Development of a stochastic based multidimensional matrix for the analysis of pavement performance data
Thesis DisciplineCivil Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
As pavement condition becomes an ever-growing problem within the ageing New Zealand road network, a challenge emerges to effectively analyse the ageing pavement databases to improve pavement performance. Establishing how the various factors affect pavement performance is complicated due to the random features of pavement deterioration and the complex relationships between different parameters. To address this, it is proposed that a new tool be developed that will combine critical indicators into one structure for performance comparisons. The tool takes the form of a stochastic multidimensional matrix which can deal with random features and complex relationships. The range of pavement technologies that will be compared is based on data available within the New Zealand Long-Term Pavement Performance database (LTPP). The data is collected by professionals with industry standard or better equipment for New Zealand conditions.
This research found a possible weak point in data quality. The location with respect to the wheel path of where the data was collected is estimated to the best of an engineer’s ability and not measured directly. If data was not collected in the wheel paths, allowances must be made. This research presented a new methodology to check and quantify the wheel paths distribution. Deploying this methodology on an LTPP test section showed that the estimation method employed by the NZTA was sufficient and no allowances had to be made to the data. This research also highlighted that the wheel path width is not as wide as originally anticipated for both light vehicles and heavy vehicles. This information was shown to be valuable for contractors in calibrating the variable bitumen spray bar.
Once the validity of data was established, the data structure and selection methodologies were investigated. From the literature review and discussion with experts, a multi-dimensional approach was chosen. This approach allowed for multiple levels of research to be conducted. Data could easily be analysed at a site, indicator or network level all within one structure. As the databases were large, the multi-dimensional structures would be filled with stochastic indicators rather than storing the entire population. This allowed for two key advantages; firstly, it allowed the structure to remain small and easily manipulated. Secondly, it allows most computational power to be conducted up front. Therefore, allowing researchers to establish trends much more quickly by simply examining the multi-dimensional structure in different dimensions.
The comparison of different indicators to identify sections of pavement that are performing well was the next objective. This involved the featurization of pavement data through the use of fuzzy logic and combining the featurization data with expert weights. This allowed different sections of pavement to be ranked and establish which pavement sections were performing well. This research presented a new method of establishing fuzzy memberships functions based on data and not on expert opinion. This research established a new tool called The Stochastic Based Multi-dimensional Matrix (SBMDM).
This research will present two examples of how the SBMDM was demonstrated through case studies. These case studies investigate pavement performance for a specific location and investigate the SBMDM at a network level. After interviewing experts in New Zealand through the implementation of the Delphi method, it became apparent that rutting is the most important pavement performance indicator for New Zealand roads. By adopting this point and utilising the SBMDM, an in-depth study was completed on LTPP sites in the Canterbury region. Results show that there is a significant difference between the LWP(outside) and RWP(inside) rutting. This research reasoned that the camber or cross fall of the roads surface, caused an uneven distribution of load, resulting in the observed results.
The second study used the SBMDM to analyse rutting from a network level. The results show that there is a significant difference in the amount of rutting in the inside wheel path compared to the outside wheel path. Using deterioration models developed in New Zealand, it was shown that the models matched the trend seen at the network level. From this result, it can be reasoned that there is a deterioration cost due to camber.
The research includes a comprehensive literature review. Each chapter will include further detailed literature as it relates to a specific topic. The scope, objectives, methodology, results, recommendations, and conclusions of the research are also detailed.