Graduation Semester and Year
Spring 2025
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Civil Engineering
Department
Civil Engineering
First Advisor
Dr. Mohammad Najafi
Abstract
This dissertation addresses the challenge of accurately predicting and optimizing productivity in Cured-In-Place Pipe (CIPP) rehabilitation projects, a prevalent trenchless technology for renewing aging pipeline infrastructure. Current productivity estimation methods often rely on contractor experience and lack the precision necessary for effective project planning and resource allocation. This research proposes a novel approach leveraging the power of Artificial Neural Networks (ANNs) to develop a robust predictive model. A comprehensive dataset of 666 CIPP projects served as the foundation for model development. This dataset encompassed eleven key variables, including pipe diameter, length, number of shots, number of service reconnection installation type, resin type, manhole depth, liner thickness, host pipe material, flow type, and road type, alongside the corresponding worker hours for each project.
A methodology was employed, beginning with exploratory data analysis to understand the distribution and characteristics of each variable. Pearson correlation analysis revealed the strength and direction of relationships between input variables and worker hours, while a Random Forest Regressor algorithm facilitated feature ranking to identify the most influential predictors. Eleven ANN models were trained, validated, and tested, each systematically omitting variables based on their ranked importance, allowing for a thorough sensitivity analysis to determine the optimal model configuration. Model performance was evaluated using established metrics, including Root Mean Squared Error (RMSE), Mean Squared Error (MSE), Mean Bias Error (MBE), and the coefficient of determination (R²).
The optimized ANN model, utilizing only four key variables, pipe length, diameter, number of shots, and manhole depth, achieved remarkable accuracy in predicting worker hours, exhibiting an average error of 0.067%. This model significantly outperformed baseline models incorporating all eleven variables, demonstrating the effectiveness of feature selection and the ability of the ANN to capture the complex nonlinear relationships between project characteristics and productivity. This research contributes to the field of trenchless technology by providing a data-driven, accurate, and implementable model for optimizing CIPP productivity, offering substantial benefits for stakeholders involved in pipeline rehabilitation projects. Furthermore, the methodology presented can be adapted and applied to other trenchless technologies and infrastructure projects, advancing the state of the art in productivity prediction and optimization.
Disciplines
Civil and Environmental Engineering
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Al Khayyat, Ghassan M., "DEVELOPMENT OF A MODEL TO OPTIMIZE PRODUCTIVITY OF CIPP BY USING A NEURAL NETWORK MODEL" (2025). Civil Engineering Dissertations. 516.
https://mavmatrix.uta.edu/civilengineering_dissertations/516