基于粒子群优化-熵权无偏灰色马尔可夫模型的沥青道面性能预测

Performance prediction of asphalt pavement based on PSO-entropyweighted unbiased grey Markov model

为提高沥青道面使用性能的预测精度,针对道面性能数据累积年限少、波动大的特点,采用熵值法对多个道面单元进行赋权,并利用粒子群算法(PSO)和马尔可夫模型对传统无偏灰色模型残差序列的状态区间和白化系数进行优化,构建可准确预测沥青道面性能的PSO-熵权无偏灰色马尔可夫模型.结合西北某机场沥青道面6个检测单元前7年的道面状况指数进行模型精度检验,结果表明:与传统无偏灰色模型相比,经马尔可夫模型划分残差序列空间并使用PSO算法寻找最佳白化函数后,优化模型第1年~第5年各单元残差和相对误差均减小,且整体预测精度分别提高0.05%、0.28%、0.05%、0.03%和0.14%,第6、7年的整体预测精度分别提高12.9%和19.2%,说明优化后的模型对实际沥青道面的预测结果更具有效性和针对性.

To improve the prediction accuracy of asphalt pavement performance,multiple pavement units were weighted by the entropy value for the characteristics of pavement performance data with few years of accumulation and large fluctuations.The particle swarm optimization(PSO)algorithm and Markov model were used to optimize the state intervals and whitening coefficients of the residual sequences of the conventional unbiased grey model.The PSO-entropy weighted unbiased grey Markov model for asphalt pavement performance prediction was constructed.The model accuracy was tested by combining the previous 7 years pavement condition index from 6 inspection units of asphalt pavement at an airport in Northwest China.The results show that compared with the traditional unbiased grey model,the residuals and annual relative errors of each unit in the first five years of the optimized model decrease using the Markov model to divide the residual series space and applying the PSO algorithm to find the best whitening function.The overall forecast accuracies of the 1st year to the 5th year increase by 0.05%,0.28%,0.05%,0.03%and 0.14%,respectively,and those of the sixth and seventh years increase by 12.9%and 19.2%,respectively.The optimized model is more effective and relevant in predicting the actual asphalt pavement.