recommended, together with painting of the metalwork. It was inspected again (after the remedial work had been finished) in 1955, 1961 and 1965.
Describe the types of load carried by:
Assuming that there was no traffic on the bridge, and no other external loading, where in the bridge were the eye-bar joints likely to be under the greatest load?
The top eye-bar joints should be under greatest tensile load because they support the complete length of chain below. The load diminishes at successively lower eye-bar joints.
The bridge appeared to fulfil its function well, despite the loading on the bridge increasing steadily after it was built with the increase in car population and road traffic. The bridge had been built at a time when the Model T Ford was the most popular automobile on the roads, but by the 1960s cars were significantly larger and heavier, further increasing the load on the bridge.
Corrosion of the steel structure of the Silver Bridge was likely, especially as the acidity of the rain was enhanced locally by industry nearby. So the entire structure was painted with an aluminium-based paint, hence the name ‘silver’ bridge. An example of how corrosion can lead to failure of a bridge structure is described in Box 5: Fall of the Kinzua Viaduct.
Many historic bridges and viaducts show the effects of long-term corrosion, which can often lead to catastrophic failure. Such a failure happened to the Kinzua Viaduct during the passage of a large storm through western Pennsylvania in July 2003. The Kinzua Viaduct had been built from cast and wrought iron by the New York, Lake Erie and Western Railroad Company in 1882 to deliver coal from local coalfields to the Great Lakes. It was reconstructed in steel in 1900 (Figure 32) to allow for the greater loads on the structure.
The viaduct was retired from active service in 1959, but the structure remained as the centrepiece of a national park, taking occasional traffic from sightseeing tours. However, the bridge required extensive renovation owing to rusting and was closed again to traffic. Renovation was started in February 2003, but on 21 July 2003 an unusually severe storm hit the region. During the storm, 11 of the viaduct's 20 towers fell, destroying the landmark (Figure 33).
Forensic investigation of the fall showed that the centre of the structure had oscillated from side to side as two tornados (with wind speeds of about 100 mph) struck the site. It was estimated that about four complete cycles occurred before its collapse.
Final failure had occurred at the rusted base bolts holding the structure to the foundations, which had fatigue cracked over a long period of time (Figure 34), assisted by internal rusting.
The lattice superstructure fell in several parts, the railway line ending up hundreds of feet away. The failure bears some similarity to the classic fall of the Tay Bridge in a severe storm in 1879, the main difference lying in the way the joints in the structure behaved. Most of the cast-iron joints of the Tay Bridge failed either before or during the fall, so the railway line ended up close to the foundations. The steel joints at Kinzua remained mostly intact.
The Kinzua bridge collapse is covered in the ‘Kinzua's weakest link’ video link below.
Kinzua's weakest link
The 39-year-old Silver Bridge collapsed suddenly at about 5 p.m. on 15 December 1967 when the roadway was filled with rush-hour traffic – 37 vehicles were trapped on the roadway.
The first signs of collapse were later recounted by the survivors. Many occupants of the cars on the bridge had felt it ‘quivering’ before it fell. Most witnesses had then heard ‘cracking’ or ‘popping’ noises, some saying that it sounded like a ‘shotgun blast’. After this, the bridge started disintegrating fast; girders and hangers fell, followed by collapse of the roadway itself near the centre of the bridge. The towers then fell, bringing the rest of the chains with them. The entire structure collapsed within about a minute, disgorging 31 of the 37 vehicles into the river below. Witnesses on the banks described the bridge falling like ‘a house of cards’, but many tried to save those who had escaped from their vehicles. Those who were trapped inside their sinking vehicles had little chance of escape, however, given that the river reached a depth of 20 m near the centre. Some broke their vehicle windows and managed to escape their sinking or sunken cars, swimming to the surface. The fall from the road deck and impact with the water rendered many of the victims unconscious and they drowned, trapped in their sunken vehicles. The temperature was about −1 °C, and the cold water of the river meant that anyone who survived the fall itself succumbed quickly to hypothermia. Despite heroic rescue attempts from both sides of the river, the disaster claimed 46 victims, although remarkably, three people from the centre section survived.
Recovery of the bodies took some time, and they were the first priority after all the swimming or stranded victims were rescued. However, it was vital to determine the cause of the accident, so the river bed was trawled thoroughly for all the metalwork that had fallen. Since virtually the whole bridge had disappeared (apart from the road deck on the West Virginia bank), this was a big job involving many weeks’ work. It was difficult work as well, because the river was deep and fast-flowing at that point, as well as being very cold.
Although there had been many other bridge collapses in the USA before 1967, only one had been worse: the collapse of a railway truss bridge in 1876 at Ashtabula, in which over 100 died. For such a total collapse to have occurred in 1967 seemed unthinkable, given the progress in analytical design, the greater understanding of loading and the improvement in construction materials that had occurred since the 1920s.
A thorough and intensive investigation was needed to establish just what had happened to cause such a catastrophic failure. Several US government agencies were involved, including the National Transportation Safety Board (NTSB) and the National Standards Bureau (NSB), as well as the Battelle Memorial Institute and several university engineering departments.
The investigation took three years to complete, although critical evidence emerged within weeks of the accident.
Some possibilities could be ruled out immediately. For example, there were rumours of supernatural forces at work that night, but very little solid evidence of the ‘Mothman’ emerged, either there or anywhere else. The Mothman was a demon purportedly haunting the bridge, which has supposedly appeared as a portent of similar disasters around the world. Such stories would have encouraged the investigators to speed their work so as to reassure the public that there were more rational causes of the bridge's fall.
It was important to establish the precise sequence of events leading up to and during the collapse. From which part had the collapse started? Why had so much of the structure been destroyed? Was there any prior warning of the failure? What part had the weather conditions at the time played?
Eyewitnesses were plentiful, and each had a different perspective of the bridge as it fell. There were some common parts to their statements. Most of the witnesses, especially survivors from vehicles on the bridge at the time, testified that they heard cracking sounds very early in the collapse. This suggested that brittle fracture was an important failure mode, if not initiating the fall then certainly playing a crucial role in the disaster.
Testimony showed that the bridge vibrated during high lateral winds, and also when traffic crossed, which meant that movement in the eye-bar joints would have been occurring quite regularly. There was generally significant traffic across the bridge since it carried an Interstate highway.
Witness statements attested to unusual vibrations in the roadway just before the collapse. They also noticed that a bolt or cap-like object was seen on the roadway prior to or during the initial period of the collapse.
A plan was needed to determine the chain of events leading up to and during the collapse. That sequence would necessarily depend on which parts had broken first, and a fault tree would enable a plan of action in isolating the cause (or causes) of the disaster. Such a systematic approach is known as fault-tree analysis or FTA, and is part of the armoury of methods used by accident investigators. With large-scale and devastating accidents, all possibilities, however remote, need evaluation in the light of all the available evidence. In this way, the list can be whittled down to the vital one or small handful of most probable causes. Such a systematic approach is vital where both the material and the witness evidence is extensive, not just for the analysis of bridge or building failures where destruction is almost complete, but also for marine and aerospace disasters.
One action that was taken almost immediately was to close a bridge of very similar design some miles upstream, at St Mary's. That bridge would not just be subject to rigorous inspection but would also become the basis for experimental work on its dynamic behaviour when loaded under controlled conditions.
The possibility of wind action could also be ruled out, because the wind at the time of the accident was parallel to the long axis of the bridge, and was only about 6 mph. Likewise, there was no evidence that the masonry piers were involved. Indeed, they survived almost unscathed (Figure 35).
As the wreckage was pulled from the river it was examined and identified, and any failures of the metal components were recognised and tagged. This was a mammoth task, given that virtually the whole bridge had fallen into the water, including all the road decks, trusses, chains and hangers, eye bars and the two towers. The parts were then reassembled and all the failed or fractured components photographed and catalogued. Over 90 per cent of the bridge components were collected together and reorganised into their original positions in the bridge. Their position on the river bed before extraction was also an important facet of the investigation. To help reconstruct the sequence of events, the metal parts were classified into different categories:
The fractured parts were then examined in detail to identify the extent of plasticity, exposure of fracture surfaces, and their chemical state.
Distinguishing the different kinds of damage to the many different structures in the bridge was time-consuming, and often difficult. Metal surfaces exposed by fracture would have rusted both in the river and later, when exposed to the atmosphere during storage and reassembly. A selection of critical failed parts needed to be identified for shipping to the many labs involved in detailed analysis of the components.
Brittle fractures were discovered quickly in the mass of debris hauled from the river. Such samples became the focus of increasing effort as time went by, simply because they were unexpected. So the possible failure mechanisms were immediately narrowed down when brittle fractures of critical components started to emerge from the river.
Suggest what mechanisms could cause brittle fracture of steel components that might normally be expected to fail in a ductile manner.
There are several possible mechanisms that can cause brittle behaviour in any nominally ductile material such as steel:
One particular broken part was recognised quickly as part of an eye bar. There were 146 eye bars in the original bridge, and they were safety-critical because if broken the main chains could be threatened. The eye bar was identified as being from the top joint in the hanging chain nearest the bank and next to the Ohio tower, one of the two lower bars on the outside of the bridge facing north, upriver (Figure 36). It was assigned the identity number 330. It was 17 m long, 51 mm thick with a shank width of 305 mm. The hole in the end, designed to fit over the pin, was 368 mm in diameter, and the width of each limb beside the hole was about 203 mm (Figure 37).
One half of the eye at the joint is shown in Figure 38(a), and it shows two breaks in the limbs either side of the pin-hole. Although both appear brittle in this picture, in fact one side showed signs of ductile deformation. The way it had fractured was unique when compared with the other eye bars collected. The missing part of the eye bar was located and examined (Figure 38b). It corresponded well to the main part, although it had been damaged in one corner – presumably when it fell off the pin and impacted with the deck or another part of the bridge, which must have been still standing at that point in time. This second part shows more clearly the ductile portion on the left, where the limb has broken with a large lip projecting from one side of the component. This surface is seen in oblique view in Figure 38(c), a view that also shows a secondary crack or branch away from the main path of the crack. The surface on this ductile part of the eye bar was much coarser than the brittle fracture side. A thin layer of rust covered all the surfaces, as would be expected from their immersion in the river for several days.
Suggest why it was important to find the missing portion of the eye bar.
It is always best if a corresponding part to a fracture surface is examined, because it can corroborate features present on the half of the surface already found. It is especially important where subsequent damage such as corrosion has occurred. If eye bar 330 had fractured at an early stage in the disaster, it would be vital to determine the cause of the brittle fracture.
When the thin coating of recent red-brown rust was removed gently in the laboratory, the original state of the surface on the lower part of the eye-bar hole, the part showing the brittle fracture, was revealed. Citric acid, present in citrus fruits like lemons, was used to remove the rust. It is a very weak acid, and so its dissolution of red-brown rust is slow. This allowed more control of the cleaning process, minimising damage to the underlying surface. The overall fracture surface showed very little sign of ductility, except for a small shear lip along a short length of the outer edge of the fracture.
Part of the fracture surface is shown in Figure 39. It was noticed that one corner of the inner side of the fracture, i.e. the side next to the pin, showed two curved features of different colour and texture from the rest of the fracture. These zones were very small, measuring only 1.5 mm and 3 mm in diameter respectively; the origin of the larger zone is shown in the figure. They were dark grey, almost black, a tone probably representing Fe3O4, the iron oxide formed in low concentrations of oxygen or air: see Box 6: Rusting. Remnants of the recent red-brown rust were visible in pockets on the rest of the surface. The lines on the curved features pointed back to the inner surface of the eye bar. It was feasible to suggest that the two zones represented brittle cracks present before the final failure that reached a critical size just before the catastrophe.
The reactions of iron and water include several end products, depending on the presence or absence of air, the temperature and the concentrations of salts in solution. The chemical reactions here are for illustration: you don't have to remember them.
The most common product is red-brown rust, formed by the reaction:
4Fe + 3O2 → 2Fe2O3
Note that in the Fe2O3 produced, the ratio of iron to oxygen is 1:1.5. The volume of red-brown rust is about 50 per cent greater than that of the metal, and so can enhance crack growth.
Hydration of the oxide is usual in the presence of water:
Fe2O3 + 2H2O → Fe2O3.2H2O
The hydrated oxide is a very weakly protective film because it tends to spall away from the underlying surface in lamellar flakes, exposing a fresh surface to further attack. The volume change associated with producing the hydrated oxide is larger than for the oxide itself owing to the water molecules in the atomic structure.
The reaction of iron with water can also form hydroxides, producing hydrogen gas:
Fe + 2H2O → Fe (OH)2 + H2
2Fe + 6H2O → 2Fe (OH)3 + 3H2
The hydrogen gas may represent a danger if the reactions occur in an enclosed environment, such as a steel tank, for example. Many welders have been injured and killed by explosions when the welding torch penetrates to the interior: the hydrogen is released to mix with air and then explodes.
If the concentration of oxygen is low, then different oxides are formed:
2Fe + O2 → 2FeO
where the ratio of iron to oxygen is 1:1, and:
FeO + Fe2O3 → Fe3O4
where the ratio of iron to oxygen is 1:1.3.
Both products are black and form preferentially at high temperatures, such as during forging of hot metal, when they are known as ‘black scale’. They are usually removed by treatment with sulphuric acid in large-scale manufacture, a process known as pickling. Black oxide is also formed in central heating systems, since the system is closed to the outer air and oxygen is depleted in the closed water supply by reaction. Hydrogen gas accumulates at the top of the system, and is liberated when the system is bled.
To explore the problem further, the eye bar was examined for signs of further cracks. The mechanism that caused the critical crack was probably at work at other points on the inner surface of the eye bar, so could be tested by several techniques.
Many such sub-critical cracks were found (Figure 40), showing that there was a single mechanism at work. The interior of many of the cracks was filled with iron oxides, often present in a lamellar form showing successive and intermittent phases of formation. An adjacent eye bar on the next joint down along the chain was also found to be cracked in a similar way at roughly the same point.
(Hint: bear in mind that a hole in a component represents a stress concentration factor of about three.)
The other eye bar of the same joint was located, and showed damage to the hole consistent with having been pulled off the pin. A large burr existed on one side of the hole only, showing that the end had been subjected to a large force in the accident.
Many sections were taken of the steel near the fracture to examine its microstructure, and were compared with different parts of the same eye bar as well as with other eye bars. The sections showed a steel core surrounded by a zone that could be identified as being of higher strength due to the presence of martensite.
Martensite is a strong, hard phase of steel usually formed by rapid quenching from a high temperature.
XPS, X-ray photoelectron spectroscopy, gives information about the elements on the surface of a material. It does this by analysis of the X-ray spectrum emitted by the surface when impinged by the electron beam.
The fracture surface was also analysed by XPS for trace elements that might give a clue to the corrosion processes at work over the 39-year lifetime of the bridge. In addition to small traces of manganese present in the original metal, the researchers found significant traces of sulphur present within the cracks; this is an element not present in the metal itself, indicating an unknown, external source. The sulphur concentration was greatest at the mouth of the crack. The steel had a carbon content of 0.6%, slightly higher than the normal content of mild steel (which is up to about 0.3%).
It was possible that the tiny cracks present on the inner surface of eye bar 330 initiated the collapse by causing brittle fracture. It therefore became important to determine the strength of the steel and, also, its fracture toughness. Steel from eye bar 330 was tested, as well as from other eye bars from the Silver Bridge. Hardness tests across a section though an eye bar showed a soft outer zone, followed by a harder zone and then a softer core (Figure 41). The hard zone extended from about 2.5 mm to 9 mm inside the section. This represented the hardened zone produced by quenching the steel during manufacture. The outer layer of the bar showed loss of carbon due to the heat treatment during manufacture.
Charpy impact tests at several laboratories showed the toughness to fall with lowering temperature, with a low value at or near the freezing point of 0 °C (Figure 42), a temperature close to that experienced at the bridge at the time of the accident.
Small samples cut from eye-bar material were also tested in simple tension at 25 °C, and gave the following results:
yield strength of outer layers = 590 MPa
tensile strength of outer layers = 835 MPa
yield strength of inner layers = 490 MPa
tensile strength of inner layers = 810 MPa.
All the samples showed high ductility with a reduction in cross-sectional area of nearly 50%.
The investigators wanted to know about the fatigue properties of the component, to find a feasible explanation of why it took 39 years for the eye bar to break. They needed information on the several stress corrosion mechanisms that were possible in the material, including hydrogen embrittlement, the effects of sulphur compounds such as H2S (hydrogen sulphide) and the effects of moisture and salt. Notched eye-bar material was loaded to failure in various environments.
In fact, no evidence emerged for hydrogen embrittlement, and a wet environment in the laboratory tests had no effect on the rate of crack propagation. However, the life of the steel samples was reduced substantially by hydrogen sulphide, a conclusion that appeared to correlate well with the detection of sulphur in the critical crack (and other, sub-critical cracks).
Although it is known that a round hole in a flat sample will theoretically produce a stress concentration of about 3, the issue was decided experimentally. A tensile test at 25 °C was undertaken on an intact eye-bar-pin assembly from the bridge, being some 8 m long and from a lower part of the chain. It yielded at about 7 MN, and fractured in the shank at a stress of about 770 MPa. The yield stress in the shank was about 520 MPa, and the failed eye bar showed ductile behaviour with a reduction in area of 30% at an elongation of 8.5%. By putting strain gauges at various points in the hole of the bar, the stress concentration was calculated to be about 2.62 at the opposing faces of the inner side of the hole where fracture had occurred in eye bar 330.
Suggest why, in the tensile tests, fracture occurred in the shank rather than at the hole of the eye bar. What factors contribute to where failure occurs?
The shank dimensions are 51 mm by 305 mm, giving a section area of 15.6 × 10−3 m2, while the limbs at each side of the hole are each 203 by 51 mm, giving a total section of 20.7 × 10−3 m2. So the section area is about a third greater in the limb compared with the shank, giving a correspondingly lower stress.
Nevertheless, the greater stress concentration at the edges of the hole should have ensured failure here rather than in the shank.
Surface roughness effects can be critical, so if the pin and eye-bar hole surfaces were smooth and the shank surface was rough, failure in the shank would be preferred.
It is worth emphasising that stress concentrations are of less importance in ductile compared with brittle-type failures. After all, a material that usually fails in a ductile manner, such as steel, can yield locally at the root of a notch or the edge of a hole. By contrast, during brittle fracture there is no mechanism for absorbing excessive load by deforming plastically, and the stress at the root of a crack may be extremely high.
The experiment established that brittle cracks had not developed in the lower eye bar: if they had, this test bar would have failed at the hole rather than in the shank.
An additional possibility was considered. It was known that there was significant movement of the bridge during passage of traffic, because users had noticed it many times when crossing. The joints would thus have been subjected to rotary motion around the pin in order to accommodate such vibrations. Could these have caused fatigue crack growth at the bearing surfaces?
Contact between a circular and a flat plate creates so-called Hertzian stresses at the contact zone: compressive at the centre, and surrounded by a tensile zone. A similar effect will occur at a circular pin joint, provided there is some clearance between the two parts. In addition, there could be considerable wear caused by corrosion. Rusting would create particles of Fe2O3.2H2O, which, being harder than the steel, would act as an abrasive powder as the surfaces moved against one another. The fact that the rust particles had a larger volume than the metal they replaced would also stimulate wear. The inner surfaces of the eye-bar holes showed deep grooves (Figure 40a), indicating fretting action. Could fretting have initiated critical cracks?
To test this hypothesis, pin and collar shapes were machined from eye bar 330 (away from the region of the actual failure), fitted together and then rotated so that the pin acted against the collar. The results are shown in Figure 43. Even with the effects of fretting, the material around the eye still showed a higher fatigue life than the material in the shank.
Suggest how fretting fatigue could occur at a pin joint in the main chains of the Silver Bridge. Indicate the most likely place for such a problem, and compare the actual position of the critical and sub-critical cracks on eye bar 330, drawing any appropriate conclusions.
Fretting wear occurs owing to repeated cyclical movement at a joint and was caused in the Silver Bridge pin joints by corrosion producing particles of Fe2O3.2H2O that were harder than the underlying steel, and of greater volume. The action will have been most severe at the upper joints on the main chains, where the loads were largest. Tension cracks through fatigue could have formed at either side of the contact zone between the edge of the eye bar and the central pin. Although fretting fatigue had been shown in the tests to be a possible failure mode, the mechanism demands that fatigue cracks could grow only very near the points of contact between the eye-bar hole and the pin. Since the main load will occur along the chain, the contact zones will be on the long axis and not at 90° to the axis. The critical crack was found on the lower edge of the pin-hole at 90° to the axis, so is unlikely to have been formed by fretting fatigue.
There is no doubt that fretting action on the inner surface of eye-bar joint 330 occurred during its 39-year life. The surface next to the critical crack is very rough indeed, showing deep corrugations aligned circumferentially: that is, at right angles to the sub-critical cracks seen in Figure 40(a). There appears to be no obvious correlation between the crack positions and the corrugations, however. Fretting action will have been most severe on the highest joints of the chain where the load on the joint was greatest.
One factor that can cause serious problems in any material is the presence of residual tensile stress. The problem often arises as a direct result of manufacturing, when hot material is shaped and then allowed to cool to ambient temperatures. For large castings like those needed to make the eye bars, such residual stress would be modified by the subsequent heat treatment to strengthen the steel, but had to be studied as part of the research effort into the catastrophic failure of the bridge.
The residual stress was investigated using several methods, including the destructive technique of removing metal layer by layer, as well as by drilling holes in the suspect sample. The surface strain was monitored by strain gauges, which indicated that there was significant stress in the eye bars near the critical inner surface next to the pin. The cuts made in order to measure the residual stress are shown in Figure 44.
The researchers reported that close to the edge of the hole in each eye bar, residual stresses were extremely high as a proportion of the total yield stress. They plotted the hoop tensile stress against the distance from the edge of the hole to produce a graph as shown in Figure 45. The upper curves show the residual stress in the inner surface of the eye bar to be of the greatest magnitude.
Such large stresses as those shown in Figure 45 point to the reason why the crack grew initially, and then, when it had reached a critical depth of about 3 mm, catastrophically. On the inner surface the largest stresses observed fall above the top measure of the graph at 160 MPa, the greatest being 190 MPa, nearly a third of the yield strength of the material. The stress tends to drop inside the bar, although in different ways; at cut 4, a compressive state is reached in the middle of the bar.
Although not sufficiently recognised at the time, residual stress in the inner edge of the eye of the bars was clearly a significant factor in the disaster. Whether or not other eye bars were examined in a similar way remains at present unknown, and the follow-up with the makers, US Steel, also unclear. The residual tensile stresses will have been formed during casting and the subsequent heat treatment, and exposed at the inner edge when the central holes were machined out. Records of the heat treatment eye bar by eye bar should have been inspected by the investigators, but whether they did see such records remains unknown. Whether US Steel knew about the problem at all also remains unknown.
The design of the original structure was governed by applicable standards in 1926. The official inquiry found that the design and build fell within those limits, the most important being the allowable stress in the eye-bar chain of 345 MPa. The steel was to be made with a maximum elastic limit of 520 MPa, with a safety factor on the strength of the steel of 2.75. It was argued at the time that over 70 per cent of the load was from the self-weight of the structure. Other suspension bridges of the same time were built with higher safety factors, however. A similar design of eye-bar chain in a larger bridge at Florianopolis in Brazil was given a safety factor of 4.61 using an allowable stress of 320 MPa. The bridge used four eye bars rather than two, so had greater redundancy. Safety factors (see Box 7) of 4.6 to 4.7 based on yield stress were usual in wire suspension bridges (such as the Golden Gate bridge in San Francisco).
The safety factor chosen for any structure is simply an expression of the state of knowledge (or lack thereof) at the time, and should allow for any future uncertainties as well as present uncertainties, such as quality of the parts used in the structure. So the safety factor might be termed an uncertainty factor. When knowledge of strength and quality is poor, then the safety factor is high. Thus at the time of building railway bridges in Britain in the 1850s, cast iron was widely used as a principal structural material. Steel was not available until much later. It was known that cast iron was brittle in tension, and following early tragedies, it was specified that a safety factor of 6 should be used for railway bridge design. This safety factor allowed for stress concentrations such as blow holes and sharp corners in beams, although it could be exceeded if such defects were close to one another, when the effect of one is multiplied by the effect of the other. Thus a spherical blow hole (Kt ≈ 2) next to a circular hole (Kt ≈ 3) gives a net effect of Kt ≈ 6.
The safety factor is most critical and important for the most highly loaded parts of a structure, simply because they will be closest to failure if, for whatever reason, the maximum permissible load is somehow exceeded. Nowadays, structures are designed for maximum loading from a variety of sources that are often difficult to predict with high confidence. Those factors include:
Some are created by the weather, a notoriously difficult area to predict, and others by movements in the earth's crust (also difficult to predict). However, hurricanes and earthquakes do tend to occur in well-defined areas of the world, and so structures built in those zones will have a higher safety factor than elsewhere to allow for the extra risk there. All these natural phenomena will produce extra loads on a structure, and frequently cause catastrophic failure. Thus a double-deck highway collapsed on itself during the San Francisco earthquake of 1989 (killing 23 drivers on the lower deck at the time). Several stadiums in Europe collapsed with heavy casualties during the winter of 2005/6 owing to high snow loading.
However, little was apparently known then about the problem of stress corrosion cracking, although the type of strengthened steel used in the eye bars had been tested before use for its fatigue properties. It was known that the material would be subject to rusting, but it was thought that painting would prevent the problem becoming serious. Unfortunately the design of the pin joints left them completely exposed to the weather, and water could gather at the lowest points within the joint. Such hidden parts of the joint could not be painted owing to their inaccessibility, so were left unprotected. Moreover, the eye-bar joint could not be inspected without disassembly, an impractical solution for a suspension bridge.
The problem of SCC was also encountered in India in the 1920s, in the premature detonation of rifle shells. It was caused by attack on the copper content of the brass shell case by traces of ammonia in the air. The gas, which is produced during rotting of animal wastes, attacked cases that had been deformed during manufacture, leaving high levels of residual stress in the upper edges of the cases, where cracks were initiated. The problem was eliminated by annealing the shell cases to lower the residual stress, and storing the shells well away from sources of ammonia.
Cable suspension bridges are also corrosion-sensitive, the most recent example being the corrosion at the base of the main hangers of the first Severn suspension bridge. Although the high-tensile steel wires had been galvanised, salt water collected in the joints at the road decks and penetrated to the interior of the joints, causing breakage of the seal. The zinc coating corrodes preferentially, but once this is consumed then rusting of the core will occur. All had to be replaced at high cost during the 1990s.
Following the discovery of the broken eye bar near the top of the northern suspension chain on the Ohio side of the bridge (Figure 36), it was possible to reconstruct the sequence of events during the collapse.
When the side chain separated, the entire structure was destabilised, simply because all the loads it supported were immediately transferred to the adjacent parts of the bridge. With its support along the north side of the bridge destroyed, the road deck below the broken chain started tipping over to the north. The hangers holding it up on the south side started breaking, and the deck below the tower broke away. The road deck crashed to the ground on the Ohio bank of the river, taking its vehicles with it as it fell.
The Ohio tower was affected rapidly by the break in eye bar 330, and it leaned over to the north east, with buckling and fractures of its bracing elements below the top. The main road deck in the centre of the bridge was also tipping and dropping as the tower toppled over. As the tower fell, it pulled the south main chain over with it, putting it under enormous lateral load. In the main span, fractures of the hangers occurred on the south side of the deck; as the main chain fell to the north, it impacted the north tower, and eye-bar joints started breaking in the south chain. It was followed by main-chain joint fractures near the West Virginia tower. According to the official account of the disaster, these events occurred up to about 10 seconds after the fracture of eye bar 330. During this phase the road deck broke in the middle, taking its vehicles down into the river. The visual sequence is summarised by one of the eyewitnesses who was about a quarter of a mile away downstream on the West Virginia river bank:
I turned around and looked and saw the Ohio towers (sic) falling. The tower legs seemed to twist counterclockwise (when viewed from the top) and fall upstream and towards the center of the river. The center span of the bridge broke in the middle and fell straight down. It looked as though the cars on the center span all fell with the bridge and looked like they were falling in a funnel – some falling backward, some falling forward. After the center span fell, the West Virginia towers (sic) and span fell … the bridge was all down in a matter of five seconds as I estimate it…
The West Virginia tower was one of the last main parts of the bridge to fall, dropping into the river and facing towards the east. At the end of about a minute the entire superstructure, apart from the stone piers, had disappeared (Figure 35).
In the immediate aftermath of the disaster, it was vital to prevent any further collapses, especially on bridges of similar design. Two other bridges were built to a design similar to that of the Silver Bridge, one upstream at St Mary's, West Virginia and the other in Brazil at Florianopolis. The bridge upstream on the Ohio river, at St Mary's, was the focus of concern, and it was closed to traffic immediately after the disaster. The eye-bar design was actually quite widespread in other bridges, but frequently eye bars were provided not in pairs but in multiple connections, increasing the safety factor significantly. In the case of a single eye-bar failure, the others could support the load until repairs were made. This is certainly true of many British chain suspension bridges as well as US structures.
The inquiry recommended several key measures, which were enacted by President Johnson. They included:
A nationwide inspection of existing bridges (about 1 million) was quickly undertaken, and many problems identified and corrected.
Now watch the video below on the ‘Silver Bridge’ disaster and then answer SAQ 8.
Describe the failure sequence of the Silver Bridge in December 1967, indicating the direct cause of the accident and any contributing factors that led to the failure. Include in your answer the evidence for the particular causes you mention.
The Silver Bridge accident occurred owing to stress corrosion cracking of a pin joint (no. 330) on the upper part of the subsidiary suspension chain on the north of the Ohio side of the structure. The critical crack occurred at the bottom of the northfacing lower eye bar of the joint. Each joint comprised two pairs of hardened steel eye bars fitted onto a steel pin with screwed caps to close the joint. The joint was the first one below the top of the Ohio tower. The accident happened about 39 years after construction, when the crack became sufficiently deep to grow catastrophically. The disaster happened for a combination of reasons, including the following:
The evidence in support included direct examination of the failed eye-bar fracture and other surfaces, and tests on eye-bar steel including residual stress experiments by removal of material. X-ray analysis was used to examine the cracks in eye bar 330 for traces of impurities.
In terms of current knowledge of failure analysis, there are several gaps that could have been addressed at the time. In particular, the fact that the critical crack occurred in an eye bar below the top of one of the towers suggests very strongly that the level of residual stress varied between the eye bars. The origin of the eye bar analysed and discussed above is not stated in the official report, and it also appears that only this one eye bar was actually studied for residual stress. It would have been of great interest to have seen the variation of residual stress levels across the upper eye bars, because it is the only explanation of the formation of a critical crack in an eye bar exposed to lower imposed dead and live loads.
A second question arises about the source of the sulphur found in the critical crack initiation region. The official report points towards H2S, but this is a rare gas to have occurred in an open environment. Sulphur dioxide is a much more common pollutant, and could have been produced by a local power station using high-sulphur West Virginia coal. There was also a foundry in Point Pleasant, close to the east side of the bridge, which probably produced quantities of the gas during smelting.
A subsequent court case was brought by the injured victims and relatives of the deceased, alleging negligence on the part of the builders of the bridge. The case was rejected on the grounds that stress corrosion cracking of the kind found in the critical eye bar was not known at the time the bridge was designed. Although the plaintiffs received no compensation, the disaster had at least raised the importance of thorough inspection of an ageing infrastructure. However, bridge failures unfortunately continue to occur.
This free course provided an introduction to studying Engineering. It took you through a series of exercises designed to develop your approach to study and learning at a distance, and helped to improve your confidence as an independent learner.
The content acknowledged below is Proprietary (see terms and conditions) and is used under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Licence.
Course image: drpavloff in Flickr made available under Creative Commons Attribution-NonCommercial 2.0 Licence.
Grateful acknowledgement is made to the following sources for permission to reproduce material within this book:
Figure 2: Reprinted from Engineering Failure Analysis, Vol. 10, No. 2, Mirshams, R.A. and Sabbaghian, M. ‘Failure Analysis of an Elbow Tube Fitting’, pp. 215–221. © 2003 Elsevier;
Figure 3(a): © Simo Bogdanovic/Alamy;
Figure 4: © Kawase Shinichi/Digitalization: Kobe University Library;
Figure 7: © Paul Carstairs/Alamy;
Figure 8(a): From Neale Consulting Engineers;F
Figure 8(b): From CALCE News.
Figure 9: © TWI Ltd;
Figure 10: Courtesy of University of Washington Libraries, Special Collections [UW20731; UW21422; UW21413].
Figures 12: © Peter Lewis;
Figure 15: Courtesy of Oceanic & Atmospheric Administration/US Department of Commerce;
Figure 17: Pipeline Accident Report – Natural Gas Pipeline Rupture and Fire Near Carlsbad, New Mexico, August 19, 2000, NTSB/PAR-03/01, US National Transportation Safety Board;
Figure 24: © Colin Gagg;
Figures 27 and 31: Courtesy of Jack Burdett Collection, Point Pleasant;
Figures 26, 31, 35, 37 and 41–45: Highway Accident Report – Collapse of U.S. 35 Highway Bridge, Point Pleasant, West Virginia, December 15, 1967, NTSB/HAR-71/01, US National Transportation Safety Board;
Figure 30: Courtesy of James Bashford/Gig Harbor Peninsula Historical Society & Museum;
Figure 32: Earl T. Kilmer Collection. Presbyterian Church and River Museum;
Figures 33 and 34: Courtesy of Gannett Fleming;
Figures 38 and 40: Journal of Testing and Evaluation, Vol. 1, No. 2, © ASTM International, 100 Barr Harbor Drive, West Conshohocken, PA 19428; Figure 39: © John Bennett.
Don't miss out:
If reading this text has inspired you to learn more, you may be interested in joining the millions of people who discover our free learning resources and qualifications by visiting The Open University - www.open.edu/ openlearn/ free-courses
Copyright © 2016 The Open University