Lessons from Bowstring Truss Failures in Chicago

To view the figures and tables associated with this article, please refer to the flipbook above.

Chicago has seen what happens when bowstring roofs fail. In 1998, a bowstring roof collapsed during a fire at the Beverly Tire and Auto Center, killing two Chicago firefighters. In 2010, another collapsed at the former Sing Way Laundry, killing two firefighters and injuring 19 others. After the heavy snow of February 2021, several old buildings failed across the city, including buildings with bowstring roofs. They are also infamous in many other states, to the point that some fire departments have historically stopped firefighters from entering a building if it had a bowstring truss roof system.

Yet bowstring truss buildings remain part of Chicago’s working building stock. They are the old warehouses, garages, service buildings, bottling plants, laundries, small factories, and retail buildings that helped the city grow. Many are still occupied, and many continue to serve the same basic purpose for which they were built: providing large, open rooms without rows of interior columns.

Examples remain visible in the city. The West Loop Branch of the Chicago Public Library reused former industrial buildings and exposed the original bow-truss ceilings and skylights. Revolution Workshop, a construction training center, occupies two adjacent bow-truss warehouse buildings that originally housed soda manufacturing and bottling. Bowstring trusses clearly have deep roots in Chicago’s architectural and structural history.

A bowstring truss works most clearly as a tied arch. The curved top chord is the bow. The straight bottom chord is the string. Roof loads push down on the system. The curved top chord carries much of that load in compression. The bottom chord resists the outward thrust created by the arch shape.

That bottom chord is the key. The roof does not only push down. Because of the curved form, it also wants to push outward at the end supports. The bottom chord restrains that outward force. If the bottom chord works, the supports mainly receive vertical load. If the bottom chord fails, the truss can spread.

Bowstring trusses can also serve as continuous wall-anchorage elements. Their long spans and high axial strength allow them to transfer out-of-plane masonry wall forces into the wood roof diaphragm over a long length. Where out-of-plane seismic anchorage forces are significant, the code requires those forces to be carried through continuous elements across the diaphragm. Because a bowstring truss already spans from wall to wall, it can often serve as that element and reduce or eliminate the need for separate wood sub-diaphragms.

In many Chicago buildings, the truss ends bear on brick pilasters or steel columns. While these supports may handle vertical loads well, they may not have much reserve strength against outward thrust. Much of the older clay masonry is unreinforced. It cannot resist large flexural stresses from a spreading roof system.

The end connections are especially important. Many older bowstring trusses used U-shaped steel straps, bolts, or other hardware to connect the top and bottom chords at the ends. These parts transferred force from the compressed top chord into the bottom tie. If any of the hardware fails, the truss can lose the action that made it stable.

Modern analysis of older bowstring trusses often reveals deficiencies that were not recognized or addressed in the original design. This has been true for many of the Chicago bowstring trusses evaluated by structural engineers. Older design methods were simpler. Some original calculations used graphical methods or simplified truss assumptions. Those methods ignored the more complicated stresses and forces that a system like bowstring trusses can generate. Modern structural analysis software can better predict the behavior of bowstring trusses, and engineers should use them when dealing with these systems.

Loads have also changed. Older trusses were often checked for uniform gravity loads. Current evaluation may include unbalanced snow, ponding, and heavier HVAC equipment that has been newly installed. Furthermore, wind uplift is an important part of roof analysis, and arched roofs like bowstring trusses can experience higher uplift loads compared to traditional roofs because of their shapes. These loads can change the force pattern. A truss web member that once acted mainly in tension may be pushed into compression under an unbalanced snow load and buckle, something the original analysis may never have predicted nor considered in the design.

Wood design values have changed, as well. Earlier practice treated wood tension capacity more favorably than later experimental testing supported. In the 1980s, the In-Grade Testing Program tested full-size members and led to reduced allowable stresses for many lumber grades and load cases. The bottom chord is the most critical member of a bowstring truss, and because earlier practice overestimated its capacity, many bowstring trusses fall short of supporting modern code-prescribed loading.

In the end, a truss that was considered acceptable under its original assumptions may be much weaker when checked with current load requirements, current wood values, and modern analysis software.

A roof collapse in June of 2025 at a retail store in Chicago’s West Lawn neighborhood showed this behavior clearly. Berkey Engineering investigated the building after the collapse. A bowstring truss roof supported by masonry framed the building. Before the collapse, a bystander reportedly noticed that a first-floor pilaster was bulging and warned the workers in the building.
In a bowstring building, an outward-moving pilaster can be a symptom that the bottom chord is no longer tying the truss together. The top chord compression system is then free to push outward. The masonry pilaster, even if constructed properly, fails under these unique forces causing it to bulge outwards.

A rainstorm occurred a few days later. The roof had already partially collapsed and moved down 2 feet in the center. The added water load likely made an already weak condition worse. The roof collapsed, bringing down roof framing, ceiling finishes, and debris into the building. Fortunately, no one occupied the building when the collapse occurred.

The investigation started with field work. The team found that the building contained five bowstring trusses, one of which had collapsed due to the rainstorm. The team documented all the geometry and measured everything it could measure. For the configuration of the truss, simple survey tactics were utilized but to further analyze the behavior of the unstable masonry pilaster under construction wind suction pressures, a LIDAR scan of the wall was done and a point cloud data was developed. The CMU wall and the pilaster configuration was then analyzed in RISA-3D as plate elements. This process of observation was the first step in designing repairs for an existing structural system. Because the roofing system was older, the team collected as much data as possible.

The team then modeled the truss in RISA-3D. The model compared the observed roof shape with the expected structural behavior. The question was simple: did the deformation make sense? It did. The outward movement of the supports matched the behavior expected when the bottom chord and its connections could not provide enough tension resistance.

The field observations and model pointed to the same main issue. The bottom chord system had lost the ability to act as the tie. The bottom chord splice capacity was low. Wood splitting reduced the capacity near connection regions. Heavy point loads due to modern HVAC equipment near the top of the truss increased demand on a system that already had limited capacity. The truss no longer behaved as a stable tied arch and had collapsed.

The masonry damage was therefore not separate from the roof damage. The pilaster moved because the roof pushed it. The roof pushed it because the tie action was weak.

The repair strategy had to do four things.

The first step was shoring. Since the bowstring trusses were extremely overstressed, the team needed to prevent another catastrophic failure or the existing collapse from getting worse. The team achieved this by supporting every bowstring truss with regularly spaced shoring posts.

The next step was to repair the collapsed truss. The team did not assume that the truss remained in its original position. The team set up the model to reflect the existing condition. From that point, the team determined how much each panel point needed to rise to bring the truss back toward its intended profile.

The jacking sequence had to be slow due to the statically indeterminate nature of the truss system. The team could not bring the truss back in one movement. A sudden lift could split more wood or push the masonry again. The team broke the work into many phases. The project team used a sequence of approximately 47 distinct steps so the contractor could raise the truss panel points in controlled increments over the course of two days.

Once the team brought the truss closer to its intended shape, it could repair the bottom chord system. The team removed the portion of the wood bottom chord that experienced such a brittle failure and replaced it with pressure-treated glulam.

The analysis also showed overstress in the top chords, and two web members were susceptible to buckling under full snow loading. The team considered several repair concepts, and the selected repair reinforced the top chords with bolted steel angles and added steel channels at the top chord-to-bottom chord intersections. Steel rods were then installed between the channels on opposite sides of each truss and tightened with turnbuckles. Tightening the rods introduced compression into the bottom chord, reduced the tensile demand, and restored a more reliable load path through the truss.

The final repair addressed the bulging brick pilaster at the south end of the building. The approach was simple and made efficient use of material that would already be available on the site. Steel channels were placed on both sides of the pilaster, effectively sandwiching the damaged masonry. A steel rod passed through the channels and tied the assembly together. That rod was then field welded to a steel element on the bowstring truss. Once the system was in place, the rods were tightened with turnbuckles. As the rods tightened, the channels drew the pilaster back in while relying on the truss and its connection to the structural roof sheathing to provide an extremely stiff point of support.

Bowstring trusses helped build Chicago’s industrial and commercial landscape. They gave owners long roof spans, open interiors, and economical buildings. Their strength came from a simple idea: the curved top chord carried compression, and the straight bottom chord tied the ends together.

That same idea explains their failures. When the bottom chords overstress, the truss spreads, pushing the masonry supports outward. The first warning may be a moving pilaster, a cracked wall, a sagging roof line, or a split bottom chord.

The West Lawn collapse showed this behavior clearly. A bulging pilaster appeared before the roof came down. The roof had already moved. Rain added load to a weak system. The collapse then connected all the pieces: bottom chord weakness, roof sag, point loads, masonry movement, and loss of tie action.

Good repair starts with that understanding. Stabilize the system. Measure the existing shape. Model the truss. Reinforce or replace its members. Then repair the masonry support. The repair must address the roof and the wall as one connected structural system.

For engineers working with older Chicago bowstring buildings, the central question is simple: can the truss still act as a tied arch? If the answer is uncertain, the building deserves a closer look. ■