Thursday, May 31, 2018

Nipigon River cable-stay bridge failed 6 weeks after opening

The Canadian transcontinental road uses the Nipigon River Bridge to connect western and eastern Canada.
Construction began in 2013 to build the first cable-stayed bridge in Ontario, two lanes at a time. In 2014 the north and center towers were finished and the north half of the bridge was opened in 2015. The old bridge was closed, and demolished. The south tower was built in 2016 and the deck was built in 2017. [enl-tbay] An animation of how a cable-stayed bridge is built. A time-lapse video shows that the towers were finished with precast segments and a tall crane after a concrete pump truck was used to build the foundation.

Fall 2017

Facebook from
Unfortunately, the $106 million bridge failed on Jan. 11, 2016, just 42 days after it opened to traffic. One end of the deck lifted 60 centimeters. Many of the articles I read reported the following:
An engineering report received by the province in July, but not public until September said there were three reasons for the bridge's failure.
  • The design of the shoe plate and its flexibility
  • A lack of rotation in a bearing that was constructed
  • Improperly tightened bolts attaching the girder to the shoe plate [they were (way) too loose]
The province reiterated in the fall that neither cold temperatures nor wind were to blame for the bridge's failure. It also stressed that overloading — not any kind of flaw — caused the bolts to break. []
It is estimated that permanent repairs will take three years, which is longer than it took to build the bridge itself. [] Evidently the demolition of the old bridge was efficient because the trans-Canada road was severed for 17 hours until an emergency one-lane repair was done and for six weeks for heavy truck traffic until a temporary repair was done.

The one-lane emergency repair was done by piling a lot of weight at the end of the other lane. Note all of the concrete barriers piled crisscross at the end of the deck in this screenshot. That is the weight they are using to close the 60 centimeter gap. [Globe&Mail]
Screenshot from the first video in GlobalNews-20160111
Until the temporary repair was done after six weeks, the failure "caused up to 1,300 trucks – carrying an estimated $100 million worth of goods – to detour each day for several weeks." [Globe&Mail] The detour is to go down into the US and go along the south shore of Lake Superior. That is, we are not talking about going 30 miles up the river to the next bridge because this is the only road bridge between eastern and western Canada!

The spans are 820 feet (250 meter), but I have not found the height of the towers.
Canam Bridges    This page also has a video of drone footage of the construction.
The precast deck panels are reinforced with fiberglass instead of epoxy-coated or galvanized steel rebar. More specifically, the deck panels are "glass fiber reinforced polymer-reinforced concrete (GFRP-RC) deck panels. The GFRP features vinyl ester resin and boron-free E-glass fibers." [ComposttesManufacturing]

As to why there is only one road (and bridge) across Canada, Ted Gregory commented in a posting:
The Canadian Pacific transcon is immediately next to the Highway bridge.When CP built their line across Canada, the stretch across the north shore of Lake Superior was the most difficult to build.CP has 5 mountain crossings and this was the most difficult!!
There's alot of profile and massive granite outcroppings on the north shore of Lake Superior. 
Another factor is the weather. The storms are ferocious. Infact it was during a winter storm that the connecting pin snapped. The Nipigon River handles a tremendous amount of water with alot of momentum. In its 30 mile length, 4 dams exist for hydroelectric power. It is the only outlet for Lake Nipigon, the largest lake in Ontario.
There's alot of profile and massive granite outcroppings on the north shore of Lake Superior. Another factor is the weather. The storms are ferocious. Infact it was during a winter storm that the connecting pin snapped. The Nipigon River handles a tremendous amount of water with alot of momentum. In its 30 mile length, 4 dams exist for hydroelectric power. It is the only outlet for Lake Nipigon, the largest lake in Ontario.
He later posted a link about the CP construction.

Some of the articles about the bridge mentioned that the province has decided that they need to start an environmental study in 2018 concerning an emergency detour across the river. The mayor of Nipigon indicated that they have been asking for an emergency detour route for 10 years.

Just before I was about to publish, I found a Technical Review. So I deleted all of my ranting about having just three bullet items of information. Now I will try to summarize 110 pages of information.

The obvious problem is that 40 bolts broke! The investigation revealed that some of the bolts had already broken during the previous six weeks of service because they were rusty. The rest broke with a cascade failure. (When a bolt fails, the remaining bolts have to hold more force, causing some to break, which means the remaining bolts have to hold more force, causing....)

Technical Report, Page ii
In the following text, CHBDC stands for Canadian Highway Bridge Design Code.

But of course the real issue is why did the bolts break. We start with a diagram that explains the terms in the bullet items.
The purpose of the bearing is to allow the girder+shoe-plate+upper-plate to slide back and forth over (for down forces) or under (for uplift forces) the lower plate+abutment-foundation as the span expands and contracts due to temperature changes.
The bearing also needs to absorb subtle rotational forces as the steel deck girders bend depending where heavy trucks are located on the deck.

In the following photo of the bearing, I put a red rectangle around the visible bolts that broke.
Technical Report, Page iii plus Paint
The root causes were that the nuts were not tightened enough and the PTFE (Teflon) sheets on the bearing surfaces lost some of their "slipperiness" because the surfaces were damaged by the design not accounting for rotations. Another source of the PTFE damage was that the bearing surfaces were not installed parallel to each other. The extra friction caused by the PTFE damage created unexpected shear forces in the bolts. The bolts were properly manufactured. Wind and cold were not an issue.

Technical Report, Page 5
The bearings were not parallel because the upper plate was bolted to the shoe plate, which was already bolted to the girder. And the lower plate was already set in the concrete foundation. But just because the girder is a straight line on the drawing and parallel to the lower plate does not mean that will be true in the real world. Dimensional imperfections in the girder bottom flange during manufacturing and deviations from the girder being parallel to the concrete foundation should be adjusted in the field by either modifying the dimensions of the shoe plate or grouting the interface between the lower plate and the concrete foundation. Neither was done so "any dimensional variations in slope and elevation would have been imposed on the bearing assembly." That is, the upper and lower plates of the bearing would not be parallel causing just a small surface area of the PTFE to absorb all of the forces. [Technical Report, Section 4.3]

The contract drawings indicated that the bearings should accommodate rotations about the horizontal and vertical axes. And the contract drawings specified rotations that were less than required by some code. The bearings were not designed to accommodate rotations. So whenever a heavy truck rolled across the bridge (Figure 7-35 above), different parts of the PTFE surface would basically get smashed.

In summary, a bad design of the bearing did not accommodate rotational forces and a bad implementation of the bearing (no field modifications to keep the upper and lower plates parallel) damaged the bearing surfaces. The increased bearing friction of the PTFE surfaces caused the shoe plate bolts to be subjected to unintended shear forces. The rotational forces that were not absorbed by the bearing caused a "non-uniform bolt force distribution." [Technical Report, Section 9, sixth bullet] This is why several bolts broke during the weeks before the cascade failure.

The second bolt problem was that they were too long and the threads were not long enough for the nut to reach the girder. This was "solved" by adding four more flat washers to each bolt. Note that this is a problem that could have been flagged by any worker or by visual inspection. This was not a structural problem. But it is an indicator that the contractor, and the government inspectors, did not care about the quality of the bridge construction. [Technical Report, Section 4.4.1]

The irony is that the extra washers may have mitigated the problem of not using a beveled washer at the bottom to compensate for any slope in the flange per the CHBDC. Since this was flagged as a problem [Section 4.4.2], I assume the flanges had an S-Section. (Having slept on it, I think that assumption is wrong. The bevel they are talking about is that the shoe plate is 52mm at one end and 60mm at the other end.) Again, this is a code violation that could have been identified during an inspection. For example, the inspector could ask to see the box of beveled washers that supplied the washers for the construction. If there were no beveled washers on sight, then the inspector would know this code had been violated.

A huge (ridiculous that it could happen and ridiculous that it was not caught) problem is that some of the nuts were not tight enough. In fact, some of them were loose! And the amount of tightness varied significantly. So the tight bolts would be the first to fail. The picture below shows the inspector should have been suspicious because of the variable number of threads showing that was not commensurate with the thickness of the shoe plate. And the inspector could have easily determined the seriousness of the situation if he was carrying a torque wrench. In fact, given that some of the bolts were loose, using just a cheap end wrench would have flagged this very serious issue. [Section 4.4]
[Technical Report, p10]
A lot of research has been performed on how to properly tighten high-strength bolts so that they are 70% preloaded to their minimum tensile strength. Not preloading the bolts not only allowed the flange to separate from the shoe plate, it allowed a significant cyclic load of the bolts. "If a 10ksi load is applied to a bolt that is preloaded to 100ksi, then the change is 10%. But if it is applied to a bolt with 10ksi preloading, then the change is 100%. Heavy trucks crossing a bridge is a cyclic load on the components of that bridge." [DeBruler] But an inspector could have used a torque wrench to at least verify that significant torque had been applied. (Torque alone is not an accurate indicator of the preloading of the bolt, but a properly installed bolt has a lot of torque.) Some of the nuts on the other girder were so loose that they needed a one-fourth turn just to become snug!

Taking another look at the shoe plate after failure, note that the middle of the shoe plate was bent up so much that it cause a plastic deformation of the plate. (If the plate had stayed in its elastic range, it would not have left a gap in the middle.) To compound the problem of loose flange bolts, the strength of the steel specified for the shoe plate on the fabrication drawing was less than the strength on the contract drawing. [Technical Report, p31, first bullet]
Technical Report, Page ii
Trucks crossing the bridge causing the plate to bend applied a bending force to the outer bolts causing another cyclic load.

Cyclic loads cause metal fatigue. Inspection of the bolts after failure showed axial bending metal fatigue. "Overload permits issued by MTO indicate that just under ninety trucks over 60 tonnes (the weight of a code design truck) potentially crossed the bridge. The number of heavy trucks crossing the bridge is similar in magnitude to the number of crack propagation cycles seen in the striations of the fracture surface of some of the bolts." [Technical Report, p87]

This is the temporary repair that was completed February 26, 2016 that allowed both lanes to be opened and allowed trucks to use the bridge again.
MTO, p1
The original bearing design was a variation on what has been used at the end of truss spans for over a century. The radical change the permanent design below indicates that cable stay bridges need a different design to resist significant uplift forces while providing more rotation flexibility than a truss needs. "This concept has been used on other cable-stayed bridges, including ones in Quebec and Kentucky." [MTO] Since it was their first cable-stay bridge, someone in Ontario should have consulted with others who have already designed cable-stay bridges before the bridge broke, rather than after it broke.
MTO, p2

Given the number of problems found during design and construction, you have to wonder about the competency of the various government agencies charged with inspecting the design and the  construction. This article asks why wasn't the lack of effective oversight also investigated. Building code departments are supposed to do more than collect fees and bribes.  (Another competency article)

Ted Gregory shared
Eastbound CP crosses the Nipigon River a few weeks ago. Behind it is the brand new (2015) Trans-Canada Highway bridge. But only 2 lanes are open, as the other half of the bridge is still under construction. I Believe the second structure has been delayed after the pictured highway structure was shutdown. On January 16, 2016, winter storm gales snapped an expansion joint that led to one half of the bridge being raised 24". Not 2.4, 24. Last I heard, no formal determination of cause has been made.
Demolition of the old bridge.
(new window)

No comments:

Post a Comment