Popular Mechanics (South Africa)
A famed building’s fall from grace:
How the Salesforce Transit Center went from being to one of the century’s biggest construction debacles.
What went wrong, the solution, and what was learnt.
ON THE AFTERNOON
of Tuesday, 25 September 2018, Marc Benioff, founder and co-CEO of Salesforce, stepped on stage at the Moscone Center in San Francisco to deliver the keynote speech at Dreamforce, his company’s annual conference. The event – a combined business meeting, marketing rally, and
New Age retreat – attracted more than
100 000 people from around the world, closing off an entire city block.
Benioff had built Salesforce and its core product of cloud-based customer management software from a Telegraph Hill apartment into a $13 billion-revenuea-year juggernaut employing 30 000 people worldwide, with 8 500 in San Francisco. Just a few days before Dreamforce, he’d sealed a deal to purchase the struggling
Time magazine, prompting an admiring profile in The New York Times.
Completing his apotheosis, 25 September 2018 was Benioff’s 54th birthday. After his speech, he could return to his office in the 326-metre-high Salesforce Tower – the second-tallest structure west of the
Mississippi, whose naming rights he’d purchased in 2017 – and look down upon the Salesforce Transit Center and Park, his native city’s new crown jewel.
Conventional wisdom warned against Benioff buying the naming rights to the Transit Center. What if there was a wreck or derailment, chaining your brand’s name to a disaster? But to Benioff, the potential pay-off seemed to outweigh the risk.
Built at a cost of $2.2 billion, the Salesforce Transit Center and Park formed the cornerstone of the Bay Area’s ambitious regional transportation plan: a vast, clean, efficient web of trains, buses, and streetcars, running through a hub acclaimed as the Grand Central Station of the West. Naming this structure – the embodiment of a transformative idea – could yield marketing gold for Salesforce. It could also make Benioff a household name on the level of Bezos, Gates or Zuckerberg.
Benioff took the gamble in 2017, pledging $110 million over 25 years, with $9.1 million up front and the rest committed to supporting operations when the trains started running. For now, the train box sat vacant on the bottom level, awaiting a 2 km tunnel connection.
The rest of the complex had been open for six weeks. Bus traffic was running through the terminal, cutting commute times to the East Bay by up to 20 minutes thanks to its direct ramp to the Bay Bridge. Visitors flocked to the sumptuously landscaped rooftop park, compared by many to Manhattan’s famous High Line. The entire four-block-long, 100 000 m²-plus construction formed a modernistic gem, environmentally sustainable and covered in an undulating white aluminium exoskeleton patterned by physicist Sir Roger Penrose. Suffused with natural light, the building had striking, playful art everywhere you turned.
As he took the stage on his birthday at the Moscone Center, Benioff must have been confident his gamble on naming rights had paid off. He couldn’t imagine that at that moment, less than 2 km away, the ambassadors trained to welcome the public to the STC were now frantically waving commuters away. Rather than Grand
Central Station or the High Line, the Salesforce Transit Center and Park suddenly resembled the sinking Titanic.
EARLIER THAT DAY,
workers installing panels in the
STC’s ceiling beneath the rooftop park uncovered a jagged crack in a steel beam supporting the park and bus deck. ‘Out of an abundance of caution,’ officials said, they closed the Transit Center, rerouting buses to a temporary terminal. Inspectors were summoned. They found a similar fracture in a second beam.
Structural steel is exceptionally strong, but given certain conditions – low temperatures, defects incurred during fabrication, heavy-load stress – it remains vulnerable to cracking. Two types of cracks occur in steel: ductile fractures, which occur after the steel has yielded and deformed, and brittle fractures, which generally happen before the steel yields.
Ductile fractures develop over time, as the steel stretches during use, explains Michael Engelhardt, PhD, a professor of civil engineering at the University of Texas at Austin and chair of the peerreview committee overseeing the STC’s response to the cracked-beam crisis.
‘Engineers can predict ductile fractures and make adjustments during design, such as redistributing the load among various parts of the structure,’ Engelhardt says. ‘Brittle fractures, by contrast, happen suddenly and release a great deal of energy. They’re concerning. They aren’t supposed to happen.’
The cracks discovered beneath the rooftop park were classic brittle fractures. The tapered 10 cm-thick steel beams – 76 cm wide and 18 m long, with a horizontal flange on the bottom – undergirded the two hectare park on the building’s fourth level, and buttressed the roof of the bus deck on the second level. By themselves, the cracks formed a point of weakness with potentially hazardous consequences. But they also suggested the possibility of a larger crisis.
If two brittle fractures had appeared in the building’s 23 000 tons of structural steel, couldn’t there be others?
AT THE PEAK OF THE
evening rush hour, the Transit Center that normally teemed with buses was summarily closed. Mass confusion, an epic traffic jam, and a stampede toward BART trains and Ubers ensued. TV crews reported live outside the STC, interviewing angry and bewildered citizens.
Engineers and officials at the Transbay Joint Powers Authority (TJPA), the agency managing the Transit Center, were trained to deal with emergencies, but this was especially shocking. The project had been built by some of the most respected firms in the industry. Pelli Clarke Pelli Architects conceived the design. Thornton Tomasetti, Pelli’s collaborators on Malaysia’s iconic Petronas Twin Towers in Kuala Lumpur, served as the designer and engineer of record. The Bay Area’s pre-eminent contractor, Webcor/Obayashi, led the construction. Skanska, the construction firm behind New York’s World Trade Center Transportation Hub and Oculus, won the $189 million subcontract to furnish the structural steel. And the Herrick Corporation, another California construction heavyweight, had shopfabricated the girders in question, using steel flange plates supplied by two subcontractors.
There had been layers of inspection and code verification, including certifications of quality for the steel in the beams that fractured. In 2011, a year after workers broke ground on the STC, the TJPA had ordered a comprehensive review of its seismic design, which halted progress for 18 months. For a massive construction project in the heart of earthquake country, however, the time seemed well spent. After the reworking of the seismic plan, Fred Clarke, the project’s lead architect, had declared the STC as ‘probably one of the safest buildings in the world’.
AT FIRST, EVENTS
moved swiftly after the cracks were discovered. To ensure safety and stability, 6 m-high hydraulic jacks were installed to shore up the affected Fremont Street overpass. Crews stripped the fireproofing from the steel so engineers could begin inspection. Reporters arrived from CBS,
The Wall Street Journal, and The New York Times. An Associated Press story cited the transit hub as the ‘latest example of problems in a city brimming with homelessness and poor infrastructure’.
Engineers conferred, contractors scrambled to dig out blueprints proving the problem wasn’t their fault, and attorneys braced for lawsuits. Then the pace slowed, as officials realised that the two central questions raised by the fractures – ‘What went
If two fractures had appeared in the building’s 23 000 tons of structural steel, couldn’t there be others?
wrong?’ and ‘Was the problem localised?’ – would take months rather than days to resolve.
On 4 October, the mayors of San Francisco and Oakland sent a letter directing the Metropolitan Transportation Commission to assemble an elite peer-review committee to oversee the investigation and repair. The agency selected Engelhardt to lead the effort.
PROFESSIONALS IN
steel-fracture mechanics tend to learn from catastrophes. For Michael Engelhardt and many of his peers, the defining disasters included the 1994 Northridge earthquake in Southern California and the 1995 Kobe earthquake that devastated Japan. While Engelhardt was earning his doctorate in metallurgy at University of California, Berkeley, the 1989 Loma Prieta earthquake knocked down a section of the Bay Bridge, destabilised the Embarcadero Freeway in San Francisco, and disabled the city’s aging Transbay Bus Terminal. These events precipitated the demolition of the highway and terminal, and the eventual construction of the Salesforce Transit Center and Park in their place. Now, to square the circle, Engelhardt had been summoned back to the Bay Area to help rescue the project.
‘Our job wasn’t to decide who was going to get sued,’ he says. ‘Our job was to find out what went wrong, determine the scope of the problem, approve the fixes, and make recommendations moving forward.’
In assembling the review committee, Engelhardt made a point of including a welding expert. ‘In the world of structural steel,’ he says, ‘it’s usually the connections and joints that tell the tale.’
These points in a steel-formed building are also the potential weak spots, places where art and error come into play.
‘For function and economy in large-scale construction, steel is quite possibly the best choice,’ says Amit Kanvinde, professor and chair of civil and environmental engineering at the University of California, Davis. ‘The tricky part is making connections in steel construction, accounting for the various geometries and the changes brought about by welding during fabrication.’
ON 13 DECEMBER 2018,
Robert Vecchio, CEO of LPI Inc, a New York City firm that provides forensic metallurgy services, rose to speak at a TJPA board meeting. The gallery was packed with city and state officials, reporters from local and national media outlets, construction and civil engineering professionals, and members of
the public, all hungry for news about the Salesforce Transit Center, whose ignominious closure now stretched into its fourth month.
An internationally recognised expert in steel-fracture analytics who had worked on the breakage in the hull of the Exxon Valdez and the collapse of the Twin Towers during the 9/11 attack, Vecchio had been hired to determine the ‘root cause’ of the STC’s fractured beams. He was about to announce his preliminary findings to the board, and along the way provide a crash course in steel-fracture analytics.
Seven weeks earlier, shortly after the Transit Center shutdown, Vecchio’s team had travelled to San Francisco to supervise the removal of core samples from the damaged beams and to bring them back to the New York lab for testing. LPI technicians then performed scanning electronic microscopy, Charpy V-notch testing, Rockwell hardness testing, tensile testing, and fractographic analysis, with representatives from the project’s key stakeholders looking over their shoulders.
Now, in a concise PowerPoint presentation, Vecchio explained to the board that the cracks were due to a ‘perfect storm’ of the three factors that Engelhardt says characterise brittle fractures: weakness in the metal, damage during fabrication, and the stress of load during use.
The investigation focused on the 5-by-10-cm ‘welding access holes’ that had been thermally cut into the beams. Vecchio displayed a photo showing the red oxidised colour of the metal around the holes, indicating that microscopic cracks formed due to the intense heat generated by an acetylene welding cutting torch. He pointed out the build-up of martensite, a brittle substance with a crystalline structure, which formed around the cuts as they cooled.
Vecchio explained the high hardness of the structural steel made it prone to micro-cracks. But at the same time, he emphasised, the metal had been tested prior to welding and met all specifications and requirements. The problem was that the martensite deposit around the cuts hadn’t been ground smooth and polished after the welds had cooled. The martensite produced micro-cracks, which eventually grew into brittle fractures.
Vecchio’s team also found a relative weakness in the metal during the Charpy V testing, in which a pendulum-driven wedge slams down on a notched plug of steel to measure its ability to withstand the stress of welding. ‘The toughness level at the surface of the sample was good,’ Vecchio said, ‘but as you went to mid-thickness, the toughness dropped down quite a bit. Toughness in the centre line was very low, so the defects were sitting in material that had very low toughness. The plate itself did meet the requirements for this type of construction.’ Along with the microfractures in the unpolished steel around the holes and the stress produced by the weight of hundreds of 15-ton buses rolling above it each day, this weakness eventually produced the cracks, which likely started in July, and were discovered on 25 September.
The micro-cracks showed up during welding, Vecchio said in summary, and the combination of stress and load popped the micro-cracks into full-blown brittle fractures.
After a moment of silence, a board member finally asked: ‘Would a failure of this type suggest other places we should look in the design and fabrication of this structure?’ In other words, could other beams crack?
Mark Zabaneh, the TJPA executive director, stepped in to reply. ‘These reports are being turned over to the peer-review panel. We will follow their recommendation.’ He later told reporters, ‘We want to make absolutely sure the building is safe before we let the public back in.’
THE YEAR TURNED,
and the centre remained closed. Thousands of commuters continued to use the Transbay Bus Terminal on Folsom Street, which felt less ‘temporary’ with each passing week. And every day, workers poured into the Salesforce
Tower, where their boss looked down on his tarnished crown jewel. Aside from an apologetic and supportive tweet after the cracks were discovered the previous September, Benioff had been publicly silent regarding the closure.
Meanwhile, at the STC, engineers pored over documents and explored every corner of the structure. Officials examined 21 000 inspection reports. Ron Alameida, director of project management for the city of San Francisco, told a reporter that ‘essentially 64 000 times, things of concern were addressed and reviewed’.
Investigators sought places in the building that could be affected by the same combination of factors that caused the cracks in the beams above Fremont Street. They wanted to find out whether the perfect storm of defect, weakness, and stress formed a singular anomaly or a more general problem.
The focus settled on the girders above 1st Street, designed virtually identical to the girders above Fremont Street. The
1st Street girders supported the same bus deck and were composed of the same steel fabricated by the same subcontractor. Nonetheless, the 1st Street beams had not fractured.
There turned out to be a difference in the construction sequence of the girders. On the Fremont Street girders, the weld access holes were cut before the main welds were performed. During subsequent welding, the stresses caused very small cracks to form in the unpolished thermally cut access holes. These small cracks grew into the brittle fractures that appeared when the centre opened and heavy bus traffic stressed the girders. For the 1st Street girders, which did not fracture, the thermally cut holes were made after the main welds were made. There were no small cracks when the buses started to roll. This minor detail proved to be critical.
Four levels of inspection – by Skanska, Herrick, Webcor/Obayashi, and Turner – had missed the detail of the unpolished micro-cracks. After a March board meeting of the TJPA, Zabaneh said, the holes ‘were not installed to code in both dimensions and treatment, [meaning] they were not ground to bright metal finish… Had the weld access holes been ground per code, fissures would not have taken place and the girder’s bottom flange would not have been cracked.’ In an April letter to the mayors of San Francisco and Oakland, an official wrote that ‘The TJPA staff believes the steel subcontractor is the party responsible for the fracture’.
The various subcontractors argued vehemently about the specific identity of that party, a dispute which may take arbitrators and courts years to settle.
For Zabaneh and the TJPA, it was enough that the investigation showed the errors hadn’t affected any other piece of steel in the building. The two cracked beams proved to be sui generis, the problem confined to a single stretch of roof above Fremont Street.
Contractors performed a relatively straightforward, old-school fix: They sandwiched the affected beams between two giant steel plates fastened to the girders by hundreds of steel bolts, no welds required. Inspectors started to recertify the building, and the TJPA began plans to reopen.
FROM A CIVIL
engineering perspective, Engelhardt views the affair as a valuable learning experience, one that will likely lead to more stringent code requirements and upgraded inspection processes. ‘Most important, nobody got hurt,’ he says. ‘And the review determined that the two affected beams, both over 60 feet (18 m) long, barely moved an inch due to the fractures. The redundancies in design guaranteed the beams’ stability. The overall safety of the building was never compromised. If those workers hadn’t discovered the cracks by chance, we still might not know about them.’
For the TJPA and the greater Bay Area, however, the ordeal took a heavy toll. The closure stretched commute times, forced people to alter their daily routines, idled contractors and STC employees, and increased traffic congestion. For almost a year, people in San Francisco had to navigate around the giant, glittering, glaringly idle structure. In October 2018, citing the cracked-beams fiasco, the
San Francisco City Council gave a vote of no confidence to the TJPA, suspending funding for Phase 2 of the transportation project, which would deliver train traffic to the transit centre.
Nonetheless, officials chose Monday, 1 July 2019, as the day to reopen Salesforce Transit Center and Park. The date fell at the start of a holiday week, when foot traffic would be lighter, and most people wouldn’t be paying close attention.
THERE WOULD BE NO
pomp or glitz for the STC’s reopening. No block party, no crowd lining up for a ‘once-in-a-lifetime’ walk along the Bay Bridge–replica ramp, and no celebratory tweets from Benioff, who has maintained
‘If those workers hadn’t discovered the cracks by chance, we still might not know about them.’
his silence regarding the closure. (Salesforce did not respond to repeated interview requests for this story.)
On 1 July 2019, only the grand entry hall and rooftop park was reopened. (The city bus service resumed in mid-July, and Transbay buses started rolling on 11 August 2019.) For the first time, however, citizens were able to board the 20-passenger gondola on Fremont Street, a few steps away from where the beams fractured, and take a 30-second ride to Salesforce Park.
On the day of the soft reopening, I arrive at the centre shortly after 6 am, just as workers take down the barriers and open the doors. I walk through the Grand Hall, resembling a cathedral with its vaulting Tower of Light Column and terrazzo marble floor, inlaid with renderings of poppies and hummingbirds.
My footsteps echo in the empty hall. Outside, pedestrians flow around the building in the morning rush hour. They could take a shortcut through the hall to their offices, but they choose to keep to the sidewalks. It will take time for the city to trust the building again.
Paul Gribbon, a civil engineer who brought the Portland, Oregon $800 million Big Pipe project in on schedule and within budget, says that, along with cost and time overruns, there’s another law regarding megaprojects. ‘Once it’s up and running, once there’s a shining new bridge or light-rail station, people tend to forget about how much it cost, in all senses of the word.’
I step on to the 27 m-long escalator, polished spotless by overnight work crews. I ride up past the idle bus bays on the third level, continuing on to the fourth level and entrance to the park, which stretches like a stream-watered canyon through the surrounding office towers.
It’s hard not to gape at a place where flowers and trees sprout out of concrete and steel rather than soil. I envision the park when the Transit Center goes fully online: 15-ton buses coming and going like freighters in the deck below, Caltrain shuttling between The City and Silicon Valley, and further down, in the deepest reaches of the terminal, the sleek cars of the
California bullet train delivering passengers from LA. The park has whimsical kids’ areas and a cosy Starbucks, yet these gestures towards intimacy only magnify the immensity of the neighbouring Salesforce Tower. The entire scene feels gigantic, and fragile as a dream.
I wander past the bamboo and cactus groves and the monkey puzzle tree. I read a plaque explaining that the park rests on a base of structural foam, designed to let the structure ride out earthquakes, which, here on the lip of the San Andreas Fault, are sure to come. I recall architect Fred Clarke’s claim that the STC was probably among the safest buildings in the world.
At the mere thought of an earthquake, however, I reflexively imagine a TransitCenter apocalypse: buses crashing through the roof of the Grand Hall, smoke rising from shattered steel.
The vision passes as quickly as it arose. With a combination of fatalism and blind faith, I again trust the minds and hands that have built the $2.2 billion dream of the Salesforce Transit Center.