Abstract:

In order to reasonably conduct a comprehensive evaluation of asphalt pavement performance， a PCA-PSO-SVM evaluation model based on principal component analysis and particle swarm optimization support vector machine is proposed to address the problems of the strong subjectivity of traditional pavement performance evaluation methods and the defects of existing models. The evaluation indicators are dimensionally reduced by Principal Component Analysis （PCA） to form mutually independent principal components. The key parameters of the Support Vector Machine （SVM）—the penalty coefficient C and the kernel function parameter g are optimized using the global search advantage of the Particle Swarm Algorithm （PSO） to improve model accuracy. Finally， 170 maintenance sections of a highway in the southwest region were used as an example to evaluate the pavement performance comprehensively using the PCA-PSO-SVM model and“Highway performance assessment standards”respectively. The results show that the PCA-PSO-SVM model overcomes the drawbacks of relying on empirical methods of determining parameters， improves the identification accuracy and makes the evaluation results more in line with the actual road conditions.

Key words: road engineering, asphalt pavement performance evaluation, principal component analysis, particle swarm algorithm, support vector machine