| In this study, modern statistical analytical procedures much more suitable for large databases were used to obtain definitive and useful results on the specific design and construction features that can be utilized to improve portland cement concrete (PCC) pavement performance. With data from the Long-Term Pavement Performance (LTPP) database, an analysis approach that combined pavement engineering expertise and modern statistical analysis techniques was used to develop guidelines for improved pavement design and construction. It included preliminary statistical analysis to learn about the data and to evaluate relationships between design and construction features and pavement performance. However, emphasis was placed on using multivariate data analysis techniques to identify design and construction features most important to long-term performance and to quantify their specific contributions.; Because of the large number of correlated explanatory variables in the LTPP database, principal component analysis was used to reduce dimensionality and simplify further analysis. Also, the results from previous regression analysis studies, analysis of variance, and multiple-comparison test of means were used to identify preliminary design and construction features for further study. For each pavement type and performance measure, cluster analysis was then used to partition the pavements into three distinct groups characterized by the categorical variables good, normal, and poor, representing their performance over time. Using observations on site condition, design features, and construction factors, discriminant analysis was used to develop models that explain these groupings. These models can be used to assign additional pavements to the appropriate performance group and, in effect, identify the key design and construction features that significantly affect performance.; To quantify the levels of the design and construction features that contribute to performance, the classification and regression tree (CART) procedure was used to develop tree-based models for each performance measure. Using rules determined by recursive partitioning, CART is able to determine by optimization cutoffs of quantities of key design and constructions features that contribute to specific performance levels. Using the approach described practical recommendations on the quantities of the major design and construction features that can be used by designers to increase PCC pavement life were developed. |
