Why do bridges fail?
03 April 2018
Engineers at the site of the collapsed pedestrian bridge at Florida International University, Miami. Image: Courtesy of National Transportation Safety Board
On March 15, 2018, we saw the shocking collapse of a pedestrian bridge at Florida International University (FIU). Six people have lost their lives, an investigation has begun, and we have already seen plenty of speculation on possible causes.
Now I have no intention of adding to this speculation, and I will not be offering any opinions in this article – I am waiting for the official investigation reports that will no doubt appear over the coming months.
But it is true to say that this collapse – certainly more than other recent failures – has resonated with the general public. There is a cruel irony in a bridge collapsing and killing people, when it was originally intended to keep people safe while they crossed the road.
So in this article I want to ask (and hopefully answer) the general question of why bridges fall down. What are the most common causes of failure, and are they as intuitive as we think.
I want to use this FIU collapse as a starting point for trying to understand why bridge infrastructure can sometimes fail us.
Bridge failures in the United States
So where do we start?
Well there are many studies we could look at, but I want to concentrate on a study published in 2003 by Wardhana and Hadipriono. (We will also talk about more recent studies as we go along, but much of the later research comes to similar conclusions.)
Now this study investigated the causes of bridge failures in the US between 1989 and 2000. The researchers found that – over this period – a total of 503 bridges failed, resulting in 76 fatalities and 161 injuries.
What is meant by the term failed?
Well, the authors’ definition of failure was: the incapacity of a bridge, or its components, to perform as specified by its design and construction requirements. This definition included bridges that had totally collapsed, partially collapsed, as well as those that were distressed, for example, exhibiting excessive deformation. And when the researchers started looking at this data they found some interesting results.
The figure (below right) shows the 10 states with the largest number of failures. (There are actually 11 states in the figure because Mississippi and Missouri came joint ninth place.)
Some states had a considerably higher number of failures than others, specifically Iowa and New York. In fact, apart from New York, Iowa had more than twice as many failures as the other states.
Why was this the case?
A potential reason that may jump to mind is: had New York and Iowa more bridges? While this may appear a logical explanation, the authors ruled it out. At the time of the study, Iowa had a similar number of bridges to Missouri, and New York had a similar number of bridges to Minnesota. But despite these similar numbers, there is a significant difference in the number of bridges that actually failed. (It turns out that Maryland had the highest percentage of failures relative to total number of bridges.)
Taking New York first, the study’s authors pointed out that New York was one of the older states, so it was more likely to have older and – therefore more vulnerable – bridges than Midwestern states. (Of the 64 bridges that failed in New York state, 25 of them were more than 50 years old.)
Another reason the authors give is that one of the study’s key sources was a database that was developed and managed by the New York Department of Transportation. And this database recorded details of bridge failures around the country – not just in New York state.
The authors speculate, I think quite correctly, that the state of New York was much more diligent about providing information to the database – after all, it was their database. So New York may not necessarily have had a markedly higher incidence of failure than other states, but it probably had a higher incidence of recording these failures.
Examining the high number of failures from Iowa tells another story, however, and to help make sense of it, the figure below shows the number of failures that occurred each year. You can see there is a definite spike in failures in 1993, almost twice as many as any other year.
And it was in 1993 that the Mississippi and Missouri rivers – along with their tributaries – overflowed and flooded many states, including Illinois, Kansas, Minnesota, Missouri, and … Iowa.
This flooding wrecked infrastructure, and of the 112 bridge failures that occurred in 1993, a total of 75 of them were due to flooding. So the Iowa failure spike may be largely a consequence of this extreme event, rather than a general reflection of Iowa’s infrastructure. (But this doesn’t quite explain why Minnesota and Missouri – also flood affected – had a much lower failure rate.)
Of the 503 failures, a total of 386 occurred during a bridge’s service life, as opposed to during construction. Only eight failures occurred during construction. And in terms of years of service before failure, the youngest bridge to fail was just one year old, the eldest was more than 150 years old, and the mean time to failure was 52.5 years. Only two of the 503 bridges were pedestrian bridges. More than 50 per cent of the bridges were steel/beam girder or steel truss in construction.
And when it came to the consequences of these failures (as opposed to initiating cause), partial collapses were the most likely result, with 80 occurring in the data set. There were 12 cases of total collapse and 17 cases of bridges experiencing distress. But there were a whopping 277 cases where there wasn’t enough data to determine if the failure had been total collapse, partial collapse, or distress.
Now the authors do say that while it was usually possible to readily interpret the cause of failure from the data available, this issue of incomplete data was a real challenge in determining consequence. Further, more sophisticated information, such as the human and procedural causes of these failures, was essentially non-existent. It is very difficult to genuinely learn from failure without this sort of data.
In fact, the authors discovered that only 13 per cent of the failures in their data set were discussed in engineering news media. How can we as an engineering profession learn from failures when we don’t even talk about them?
Technical causes of failure
What were the prime culprits in terms of causation?
In some cases multiple causes contributed to failure, but the authors focused on the cause that primarily failed the bridge. And, in terms of primary causes, they found that design and construction errors caused few failures – a mere 2.6 per cent and 0.6 per cent, respectively.
And the number one primary cause of failure was hydraulics, such as flooding and flood related debris strikes – a total of 53 per cent of failures were attributable to hydraulic actions. But we need to be a little careful here, because the authors believe that the term ‘flooding’ also appears to relate to – or at least is interchangeable with – scour.
And if you are unfamiliar with scour, it is the erosion of a riverbank or riverbed (or sea bed) by flowing water, which undermines a bridge’s foundations. So this 53 per cent of hydraulic causes includes scour failures, and scour has long been known as a significant cause in bridge failures – it was the primary cause in the Malahide viaduct collapse.
A later study by Cook and Barr, based on New York state bridges, also identified scour as a major cause of failure, but, worryingly, found that 57 per cent of the bridges that failed because of it were given scour vulnerability ratings indicating they were stable.
Moving from hydraulic actions to the second highest cause of failure takes us to collisions. These accounted for 11.7 per cent of failures – markedly lower than hydraulic causes. Collisions relate to vehicles striking a structure. In this data, set bridges were impacted by 14 trucks, three trains, 10 ships and barges, with 32 collision causes remaining unknown.
A later study by Cook, Barr, and Halling, which included nine pedestrian bridge failures, found that eight failed because of collisions – the ninth failed due to a lack of lateral bracing in construction.
The third primary cause of failure was overloading. A total of 8.8 per cent of failures were due to bridges being subject to heavier loads than they were intended to carry. There were two pedestrian bridges in this study, and both failed due to overloading.
An interesting aspect of overloading failures, highlighted by Cook and Barr, was that of the seven bridges in their data set that failed due to overloading, six were load restricted to below the legal limit. (Further, five of these seven bridges were classed as structurally deficient before the collapse.)
The authors concluded that there was certainly a relationship between overloading failures and load-restricted bridges. While this may sound obvious, the lesson I take away is that these overloading failures are perhaps less an issue with dramatically overloaded vehicles, but rather an issue with under-strength bridges.
Now the striking thing is that hydraulic, collision, and overloading failures account for 73.4 per cent of failures. The vast majority of failures were caused by what the authors called external events. The bridges were subjected to conditions they couldn’t cope with. And later studies would reach similar conclusions – percentages for each cause may jump round a little, but primary causes of failure tended to remain the same.
While the figure above shows other causes, such as earthquakes and fire, I want to finish by discussing the fourth most common cause of failure: deterioration (for example, corrosion of the structure), which accounted for 8.6 per cent of failures.
But, again, we need to be a little careful with this percentage because it doesn’t quite tell the whole story. It looks like it suggests that deterioration issues only account for 8.6 per cent of failures. But this is for failures where deterioration is the primary cause. It also turns out that deterioration appears to make a significant contribution to facilitating other primary causes of failure.
In other words, deterioration weakens a bridge, thus allowing other primary causes, for example, scour, to fail it. Cook and Barr published research in 2017, which examined bridge failures in New York state between 1992 and 2014. A total of 98 bridges in their data set had collapsed – 45 of them were classed as structurally deficient, while 53 were classed as non-structurally deficient.
So a broadly similar number of structurally deficient and non-structurally deficient bridges collapsed. But when you compare these numbers with the fact that there were 2,012 structurally deficient bridges and 15,448 non-structurally deficient bridges in the state at the time, we see that structurally deficient bridges have a much higher likelihood of failure.
In other words, these numbers tell us that the more structurally deficient bridges we have in the network, the more likely we are to see bridges fail. This, of course, appears intuitively correct.
Now hold that thought for a moment.
And consider that the 2017 Infrastructure Report Card identified that of the 614,387 bridges in the US in 2016, a total of 56,007 (9.1 per cent) were considered structurally deficient – although this percentage is down from more than 12 per cent in 2007. To put it in human terms, there are, on average, 188 million trips taking place across a structurally deficient bridge every day in the US.
What does all this mean?
All of these studies tell us that bridge failures can and do happen, and they are likely to continue to happen. One 2015 study estimates the average annual bridge failure rate for the US is 128 failures per year, with many failures going unreported.
So while we wait for the findings on the Florida collapse, it reminds us that these pieces of infrastructure, which many of us consider so solid, so permanent looking, can actually be quite vulnerable.
It is only by understanding why failures happen, that we can hope to better manage this vulnerability.
All above charts are reproduced from data presented in the Wardhana and Hadipriono study
This article was first published as a blog on March 22, 2018, in Sean Brady’s LinkedIn
Sean Brady is the managing director of Brady Heywood (www.bradyheywood.com.au), based in Brisbane, Australia. The firm provides forensic and investigative structural engineering services and specialises in determining the cause of engineering failure and non-performance. Follow on Twitter @BradyHeywood