Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Engineering Science


Civil and Environmental Engineering


Nehdi, Moncef


With increasing accidental and intentional explosions and blast events inflicting life loss and economic damage to civil infrastructure, greater attention is given to the analysis and design of blast-resistant structures. Accordingly, this thesis introduces state-of-the-art machine learning models dedicated to predicting the structural behavior of various reinforced concrete (RC) members under blast loading, including slabs, columns, and beams. Moreover, extended prediction models were developed for RC members that employ fiber-reinforced polymer (FRP) retrofitting and steel fiber-reinforced concrete as blast mitigation strategies. For each model, extensive validation was conducted through statistical performance measures and comparisons to existing prediction methods. Additionally, feature importance analyses were performed to investigate the extent to which each proposed model captured its respective application. Overall, the developed prediction models achieved accurate and computationally efficient performance for the complex application of blast-loaded structures.

Summary for Lay Audience

In the event of an accidental or intentional explosion, reinforced concrete structures are highly susceptible to structural damage that may lead to severe consequences for both the structure and its occupants. Therefore, appropriate analysis and design considerations should be adopted to provide a desired level of protection. A part of this procedure is to accurately predict the response of structural members to different blast loading scenarios. Current simplified response prediction approaches are laborious and produce limited responses, whereas more detailed approaches require competent skills in finite element modeling and are computationally intensive.

To expand the state-of-the-art in predictive modeling for structures under blast loading, this thesis explores the use of machine learning methods towards developing more simplified and flexible approaches. Throughout the thesis, structural behavior prediction models were developed for reinforced concrete (RC) members including slabs, columns, and beams exposed to blast loading. The performance of each model was thoroughly investigated and found to be competitive with existing approaches. The use of machine learning for developing behavior prediction models was also extended to complex members which considered strategies for mitigating blast-induced damage. These include RC slabs with fiber-reinforced polymer surface retrofits and RC beams designed incorporating steel fibers. The resulting extensions showed that the adoption of ML methods was highly effective in considering exceedingly complicated design considerations. Overall, the proposed models throughout this thesis provided a simplified, accurate, and time-efficient approach for structural blast applications. With the expressed convenience and applicability of these models, further future developments of these models are encouraged.