PhD Oral Exam - Mehdi Nikoo, Civil Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

The demand for sustainable construction has seen a corresponding rise in the utilization of cross-laminated timber (CLT) as a structural material on the strength and environmental advantages it bears. Optimizing structural design and making it more reliable requires precise estimation of rolling shear strength in CLT. This study presents an integration of experimental testing with finite element modeling as well as predictive techniques based on artificial intelligence in order to improve the estimation process of the rolling shear strength and flexural performance of the CLT panels.

Research was conducted in two distinct phases. In Phase I, information on the rolling shear strength of 26 CLT samples was produced from a Modified Planar Shear (MPS) Tests which was performed on samples produced by a Canadian company and tested in Structural lab in Concordia University. The data set encompasses experimental data for 3-layer, 4-layer, 5-layer, and 7-layer specimens, which each provide general information on the shear response. As another goal, machine learning algorithms were designed for enhancing predictive equation models. For estimating the rolling stress, the Moth-Flame Optimization (MFO) Algorithm with an Artificial Neural Network in a Feed-Forward model was utilized. To aid the regression modeling, the MFO and the Grey Wolf Optimizer (GWO) algorithms included variables such as the modulus of elasticity (MOE), the number of layers, thickness, width, height, displacement, and grain orientation angle. Optimization methods include GWO as one of the methods which outperformed others in the adjustment of regression equations. It modified the coefficients in order to make a solution with less error. A comprehensive analysis of the performance of the developed models revealed that contrary to popular assumption, rolling shear strength was predicted with more precision using the MFO-ANN model than straightforward regression methods.

Besides strength prediction, the aforementioned experimental load–displacement curves were employed in post processing an invariant set of mechanical performance indicators that encompasses stiffness evolution, deformation capacity, and also energy dissipation. Some other indicators included stiffness-related measures, strength and deformation metrics and energy-based parameters. The indicators of choice were then min–max normalized to provide fair comparison among samples with different layer configurations. Besides, two quantifiable and data-driven indices enhanced the interpretation of damage progression in a performance-based manner which includes a Shear Response Index (SRI) with standard-deviation-based weighting and, a CRITIC-based index with respect to both parameter dispersion and inter-parameter correlation (using nonlinear extension in order to portray progressive damage behavior). In such a way, it becomes possible to concisely and quantitatively describe the processes of classification specimens into distinct damage levels, as well as the analyzed rolling shear behavior to stability, ductility and post-peak strength.

Phase II was the establishing of the correlation between the rolling shear strength and the maximum Moment in case when conducting a four-point bending test. The Finite element method (FEM) analysis was carried out in Abaqus on 3, 5 and 7-layer CLT panels having a width of 0.5 meters and a length of 3 meters. Additionally, data from 73 CLT specimens obtained from literature were incorporated. An enhanced AI model was proposed with a Genetic Algorithm (GA)-based ANN model to predict the CLT panels’ flexural performance. 70% of the data was utilized for training phase while 30% was set aside for testing phase and linear and nonlinear regression were optimized using GA by adjusting coefficients iteratively to reduce errors. Analysis for the presented model focused on panel width, span, thickness, and flexural rigidity (EI) parameters in both minor and major directions. The comparison showed that the maximum bending moment as well as the Flexural strength was improved with the GA-ANN model rather than requesting for the regression approaches.

The purpose of this study is to introduce the reader to AI-based techniques and predictive equations designed to estimate the shear behavior of CLT when subjected to experimental testing and numerical modeling. An index for evaluating damage in CLT is also introduced. The results aim at better performance based CLT design through rolling shear and flexural response enhancement which in turn amounts to mitigating the requirement for extensive large-scale testing that would be needed before.