基于线性混合效应模型的道面使用性能预测

Linear Mixed Effect Model for Airport Pavement Performance Prediction

摘要:

针对道面状况指数（PCI）历史资料的面板数据特征，以多项式作为预测函数，将部分参数视为随机变量，建立二类线性混合效应模型.结果表明：线性混合效应模型能够有效分解变量之间的异方差，弥补传统回归模型的不足，为多区域道面使用性能数据分析提供新的技术方法；混合效应模型能够综合利用多区域信息，改进道面个体的参数估计，不仅解决部位道面个体因数据不足而无法获得模型参数的问题，而且显著提高了道面个体的预测精度.

Abstract:

Accorling to the characteristics of the panel data of the pavement condition index(PCI) field data collected from different sections, a two linear mixed effect model was established with an extension of a three order polynomial by employing the fixed effect and random effect as random parameters to predict the airport pavement performance. The parameters were estimated by the mixed effects approach. Results indicate pavement sections’ heterogeneity may be captured not only through mixed effect but through random effect as well, which makes up traditional regression model’s deficiency and provides a more appropriate method for analyzing multi sections pavement performance data. At the same time, the effect model can use multi sections pavement information and improve individual’s estimation, therefore, it can estimate individual pavement condition which has little time series with a significant higher accuracy in predicting specific pavement conditions in comparison with ordinary least squares(OLS).