AI-enhanced design of fiber-concrete for novel functionality
Project Description
By micromechanics-based engineering, fiber-reinforced cementitious composites (FRCC) can engage novel functionalities beyond mechanical strength, e.g., self-healing, self-sensing, CO2-sequestration. However, transforming the micromechanical behaviors revealed in laboratory into design parameters in real engineering is too complicated for conventional analytical or numerical methods. In this project, we will employ multiple machine learning approaches (from basic models to transformer deep-learning models) to train the lab data, and thus to predict and optimize the FRCC properties, including cracking resistance, fatigue resistance, and self-healing ability.
Supervisor
QIU Jishen
Quota
2
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
The applicant, if accepted, will work with a PhD student or postdoctoral fellow who is familiar with the topic. The applicant will be responsible for two types of jobs: (a) assisting experimental work that creates data for the AI training purpose, the experiment will include but not limited to single-fiber pullout test, self-healing test; (b) assisting organizing the data for training.
Applicant's Learning Objectives
In principle, we hope the applicant can work with us for at least 2 UROP projects (e.g., 1100 in Spring and 2100 in Summer). After the training, he or she will be able to (1) plan and execute mechanical tests of engineering material in the lab (even the relatively smaller and more challenging ones); (2) analyze and organize raw data for different machine learning models (which requires him/her to understand the basic mechanism and operation method of the models); (3) acquire general knowledge and basic practical know-hows on employing ML in construction material design.
Complexity of the project
Moderate