University of Twente Student Theses

Login

Do we build what we design? Using data-driven approaches to couple the construction process into air void ratio regression

Castro Illusanguil, B.D. (2023) Do we build what we design? Using data-driven approaches to couple the construction process into air void ratio regression.

[img] PDF
1MB
Abstract:A high percentage of the roads are paved with asphalt in the Netherlands. The construction process of asphalt pavement has many activities, such as paving, compaction, marking and striping, etc., but the most important process is the compaction because it can affect the asphalt quality. In the conventional quality control scheme in road construction industry, to evaluate if the compaction process is effective, the verification of various pavement mechanical properties will be conducted, focusing on factors such as density, air void ratio, permeability, stiffness, etc. Among these mechanical properties, the air void ratio plays an important role in providing the required structural durability, particularly for the porous asphalt, which is currently the primarily utilized material for the surface layer of the Dutch highway network. More specifically, if the achieved air void ratio is really low or really high compared to the target values in the design phase, the quality of the asphalt pavement will be affected significantly and more prone to grow structural failures, such as crackings, ravelings, etc., in the operation phase. During the design phase of the road construction project, the target value of the air void ratio will also be determined as the functional requirement to verify whether or not the construction process implemented can result in ideal quality. However, nowadays, there is no effective and efficient measuring methods that can accurately measure the asphalt air void ratio covering the entire pavement without destroying the pavement. Thereby this research project aims to develop an empirical model that can provide an accurate prediction of pavement air void ratio based on construction characteristics. To develop such a predictive model for the air void ratio, a data-driven approach was proposed. Firstly, a literature study was performed to explore which factors will affect the air void ratio of pavement. Subsequently, existing asphalt air void ratio predictive models from previous studies have been investigated regarding the performance, considered input parameters, and regression methods. One model was selected as the baseline model for the follow-up research, given its adequate performance. Because this model did not sufficiently include the two most relevant parameters, namely the deviation of roller passes to the target passes and the percentage of roller passes within the temperature window, which can comprehensively represent the significance of the construction process quality, an improvement was made by integrating these two input parameters into the original input-output structure used by the baseline model. In addition, to address the nature of the system’s complexity and non-linearity, another improvement has been made regarding the utilized regression method. Thus, two improved models were built considering with the updated input-output structure, using linear regression (as utilized by the baseline model) and Random Forest (RF), respectively. To validate the models, a case study was performed, which provided construction and verification result data from a highway construction project on A50 in the Netherlands. Subsequently, the predictive performances of the two models were compared based on R-squares. The comparison shows the empirical model using linear regression failed to provide desirable results, with R2 of 0.0076, while the obtained RF model achieved a satisfying predictive r-squared of 0.84. This has shown that the RF model can predict reliable air void percentages and help to have a better overview of how the air void ratio on the road will look when it is being paved. Lastly, the air void ratio predictions and other information will be visualized in GIS as the extension to the construction process quality assessment scheme of ASPARi.
Item Type:Essay (Bachelor)
Faculty:ET: Engineering Technology
Programme:Civil Engineering BSc (56952)
Link to this item:https://purl.utwente.nl/essays/95023
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page