Prediction of rigid pavement roughness in rural Myanmar using artificial neural networks (ANNS) | |
| Author | May Mon Phyo |
| Call Number | AIT Thesis no.TE-26-03 |
| Subject(s) | Rural roads--Technological innovations--Myanmar Pavement Management Systems--Myanmar Artificial intelligence |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Transportation Engineering |
| Publisher | Asian Institute of Technology |
| Abstract | The road transportation network is crucial for socio-economic development and accessibility in rural areas. However, rural road networks are often loosely connected, and collecting comprehensive data is time-consuming and costly. In low-volume rural roads (LVRRs), pavement deterioration is more influenced by environmental conditions than traffic, particularly in Myanmar, where significant seasonal variations in temperature and precipitation exist across different climate zones. Therefore, accurate prediction of pavement roughness is critical for effective maintenance planning under data limited environments. The aim of this study is to develop a climate zone-based Artificial Neural Network (ANN) model for predicting the International Roughness Index (IRI) of rigid pavements in rural Myanmar using readily available data and to compare its performance with traditional regression methods. Data were collected from selected rural road sections under the Department of Rural Road Development, combining historical records and field data. IRI values were measured using the RoadLab Pro smartphone application. Key input variables include concrete paved width, concrete slab thickness, subbase layer presence, subbase thickness, hard shoulder width, earth shoulder presence, pavement age, rainfall, max temperature, and ADTT. Model performance was assessed based on R2 , RMSE and MAPE. The results show that the ANN models outperform the traditional regression approaches, indicating better capability of capturing complex nonlinear relationship. The best results were obtained for moderate-complexity ANN architectures with two hidden layers, such as 64–32 and 128–64 neurons. The Adam optimizer showed consistently better results, while ReLU and Tanh activation functions performed differently depending on dataset size and regional conditions. Permutation Feature Importance analysis identified hard shoulder width as the most influential variable. The study overall demonstrates the effectiveness of ANN as a practical and scalable tool to enhance pavement maintenance planning and support sustainable rural road management in Myanmar. |
| Year | 2026 |
| Type | Thesis |
| School | Faculty of Civil and Environmental Engineering (2026) |
| Department | Other Field of Studies (No Department) |
| Academic Program/FoS | Transportation Engineering (TRE) |
| Chairperson(s) | Kunnawee Kanitpong |
| Examination Committee(s) | Bhatt, Ayushman;Ampol Karoonsoontawong |
| Scholarship Donor(s) | AIT Scholarship |
| Degree | Thesis (M.Eng.) - Asian Institute of Technology, 2026 |