Sponsor: PSR9 UTC
PI: Dr. Chun-Hsing Ho
The project presents a pavement sensing system along with a list of promising computing models that can be used to predict pavement conditions using a vehicle-based sensing technology. The project started with data acquisition obtained from the previous field data collection followed by a series of data computing using machine leaning methods to determine a promising computing algorithm. Subsequently, statistical analyses were performed to evaluate the effect of sensor placements/locations within a vehicle on the accuracy of pavement condition assessments. Based on analysis results, random forest algorithm is the best fitting machine learning algorithm than other three algorithms (Linear Regression, Support Vector Machine, and Neural Network) for the pavement condition assessment. It is also found that the pavement temperature significantly influences the number of significant points (pavement distress) provided the fact that the number of significant points decrease during cold weather condition while the number of significant points increase as the pavement temperature is getting warmer. , The Time-Series analysis indicates the number of the significant points will increase quickly in the following two years, which indicate that the pavements will be deteriorated if the maintenance and rehabilitation will not be scheduled.