Long bolted HSS-to-HSS connection for modular structures: a solution for indigenous housing challenges

Abstract

Remote Indigenous communities across Canada are experiencing severe housing shortages and widespread repair needs, both stemming from engineering challenges encountered during construction. To address these issues, prefabricated steel moment-resisting frame modular houses are increasingly being adopted as alternatives to conventional housing. These structures offer several technical advantages, including enhanced quality control, lightweight components, and ease of transportation to remote and rural regions. This thesis explores the application of hollow structural sections (HSS) in panelized steel modular structures by introducing an innovative steel-bolted connection for joining HSS beams to HSS columns using high-strength long bolts as an efficient solution to the Indigenous housing challenge. The proposed connection is designed for easy installation without requiring skilled labour or heavy machinery. This thesis utilizes experimental testing, numerical analysis, and machine learning techniques to comprehensively address the overarching research problem through four integrated studies. In the first study, specimens were fabricated and tested under monotonic loading, with a Digital Image Correlation (DIC) camera used to capture full-field 3D displacement and deformation until failure. This research studies the structural performance of various geometric parameters, including bolt arrangement, extended plate thickness, number of bolts, and the presence of stiffeners in terms of joint stiffness, ductility, ultimate capacity, and failure modes. The second study focused on key connection parameters, including extended plate thickness, bolt arrangement, bolt diameter, and the number of bolts. Results highlighted the critical role of bolt configuration and quantity in maintaining the "strong column weak beam" principle, ensuring plastic hinging at the beam. The third study evaluated the effectiveness of three reinforcement techniques: (1) a stiffener below the extended plate, (2) concrete infill within the column, and (3) a combination of both. These techniques significantly impacted the structural performance of different bolt configurations, preventing premature failures such as extended plate rupture and local column buckling and reducing the ultimate rotation of the connection. Failure mode charts were developed to predict failure modes for both unstiffened and stiffened configurations to streamline the design process. These charts eliminate the need for several iterations of full design analysis, reducing design processing time and trials. In the fourth study, machine learning techniques, including genetic regression, decision trees, and neural networks, were applied to predict the ultimate moment capacity and failure modes of the steel bolted connection under monotonic loading. A dataset of unstiffened configurations from the first study was used to train and test these models, demonstrating high predictive accuracy. Additionally, genetic regression was employed to develop mathematical formulas to predict the ultimate capacity of the connection. These models underscore their potential as reliable computational tools to complement both experimental and analytical approaches.