Concrete Masonry Compressive Strength Prediction using Mechanics-based Modelling and Gaussian Process Regression with Error Evaluation based on Experimental Data

  • Author / Creator
    Liu, Wanyan
  • The compressive strength of masonry is an essential mechanical parameter considering its influence on structural design. Among different types of masonry, hollow concrete block masonry is the most commonly used one in North America. Over the past decades, various methods were developed to determine the compressive strength of hollow concrete block masonry, namely the physical prism testing method and the empirical method (i.e., unit strength method). Regarding the physical prism testing method, the variations and uncertainties in the testing programs lead to a need for a tool that can be used to understand the effects of different factors on masonry strength prediction. Meanwhile, the unit strength method adopted by North American masonry design standards/codes is based on research that is now outdated. Therefore, tools or models that can effectively achieve the goal of accurately predicting the compressive strength as well as the behaviour of hollow concrete block masonry need to be developed.
    In this study, an automated micro-nonlinear finite-element model, using one of the most popular hard-computing techniques, is first proposed and verified to simulate the behaviour of hollow concrete block masonry prisms. Different masonry design standards/codes as well as empirical models are reviewed and compared with the proposed finite-element model over the collected experimental database. After an extensive literature review, a large database is compiled based on existing experimental studies. A global variance-based sensitivity analysis for evaluating the effect of material parameters on masonry compressive strength is then conducted based on the developed finite-element model, in which Latin hypercube sampling technique and Polynomial chaos expansions technique are adopted. Six input parameters are considered. The results show that the compressive strength of concrete masonry units is the most influential parameter, while the other parameters have different levels of impact depending on the compressive strength combinations of concrete masonry units and mortar.
    Subsequently, the experimental database previously compiled is used for developing a Gaussian Process Regression model (soft-computing model) to predict the compressive strength of hollow concrete block masonry. The whole database is divided into two groups based on mortar type (e.g., Type S, Type N), and Gaussian Process Regression models are built upon each type. A parametric study based on different covariance functions is carried out and the optimal covariance functions are selected based on mortar type. A case study based on different input parameters is also conducted and three input parameters are selected for both mortar types. The same masonry design standards/codes, which were used to compare with the finite-element model, are compared with the proposed Gaussian Process Regression model. The results indicate that the proposed Gaussian Process Regression model can provide a more accurate and reliable prediction for the compressive strength of hollow concrete block masonry, with prediction error quantified. After the proposed Gaussian Process Regression model is validated through its comparison with other models, prescribed hollow concrete block masonry compressive strength values are re-evaluated. The results indicate that the current prescribed values need to be updated and new values are proposed.

  • Subjects / Keywords
  • Graduation date
    Fall 2023
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.