Artificial Intelligence and Natural Language Processing for Quality Control and Management

Established engineering standards are facing challenges in adapting to the evolution of novel material properties, designing entirely new materials, and uncovering new mechanisms that transcend intuitive understanding. This research aims to investigate a new framework of domain-specific language models to automatically generate feasible engineering designs based on requirements. The work is anchored in the understanding of natural language processing and the fidelity requirements of machine-learning models for the civil and construction engineering domain. Relevant reports from the Transportation Research Information Service and standard specifications from various departments of transportation were included in the data collection. Another dataset was the summarized literature of 36 highway agencies. After the comparison of Support Vector Machines (SVM), Long Short-Term Memory network, and linear regression, the SVM algorithm was implemented in the framework to support decision-making through textual and tabular communication, and the results showed improvements in accuracy (130%) and F1 Score (65%). In this research, utilizing case-based content extraction alongside reliable statistical uncertainty estimation has shown the potential to generate valuable decision-support tools and recommendation systems for both engineers and managers. The applications encompassing textual content extraction alongside reliable statistical uncertainty estimation demonstrate the potential to create valuable decision-support tools and recommendation systems for engineers and managers.