It’s a dangerous business when things go wrong in the public eye. I’m not talking about the physical danger of an accident in elite track cycling, although that certainly is hazardous; without a doubt, a crash at 70 km/h is severe, especially when surrounded by other bikes travelling at the same speed. No, the danger to which I refer is the danger inherent in the speculation and blame game that inevitably follows a public mishap, in this case a part failure, and that can lead to unintended consequences if rumour and supposition take hold without a full understanding of the facts.
There is something about us humans; it is all too easy to play the blame game and, sometimes, see fortune in the misfortune of others. The Germans even have a word for it: Schadenfreude, ‘pleasure derived from another person’s misfortune,’ a portmanteau based on the German words Schaden (harm) and Freude (joy).
I don’t mind admitting that when I first heard about the part failure of the additively manufactured titanium handlebars experienced by Alex Porter in the Australian Men’s Team Pursuit race at the Tokyo 2020 Olympics (held in 2021), I had two initial thoughts: “That must have hurt,” and – you guessed it – “I wonder which metal AM machine they were manufactured on?” The footage is quite shocking and Porter was fortunate not to have sustained much more serious injuries. He was also angry, and it certainly prevented AusCycling (a new organisation created in November 2020 out of a merger of Cycling Australia with two other cycling bodies) from achieving anything better than a bronze medal in the Team Pursuit. And nobody trains for a bronze medal.
Of course, I was utterly wrong. I was wrong to jump to conclusions without knowing the facts; I was also wrong to immediately associate the part failure with a metal AM machine of a particular type and infer that some devices are materially better than others. In my experience, most metal AM machines can be persuaded to work well; it is often the demands placed on the user’s know-how that are the most critical factor in the outcome.
Having absorbed the excellent 170-page report commissioned by AusCycling, freely available to download (see end of article), I will endeavour to provide a more balanced overview and try and seek out the useful lessons for our still relatively young – but increasingly professional – industry.
Elite sports and the performance race
There is a lot at stake in elite sport: vast sums of money, the pride of nations, the hard work of the athletes and support staff, and the expectation of medals. Teams compete for funding for each discipline, and cash is often directly tied to medal performance. So, the better each team does in its chosen sport, the more likely it will receive funding to invest in the latest equipment, coaching and sports science, which ultimately should advance its performance to achieve an even better medal tally next time. But the more you get, the more the expectation rises, and so the pressure increases. This causes teams to seek out every possible (legal) performance gain available. When it comes to equipment, this usually involves advanced engineering. Due to the strict timing of events and the drive to discover and implement performance gains quickly, advanced engineering, delivered to the point of need as fast as possible, is paramount. Then, the performance benefits can be assessed and further design iterations executed.
It’s hard to imagine a more high-pressure scenario outside of critical infrastructure, defence, aviation and emergency healthcare. And, as I’m sure we can all attest, from our own experience, it doesn’t matter if it’s not important to anyone else; if it’s important to us, then it’s our prime focus.
For reasons of balance and to get a perspective on just how far equipment in elite sport has come, I’m going to reference something that might astonish our younger readers and jog the memories of those in advancing years like myself, at least if you followed cycling or the newspapers at the time. Back in the early 1990s, there was a Scottish cyclist called Graeme Obree, known in the press as The Flying Scotsman. He won two track cycling world championships and contested and beat the Hour Record held for about eight years by Francesco Moser, where riders cycle as far as possible in one hour. He broke the record a second time, surpassing a new record set by Moser, in 1994.
Graeme was not one to follow convention. Besides being an amazing athlete who could push himself to and beyond the limit, he was a disrupter in bike technology. He was also willing to test the competition rules to the limit. He hit the headlines having designed his own bike, produced in his own small workshop, without finite element analysis or CAD. He used the bearing from a washing machine drum to allow a narrower profile to the crank and chainrings, and very quirky handlebars that tucked the rider’s arms and elbows underneath the torso, thus creating as low a profile aerodynamic form as possible. The bike almost looked like a throwback to the velocipedes of the early 1800s. Looks were deceiving, though, and the performance gains achieved resulted in a ban for this style of bike, with later designs after the ban ultimately resulting in the rules set in place by the UCI (cycling’s governing body) today, which strictly govern the rider’s position.
Why am I telling you all of this? Well, I guess, for a few reasons. If Obree’s bike had failed on television when he rode in (and won) the individual pursuit at the 1995 world championships, most people would have put it down to its homemade nature. There probably would not have been an extensive report, because it would have been clear where the fault most likely lay. Obree would have gone home, scratched his head and quietly resolved the issues; that’s the advantage of a small team. Besides, Graeme was looking for a great leap forward, so he wasn’t so much pushing engineering boundaries as he was pushing the boundaries of convention, so the engineering could be robust. At the time, many other riders saw Obree as an outsider, and did not take so kindly to his maverick nature; opinions were divided on where he got his performance from: was it him or his rule-breaking bike? The answer? Have a dig around on YouTube for the video ‘Graeme Obree – Athlete or Genius.’
There are no such issues today, with the rules around the bike design tightly governed by the UCI. In many ways, this means that today’s teams must search much deeper and look far harder for the performance gains that will give them a competitive edge. Innovations and advancements, in today’s world, tend to be incremental, rather than seismic. This means a highly data-driven approach; larger teams; more management; tighter product specifications; and marginal changes to shave off a hundredth of a second here and there. It also means lots of iterations, which piles on the time pressure, especially when it comes to hardware development. On the face of it, this sounds like the perfect opportunity for a flexible, digitally driven manufacturing technology, capable of producing parts that are nominally the same, but with minor geometric tweaks to improve performance. That’ll be industrial metal Additive Manufacturing, then.
The emergence of AM for bikes
There is no doubt that when Rapid Prototyping first emerged in the mid-1980s, its primary application was to accelerate product development. Still, it would be more than a decade before anything remotely structural could be produced. Even then, nothing that would withstand significant forces, such as those at play in track cycling. The closest the technology got to manufacturing bike components was making cores for a conventional investment casting process, which might then be used to produce handlebars.
The first metal AM technologies started to hit the market at the turn of the century and the first glimmer of flexible on-demand manufacturing for one-off or small batches of end-use AM parts emerged. The three words ‘metal Additive Manufacturing’ don’t fall into wider use until around 2006/7. The point is that it’s still early days for a relatively new technology, but, as with many aspects of our world, the pace of technology adoption is accelerating. Metal AM is now firmly in use to produce aerospace components and medical implants, two highly regulated industries that have spent at least fifteen years climbing the adoption curve and now have whole factories dedicated to metal AM production. Surely it’s possible to make some bike components?
The most distinct aspects of the aerospace and medical sectors are that they are highly regulated, high value, large scale, and highly experienced in controlled manufacturing. Significant constraints are placed around all processes. While speed to market for new innovations is an attractive feature of AM, it is generally not the primary motivator in these sectors. Instead, these industries are adopting metal AM to unlock performance enhancements in the component, reduce inventory by part consolidation, or perhaps exploit the geometric freedom offered by the process, maybe even process a challenging material that is too expensive to form by other means. Generally, they are not just using metal AM as a shortcut to get a product significantly faster than via a conventional route.
The demands of regulated sectors actively prevent the rash use of any technology, instead demanding rigour from fear of litigation and/or serious commercial penalties, either imposed or delivered by an unfavourable market response. Nobody wants to go through the trauma and reputational damage caused by a product recall, so companies develop systematic ways of working and follow international standards to safeguard against this as far as possible.
While I’m not suggesting that the field of elite sports plays fast and loose with how it uses technology, I imagine the perspective is different from that of the high-value manufacturing sector. Problem-solving is perhaps more centred around the human physiology of the athletes – “What do they need to perform better?” – rather than “How do we build a robust business using manufacturing technology to produce advanced products to serve a market and make a profit?”
So it’s understandable, especially when market adoption of metal AM technologies in regulated advanced manufacturing companies is relatively high profile, that elite sports teams should want to embrace such a flexible and increasingly well-proven technology; one that can make parts that fly into space and on commercial airliners, and are implanted into people’s bodies. But without the deep pockets and prospect of an ongoing commercial return, how can elite sports teams benefit from the potential of metal AM without running the risk of an engineering overhead that is too great a burden to bear?
So what led to the catastrophic part failure in Tokyo?
It’s probably worth setting out precisely what happened to AusCycling, and the chain of events leading up to the upsetting scenes at the Izu velodrome in Tokyo, before we go into the why. At this point, I want to note that the report precludes selective quoting, so I will do my best to avoid this and, as far as possible, paraphrase the events and findings. I’ll also focus on the most relevant findings for readers of Metal AM magazine, rather than list every nuanced point made in the 170-page report, which is available in full online for anyone wishing to take a deeper dive. Metal AM is a publication whose remit is to be a positive force for the development and promotion of AM technology worldwide. The aim is to provide a balanced view of the report’s findings, give credit to all those involved and highlight learnings that apply to all companies and organisations seeking to adopt metal AM.
COVID-19 in the mix
Like many competitive teams, Cycling Australia will evaluate several high-performance equipment suppliers for their elite riders. In the case of the Olympic cycling team, they settled on a Canadian manufacturer, Argon 18. The bikes were designed in carbon fibre specifically for Cycling Australia and custom-built for each rider; the frames were manufactured in Asia. Carbon fibre is a process that generally requires tooling and carries a specific lead time that can become challenging if late changes are needed. At this point, it’s also worth mentioning that the pandemic was to have an effect later in the programme, in multiple ways. First, the delay to the Olympics, disruption to supply chains and international travel ban further hampered both progress and potentially decision making, not to mention training and performance advancement; the incremental gains I talked of earlier.
I can imagine it was a less-than-perfect environment to develop and organise a team to challenge at the highest level. Whilst I know all the teams were in the same situation, it’s not unreasonable to think that the disruption led to an overall increase in risk that would most likely appear somewhere in one of the teams; still nominally a level playing field, but perhaps a bumpier one.
With long lead times on the frames, the trigger for production and the end of the frame design phase came around two years ahead of the original schedule for the games in 2020. After this point, changes to the bike would need to be accommodated somewhere other than the frame.
The human form factor
In parallel to the equipment development, which also involved changes to the frame to reduce its size and aid aerodynamics, the coaching team worked on the starting technique. This involves the rider using the inertia in his body to lunge forward, effectively catapulting the bike below them forwards and setting it in motion. This technique places significant loads on the frame, and the handlebars in particular. Back in Obree’s day, the starts were very different, with a slow build-up to wind up the massive gear needed for speed. This development in the starting technique revealed a problem with Porter’s knees hitting the lower part of the handlebars, the base bars, and a request was made to increase the handlebar clearance.
This left the team with limited options except to change the handlebars to introduce additional clearance. Using the original handlebars, which had been in use for some time in ‘standard’ form without any issues, the design was modified to alleviate the knee clash and a trial polymer AM part from a non-structural material was produced to check the form and fit. So far, so good; and the Olympics were still scheduled for July 24, 2020. The games were around fourteen months away; it seemed like there was plenty of time. The announcement that the games would be postponed did not come until about March 24, 2020.
Timing, and the attraction of AM’s speed to accommodate late design changes
In parallel to the Olympic build-up, AusCycling had already worked with Bastion Cycles in Melbourne, which had developed significant expertise in the metal Additive Manufacturing of high-performance structural cycle frames and components to their own designs. AusCycling was using Bastion components for its training bikes, so it was engaged in designing and producing the revised handlebar design.
It’s unclear why, but there was a six-month delay from making the additively manufactured model to check clearances before the final handlebars were ordered from Bastion. The report makes clear that the timing and sequence of the events leading up to the accident were contributing factors in the part failure. Further analysis of the email communication reveals that the problem with the knee clash had been known for more than a year prior to ordering the replacement modified handlebars from Bastion. So, why was production not progressed as soon as the knee clash was identified?
One aspect of AM is the genuine benefit of responding to late changes in designs; indeed, it is one of the many advantages marketed by AM companies. I know this, as I have peddled this idea myself. So, it’s perhaps understandable that an assumption may have been made that it was OK to postpone producing the new design due to the quick nature of metal AM. However, this overlooks the other discoveries that can be made once the parts have been tested. In the case of AusCycling, this delay in production had the knock-on effect of reducing the time available for thorough durability testing.
Material and design changes
The report talks about two other critical aspects of the project: the choice of materials and the modification of the existing handlebars, specifically the precise specification of the revised handlebar design. Having read the report in some detail, it’s unclear precisely what material was used for the original handlebar design from Argon 18. The report notes that an attempt was made to revise the handlebar design with Argon 18. However, this was impossible due to China’s production constraints, indicating that the original material was not AM titanium; most likely, given the production route, the actual material was carbon fibre.
This is perhaps where things start to unravel a little, although I can imagine, at the time, it probably didn’t seem like a tipping point. It’s only with the benefit of hindsight that the potential impact of each event becomes clearer. Once it was decided that a different route was needed to manufacture the modified handlebars, this should have led to a complete re-evaluation of the design requirements; taking in the specifics of the geometry, the loads and forces, the manufacturing process, the available materials, and the time available for production, post-processing and testing.
I can see how the actual events may have unfolded from reading and absorbing the report. The report picks up on issues that initially seem relatively trivial (or perhaps a better description would be incidental). There are a couple of more serious issues, but I can see how it would be possible to ‘sleepwalk’ into them if the chain of events was disrupted – say, by a pandemic, or a change of management structure. Both occurred in the run-up to the Olympics.
Before producing the non-structural polymer AM test handlebars, a request had been made to the Australian Institute of Sport for the extended design to be additively manufactured in titanium. For some reason, this did not happen; instead, the non-structural polymer AM handlebars were made and tested for rider position. Six months passed before the project was picked up again and an order was placed with Bastion. Unfortunately, Bastion was unaware that the detailed FEA and load calculations that governed the first design had not been revised for the second design. However, considering the nature and relative infancy of metal AM, Bastion did build several coupons to verify the values used in the FEA modelling to validate the actual mechanical properties of the coupons for the design calculations, demonstrating a thorough approach.
The cut-and-shut redesign to extend the base bar handlebar positions forward by 35 mm was adequate for the fit and form tests used to establish that the additional clearance was satisfactory. However, they were not sufficient for the structural part. In addition, the approach taken did not account for the change of material, and the new and significantly altered load conditions the new handlebars introduced due to the increase in the moment of force on the handlebars.
While a presentation capturing the design requirements shared with the Australian Institute for Sport featured the original request for titanium AM handlebars, this information was not shared with Bastion, perhaps because of the time lag or the potential change of management. Instead, the model data (the report calls this a skin-only drawing) used for the non-structural polymer AM part was used as the basis for the redesigned/remanufactured extended handlebars. As with most AM processes, this skin-only data is most likely STL data, although the report is not specific.
Maintaining continuity is challenging in all organisations, and can generally only be achieved with good documentation and rigour; often resource-heavy and requiring a particular mindset. The report refers to a culture within the AusCycling technical development team that was not fully aligned with the level of rigour necessary to effectively control and impart information.
Evaluating the impact of load, combined with the unique properties of titanium
A crucial opportunity had been missed in not evaluating the new load conditions and change in material. In particular, the report references the fact that, while titanium is around four times stronger than steel, its behaviour under cyclical loads is limited by its lack of ductility, loads such as those at play on a bicycle (there is a clue there, I think). In engineering terms, this reduced resilience to fatigue or lack of ductility must be accounted for in the design calculations and the subsequent testing.
In addition, a greater factor of safety can be designed in. The fatigue properties of titanium were discussed within the engineering team at AusCycling, and testing to appropriate standards was specified in the purchase order to Bastion. A suitable test rig was designed to apply both static and dynamic loads to the handlebars, and a static force of 1000 N – called for in ISO 4210-5; 2014 – was used, with the handlebars passing this test.
But this was not representative of the in-service conditions for the handlebars, so the test called for specific conditions and a number of cycles for the dynamic test, which is designed to replicate actual riding conditions more precisely. ISO 4210-5; 2014 calls for 200,000 cycles at a specific force and the automated test rig controls the frequency and conditions. I can only imagine that time was running short at this point, because a decision was taken by AusCycling to reduce the number of cycles from 200,000 to 50,000. The premise was that the 75% shortfall in fatigue testing cycles would be somewhat mitigated by rider testing, which could, presumably, commence sooner than if the total number of fatigue cycles had been completed. The inference is that the rider would uncover any lack of durability in track testing.
While this might be an acceptable method for a more ductile material that may show tell-tale signs of deformation before failing, titanium’s known lack of ductility and durability results in catastrophic part failure with little warning. This also points back to the need to clearly understand the choice of materials early in the design process so that appropriate tests and test rigs can be considered as part of the design-for-production case.
While the properties of AM Ti6Al4V were understood by Bastion, and an appropriate heat treatment cycle to ensure the mechanical properties was undertaken, further data was gathered from heat-treated coupons that indicated little difference between highly oxidised coupons and mildly oxidised coupons. Oxidation takes the form of discolouration to the material’s surface, with mild oxidations showing up as a straw colour and heavier oxidation a deeper blue. This minimal difference in the ductility of the highly oxidised samples versus the mildly oxidised samples is somewhat surprising. However, from my experience working in Laser Beam Powder Bed Fusion, I know that ductility in Ti6Al4V can be significantly improved by reducing oxidation and choosing the right heat treatment conditions and temperature profile. Understanding this is a must for anyone working in titanium.
The report clearly outlines that the reduction in testing and the incomplete and inadequate design specification were probably the two most significant factors in the unexpected part failure during the competition. Of course, there is no guarantee that the part failure would have been initiated had the full 200,000 cycles been completed and the handlebars all passed static and dynamic tests. This also raises the question of the expected performance life for components at the edge of their design capabilities. The report found comprehensive record-keeping about the lifecycle of each set of handlebars was lacking, and, with designs at the extremes of capability, record keeping becomes paramount.
When considering the structural requirements for the handlebars, precise rider conditions and forces need to be established. In the case of AusCycling, this was achieved by using force meters within the cranks, which measured both the forces applied in the downward pedal stroke and the upward pedal stroke, with the upward force accounting for around 20% and the downward force around 80%. The upward force was subtracted from the downward force, and the rider’s weight subtracted from this value. The result was the opposing force that the rider would exert when pulling up on the handlebars. These values were gathered from multiple riders. In the ISO tests, the force used was 1,000 N. Following the part failure and further analysis, the original specification provided by AusCycling proved to be incorrect for the modified handlebars; the correct value for the upward force was, in fact, 1,411 N.
When considered as a safety factor, a value of one is regarded as the minimum required; with the revised calculations, this resulted in an unacceptable value of 0.65 under fatigue conditions, effectively only 65% as strong as needed. The report analysed how the mathematical model had been constructed and established that some of the factors affecting the calculation were not correctly accounted for. Neither had been appropriately checked, ultimately resulting in under-reported values for the forces on the revised handlebar design.
It’s commendable that AusCycling has been so open with its findings, and the report does not pull any punches about where things went wrong and how improvements can be made. Elite sport at this level is generally funded from the public purse, and a duty is owed to all those involved to ensure integrity and to allow others to benefit from the report’s outcomes. The accident could have been much more severe, and Porter was lucky to escape with relatively minor injuries. With better oversight of engineering decision making and more robust processes in place, the risk of a part failure of this type – and the accident that resulted – could have been significantly reduced or prevented.
Metal AM demands the same levels of rigour as other manufacturing processes
Companies considering implementing metal AM must recognise that it is a process that requires at least the same levels of rigour as other manufacturing processes. It cannot be used as an easy gap filler or a quick fix, particularly in structural applications. Metal AM is one of the most flexible manufacturing processes available, but many process variables need to be understood and controlled in a comprehensive specification.
The report shows record keeping at Bastion was very thorough, particularly around machine checks and machine monitoring. Whilst not the state of the art when compared to the latest generation of AM machines, the best of which can gather vast amounts of data, the record-keeping at Bastion required user discipline and manual data entry. However, it is robust, and no corners were cut. The report also praises Bastion for its openness and engagement with the report’s author. The company’s desire to learn from the experience and improve its own processes shone through.
When an incident with such a high profile happens, especially in the public eye on TV, it is human nature to speculate about the cause. I know not everyone watches the Olympics, and this incident may have passed many people by, but a thorough post-event evaluation and report provides invaluable learning to those interested in cycling or sports technology development and the wider application of metal AM. Identifying the influencing factors and the root causes has many benefits. Most importantly, it protects the athletes – i.e. the end-users of the equipment that failed. When new technologies are used, it provides valuable insight into how these technologies and processes are used to the best effect and highlights the pitfalls that can often only come with experience. It also demands that those involved are robust enough as individuals to put themselves under the spotlight. The cultures they operate in must therefore be open, honest, and forgiving enough to make it safe for their employees to come forward and contribute.
“If you think health and safety is expensive, try having an accident”
This phrase is attributed to Stelios Haji-Ioannou, the EasyJet founder. Safety is, increasingly, part of our culture, and I suspect that, if this part failure resulted in more serious injuries, it would have attracted the attention of the governing authorities in Australia, as it would have done in many countries. That really is the kind of attention organisations don’t need.
I know from personal experience, and I suspect this was potentially an aspect of the failure at AusCycling; just defining the tests or completing the risk assessments is only part of the process of minimising hazards. The level of rigour in applying the tests or mitigations is what matters most. AusCycling defined the tests and procedures, but made flawed decisions when applying them, which lacked rigour and a deeper consideration of the consequences of those decisions. I’m sure that it felt as if they had done everything right, and this perhaps led to a false sense of security.
If there is a lesson here, it is to continue questioning throughout the progress of each project, particularly at inflexion points such as a change of strategy. When looking at the images of the original and modified handlebars in the report, there is an obvious visual difference: a clue. I’ve no idea if the two designs ever sat side by side in the workshop; had they, I’m sure it would have raised the question, “Are they still OK?” Relying on numbers when the mathematical model is incorrect will not give the correct answer, but it’s also much harder to see the difference in numbers than real-life parts. Hence, a blend of data and eyeballs is, perhaps, a more prudent approach.
Metal AM and its widening adoption will provide many more companies with the opportunity to benefit from its valuable attributes. However, the challenge of understanding its performance envelope, and how to apply it safely and to the best effect, demands respect. AM is not a magic bullet or a quick fix; one can’t just throw parts onto the machine and expect to get the best results. The key is to neither underestimate it, nor overestimate it. Strive to understand it before you ask it to fly to the moon or win a gold medal.
Cheshire, United Kingdom