Balustrade Design Loads: Failures, Fatalities, Research & Global Design Practices

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

In general, global building/loading codes are in close agreement for most loading types, with variations of +/- 20% for comparable conditions. However, the loads for handrails, barriers, and guards in the United States (ASCE 7, IBC, ASTM E985 and ASTM 2358, NAAMM AMP 521-01 et.al.) do not currently reflect values for crowds and assembly areas that are in widespread use elsewhere in the world, in particular Canada (NBC), Europe (EN), the UK (BS), Brazil (ABNT) and Australasia (AS/NZS).

Comparison of the loads for typical cases, without assembly or crowd loading, have good agreement with the 50 pounds force/foot (lbf/ft) (0.73 kilonewtons/meter (kN/m) versus the 51.5 lbf/ft (0.75kN/m) distributed loads; however, for crowd and stadium loadings, ASCE 7 and IBC have no additional requirements, whereas NBC, EN, BS, ABNT and AS/NZS require 3 kN/m (~200 lbf/ft) for stadiums. Additionally, EN, BS and AS/NZS also have an intermediate level of 1.5 kN/m (~103 lbf/ft) for specific areas of assembly, but this is not in the NBC code. In total, of 45 countries in which balustrade loadings were able to be found at the time of submission: 40 countries have a maximum crowd loading of 3 kN/m (~206 lbf/ft) or greater (>4x U.S. code loading), one country (India) has a loading of 2.25 kN/m (~154 lbf/ft) (~3x the U.S. code loading) and none have a lower loading.

A historical study of U.S. documents pertaining to loading yields three main references: ASTM International ’s ASTM E985 Standard Specification for Permanent Metal Railing Systems for Buildings, the American Society of Civil Engineers’/Structural Engineering Institute’s ASCE/SEI 7 Minimum Design Loads and Associated Criteria for Buildings and Other Structures, and the International Building Code. Significantly, ASTM E985 was originally released in 1984 and its references include Australian Standard AS 1170.1 Structural design actions, Part 1: Permanent, imposed and other actions, and British Standard BS6399 Loading for buildings – Code of practice for dead and imposed loads which at the time also showed 0.75 kN/m (~51 lbf/ft), but the last technical update to E985 was in 2000, and it was withdrawn from 2012 through 2024. (ASTM E985 was recently reinstated, ostensibly as it was in 2000, for the purpose of bringing it up to date.) Crowd and assembly loadings were adopted by BS6399 in 1996, AS/NZS 1170.1 in 2002 and EN1991 Eurocode 1: Actions on structures—Part 1-1: General actions—Densities, self-weight, imposed loads for buildings in 2002. In the interim period, without maintenance of standards in the U.S., balustrade loadings have languished; thus the question becomes: “Why did they change elsewhere, and are those changes justified?”

The balustrade load categories proposed here were introduced in BS6399 – 1996. The history of its introduction is unknown to the author; however, it was possibly a reaction to the 1989 Hillsborough Disaster (UK) where 96 people died in conditions of overcrowding, following a collapse of a crowd barrier at an FA Cup semi-final soccer match. The investigation listed several other precedent collapses. Railing collapses have also been reported at a Philadelphia (PA, USA) stadium in 1998, in Maryland (USA) in 2022, and at the Public University of El Alto in Bolivia in 2021, which resulted in at least six deaths and multiple injuries. The latter item is relevant as the Bolivian loading standard for Actions on Structures (NB 1225002-1) requires 1.0 kN/m (69 lbf/ft), which is greater than the maximum balustrade loading in ASCE 7/IBC/ASTM E985.

Other collapses without serious injury included events at NFL football games. These are significant because there was not a great depth of people pressing behind the barriers.
The Hillsborogh Disaster was thoroughly investigated and is well documented by R.A. Smith from University of Sheffield in “The Hillsborough Football Disaster: Stress Analysis and Design Codes for Crush Barriers,” published in Engineering Failure Analysis (1994). In the paper, Smith quotes the investigation report by Lord Taylor, noting that similar investigations had occurred in 1924, 1946 (33 deaths), 1972 (66 deaths), and 1986 (56 deaths). Smith (writing in 1994) also noted other crowd crushing events in 1990 (Mecca, 1,426 deaths), 1991 (Shanghai, 105 deaths) and 1992 (Madras, 65 deaths). The Wikipedia page https://en.wikipedia.org/wiki/List_of_fatal_crowd_crushes#21st_century summarizes hundreds of events and thousands of deaths due to crowd crushing. While all of these do not include details of barrier collapse, they demonstrate that fatal crowd loading is sufficiently frequent to be a design consideration; we know that if a barrier collapses it is more likely to cause crowd collapse and fatalities. It is important to prevent crowd loading situations from turning into fatal ones.
In a report “Going Off the Rails” (2021), The National Center for Spectator Sports Safety and Security (NCS4 at the University of Southern Mississippi) notes a history of fatalities and serious injuries at U.S. sporting arenas, both old and new.

While the Hillsborough Disaster was attributed in part to over-crowding, post-failure analysis indicated that the railing failed at approximately 8 kN/m; thus, the proposed loads of 3 kN/m (~200 lbf/ft) for crowd loading and 1.5 kN/m (~100 lbf/ft) for assembly spaces is not excessive for reasonable design load situations. The El Alto incident reinforces that elevated rail loading exceeding 1kN/m (69 lbf/ft) is possible at assembly areas other than at stadiums, so the approach in AS/NZS and EN, which is broader than in NBC, is justified. Eurocode includes a range of 3-5 kN/m (~200 lbf/ft – 340 lbf/ft) with 3 kN/m (~200 lbf/ft) recommended. This approach highlights some circumstances in which the designer may wish to select a loading greater than 3 kN/m (~200 lbf/ft) if it is considered appropriate.

R.A. Smith includes the formulation of a “leaning crowd” model used to estimate loads generated on a barrier on a stadium with terraced seating.

Studies of crowd crushing by Fruin (1993) indicated that crush forces of up to 3430 N (766 lbs) can be applied to a single person in overcrowding situations; hence, similar loads should be anticipated at barriers that contain crowds. This is important because the collapse of a barrier can lead to people falling over an edge, or by falling down and being crushed by those that land on top of them (Fried and Grant JASM Venue Safety Strategies, 2023).

An animation at CrowdRisks.com/research.html provides a computer simulation of “crowd collapse,” a situation in which a disturbance causes one person to fall and be unable to shift and regain balance without pressing on the person next to them, who is then also pushed off balance, creating a domino effect—with increasing mass and synchronous dynamic effect and/or multi-cyclic impact as the wave passes through the crowd. Review of the video of the El Alto incident indicates that there were “pressure waves” within the crowd and that a scuffle added a dynamic component to the static load at the time of collapse.

An Australian Study by C.T. Styan, M.J. Masia, and P.W. Kleeman “Human Loadings on Handrails” in the Australian Journal of Structural Engineering takes an experimental approach to measuring the horizontal loads possible on rails and finds that the loads in the BS, EN, and AS/NZ standards are reasonable and that in a significant number of cases, exceed the loads in ASCE 7. The Australian study also documents other failures due to overloading in the introduction to their study. In the 1998 Philadelphia barrier collapse professional reports stated that the failure was due to overloading, not a design fault (relative to the design loading) and further notes that the audience was only one or two rows deep. The test simulated loadings associated with various types of occupancy and compared them with the design loading. Testing found that a single row of people could generate static loads of 1.43 kN/m (98 lbf/ft), two rows generated 2.12 kN/m (145 lbf/ft) and three rows generated 2.66 kN/m (182 lbf/ft); adding dynamic “bouncing” and a single row at the barrier generated 2.43kN/m (167 lbf/ft). In each case, test results showed the design loads were appropriate; the results also found that in circumstances subject to “unruly behavior,” such as kicking the barrier, loadings higher than the design loading were possible. The testing also confirmed a 1.5 kN/m (~100 lbf/ft) loading for an occupancy with assembly but without crowd loading.

As the occupancy categories used in the U.S. are different than in the other standards referenced, the tables in BS, AS/NZS and EN have been grouped and summarized by load. Notably, in Table 1 Category “A” there is a reduced point load relative to ASCE 7 for interior single residential usage, but a 25 lbf/ft uniform load is required where ASCE 7 would exclude a requirement. The loading in the U.S. is concentrated load based on a “grab load” and “soft body slip impact load” as investigated by ANSI; hence, reductions are not proposed, and the table is a conservative bounding of the criteria. Additionally, concentrated loads do not match the 200 lbf point load requirements in the Occupational Safety and Health Administration (OSHA) Regulations 1910.28.

The proposed design loads in Table 1 are an amalgamation of the referenced EN and AS/NZS standards and the existing precedence in ASCE 7. The load categories do not align directly between the standards; whereas ASCE 7 has concentrated loads for the top rail and the components, and a uniform load for the top rail, the AS/NZS standard has all of these, and a distributed load applied to infill panels. The EN/BS standards have a concentrated load applied to the infill only (albeit that this is similar to or greater in magnitude than the ASCE 7 concentrated loads for top rails.) The reference standards’ inclusion of distributed loads applied to infill components is incorporated in the recommendation as it is likely important for barrier walls and fences. Canada’s National Building Code has incorporated allowances of 0.5 kPa (~10 psf) for walls as barriers.

ASTM E985 continues to be referenced and was recently reinstated in its prior form so that it can be updated. As such, it is noteworthy that not only does it not recognize assembly or crowd loading, it regards the 50 lb/ft as a test load for both the balustrade system and the attachments to structure as tested in ASTM E894 Test Method for Anchorage of Permanent Metal Railing Systems and Rails for Buildings and E935 Test Methods for Performance of Permanent Metal Railing Systems and Rails for Buildings. Such testing protocols do not currently allow any variability in the materials or testing to provide a safety factor. For example, the National Design Specification (NDS) for wood recommends a safety factor of 5 for withdrawal of fasteners relative to test data to allow for system variation. For post-installed concrete anchors under static loading a factor of safety of 4 is common but is greater for dynamic loading. Post-installed concrete anchors should pass the requirements of American Concrete Institutes’ ACI 355.2 Qualification of Post-Installed Mechanical Anchors in Concrete and Commentary. As noted in a report “Going off the Rails” by The National Center for Spectator Sports Safety and Security, where railings fail, failures at the anchors are the most common cause. The Australian Standard 1170.0 Structural design actions, Part 0: General Principles Appendix B provides proof load testing factors based on sample size and coefficients of variation. Following ASTM methods, the combination of lack of crowd loading and safety factor for testing results in metal railing systems attached to a wood structure, tested and approved to ASTM E985 and E935, are one fifteenth of the loads prescribed by AS1170.1 and testing to AS1170.0 Appendix B . Even for the current ASCE 7 design loads, the lack of a safety factor in testing means that systems approved by testing potentially have a significantly lower strength than systems justified by calculation to the relevant materials standards.

The proposed loads in Table 1 are reasonable and validated. The reference in Eurocode EN 1991 indicates a range of crowd loads between 3 kN/m (~ 200 lb/ft) and 5 kN/m (~ 340 lb/ft) and the post-failure analysis of the Hillsborough barriers indicated failure at ~8 kN/m (~ 550 lbf/ft); however, the latter was in a case of extreme overcrowding which might be considered greater than a reasonable design case. Crowd loading of 3 kN/m (~ 200 lb/ft) has also been adopted by the Canadian code for stadiums.

Justification by testing to meet standards needs to incorporate suitable safety factors based on the materials they are being attached to as well as testing variations. For testing with samples of 6 or more, a proof load factor of 2.5x for steel, 4x for concrete, and 5x for wood is consistent with achieving statistical significance consistent with the respective materials’ standards and coefficient of variation.

There is a large discrepancy between balustrade loadings in the United States and most other countries. Changes elsewhere were based on multiple disasters, investigation of those events, and validation by testing. The balustrade loads similar to Table 1 have been adopted in over 40 counties. These loads have been proposed to ASTM and ASCE for future incorporation. In the interim, design professionals and project specifiers have the option to follow international best practice when selecting appropriate testing and design loads for balustrades and guards.

Richard Green is the Founding Principal of Green Facades. He has over 30 years’ experience designing and engineering facade systems and specialty structures. He has participated in building standards committees in Australia, United States, Europe and ISO with a specialization on structural use of glass in buildings. (Richard@GreenFacadesLLC)