An excavation support system distributes load through several connected elements. Losing one strut can change the demands on the members that remain. A peer-reviewed study by Lei Bian and colleagues examined that redistribution in a laboratory model with different excavation depths on its two sides. The experiment offers a view of the mechanism, alongside substantial limits on transferring its numbers to a full-scale site.
The research question and the model
Bian, Xinyang Li, Fang Tan, Huanwei Wei and Cong Liu published the study in Buildings on 14 August 2026. It asks how asymmetric excavation and successive strut loss affect retaining-wall movement, ground settlement and internal forces. The authors used a model box measuring 1,100 × 900 × 1,200 mm, with a glass face that allowed the movement of the sand to be observed.
The model represented excavation depths of 500 mm and 300 mm using retaining plates and four horizontal supports. The soil was one dry standard sand at one relative density. Instrumentation included 40 strain-gauge points, nine earth-pressure points and six displacement points. The reported programme followed excavation stages and then removal of struts numbered 1, 2 and 3 in sequence.
The supports were withdrawn manually over approximately one to two seconds. The acquisition system sampled at 1 Hz, and readings were allowed to stabilise for about ten minutes before the next stage. The results therefore describe the settled response after each change. The test did not capture the peak transient response at the moment a member was removed.
What happened to the remaining supports
Table 4 reports that removing the first support increased the force in its neighbour, support 2, to 1.48 times the preceding stage. After the second removal, support 3 carried 1.86 times its previous force. Following the third removal, the remaining support 4 carried 2.11 times the preceding value. These comparisons use a different preceding stage each time and should not be added together.
The authors also observed greater horizontal movement on the deeper excavation side. Their interpretation connects this to the different embedment and restraint of the retaining plates. Earth-pressure changes varied with position, indicating that the soil and support system redistributed load across the model. The findings illustrate why watching only the damaged member can miss changes elsewhere in the arrangement.
How reliable is the evidence?
The model’s transparent face and several types of instrument make it possible to compare movements with changes in force and pressure. Following a stated excavation and removal sequence also makes the reported mechanism easier to examine. However, this is a small, one-gravity model with a single soil condition and simplified connections. It provides a controlled observation of a particular arrangement.
Stress levels and soil behaviour in a small one-gravity box differ from those around a full-scale excavation. The retaining plates and PVC supports also simplify real walls, joints and struts. Manual withdrawal represents loss of a member’s contribution without reproducing the crushing or buckling process that may occur in practice. Groundwater and saturated-soil behaviour were outside the tested condition.
The sampling rate is especially relevant to the interpretation. A stabilized force increase says little about the highest demand during the loss event. The authors acknowledge that dynamic amplification could make a real response more adverse. Repeated tests with other failure locations, soil conditions and support arrangements would be needed to establish how consistently the pattern occurs.
There is also a reporting issue. One paragraph in the axial-force discussion assigns a force to support 3 after its stated removal, while Table 4 treats a removed member’s force ratio as undefined. Civil Insight uses the table’s stage comparisons above. That inconsistency warrants clarification before anyone relies on the detailed numerical sequence.
Implications for construction practice
Civil Insight’s interpretation is that the experiment gives project teams a reason to consider the remaining system when reviewing a damaged or altered support. The designer’s assessment needs to account for the actual wall, connections, ground conditions and construction stage. Monitoring arrangements and response procedures should relate to that design assessment and the project’s defined responsibilities.
The laboratory ratios cannot establish a universal reserve capacity or a safe removal sequence. Their practical contribution is to show how load can move to adjacent members as the support arrangement changes. For a site engineer, the corresponding task is to maintain a clear record of the installed configuration, observed condition and authorised changes so the responsible designer can assess the system that actually exists.
Original peer-reviewed paper
Bian, L., Li, X., Tan, F., Wei, H. and Liu, C. (2026). Experimental Study on the Influence of Support Failure on Remaining System Under Asymmetric Excavation. Buildings, 16(16), 3230. Published online 14 August 2026. No correction or retraction notice was identified on the publisher record when checked 6 November 2026. DOI and original study
Model test and retaining-system assembly from the reviewed study, Figure 2. Lei Bian, Xinyang Li, Fang Tan, Huanwei Wei and Cong Liu · CC BY 4.0.
