Fragility Assessment of Floating Roof Storage Tanks during Severe Rainfall Events

Abstract

In 2017, Hurricane Harvey, highlighted the vulnerability of floating roof storage tanks to severe rainfall events. Several floating roofs failed due to rainwater accumulating on them, causing the release of hundreds of tons of pollutants in the atmosphere. Despite the vulnerability of floating roofs, tools currently are lacking to evaluate their anticipated performance during rainstorms. This paper presents the development and application of fragility models to assess the vulnerability of floating roofs subjected to rainwater loads and help prevent future failures. First, a finite-element model of floating roofs and a load-updating method are presented to assess the potential for failure. By coupling the finite-element model with a statistical sampling method, fragility models were derived for two damage mechanisms: sinking of the roof, and excessive stresses due to the rainwater weight. Fragility models were developed for undamaged roofs and roofs with pre-existing damage (i.e., punctured pontoons). To allow their use for future or historic rainfall events, a framework is presented to estimate the maximum amount of rainwater standing on a roof and the probability of failure during a rainstorm. Lastly, forensic investigations of a floating roof failure that occurred during Hurricane Harvey were performed to illustrate the viability of the fragility models to understand the conditions leading to failures and to propose mitigation measures. Insights from the fragility analysis indicate that small floating roofs are more vulnerable to rainwater loads than are large roofs, whereas insights from the failure investigation reveal the importance of efficient roof drainage, as well as terrain drainage to prevent water accumulation around the storage tank, which can lead to inefficient roof drains. Results also indicate that in preparation for a rainstorm, storage tanks should be filled with product to improve floating roof drainage.