基于GEP的高速铁路路基沉降预测系统研究

发布时间：2018-09-01 09:30

【摘要】：路基作为无砟轨道的基础结构,在铁路轨道铺设过程中最主要的环节就是严格控制路基的工后沉降,并对其进行科学的预测。目前常用的理论计算法和实测数据分析法,存在对数据要求严格、人为因素产生误差偏大,预测精度不高等问题。针对传统沉降预测方法存在的不足,论文提出了将云适应GEP算法用于路基沉降预测中,对促进高铁路基沉降预测的发展具有重要意义。论文对路基沉降的基本理论进行研究,分析并讨论了几种常见预测模型诸如:回归分析模型、时间系列模型、灰度模型、BP神经网络模型。并在此基础上介绍了一种演化算法—基因表达式编程(GEP),并对其算法流程及实现方法进行具体分析。经过研究发现,其变异率、交叉率及数值常量对GEP算法的性能影响比较大。然而传统的算法依然采用固定的交叉率和变异率,对数值常量处理方法也比较复杂,导致存在局部收敛,使运算结果精度不高。对此本文提出了一种将云模型与GEP相结合的改进算法,称为云适应GEP算法(CAGEP),该算法分别利用云适应常数创建策略、云交叉变异算子策略和种群有效交叉策略对传统的GEP算法进行改进,利用适应度的动态调整变异和交叉概率,使GEP算法可以跳出局部最优,提高了收敛速度和准确性,从而优化了算法的搜索性能。并通过函数发现的实验进行对比,验证了云适应GEP算法的有效性。将云适应GEP算法应用到铁路路基沉降变形预测模型的构建中,利用编程语言开发基于云适应GEP的路基沉降预测模型。再利用传统的GEP算法、云适应GEP算法、BP神经网络及灰度算法分别对路基沉降变形的样本数据进行预测并对比分析,实验结果表明云适应GEP算法预测精度最高,更适用于对路基沉降进行变形预测,有很好的使用价值。
[Abstract]:As the foundation structure of ballastless track, the most important link in the course of railway track laying is to strictly control the post-construction settlement of subgrade and predict it scientifically. At present, there are some problems such as strict requirement of data, large error caused by human factors and low precision of prediction, which are commonly used in theoretical calculation method and practical data analysis method. In view of the shortcomings of traditional settlement prediction methods, this paper puts forward the application of cloud adaptive GEP algorithm in subgrade settlement prediction, which is of great significance to promote the development of high-speed railway subgrade settlement prediction. In this paper, the basic theory of subgrade settlement is studied, and several common prediction models, such as regression analysis model, time series model, gray model and BP neural network model, are analyzed and discussed. On this basis, an evolutionary algorithm-gene expression programming (GEP),) is introduced, and its algorithm flow and implementation method are analyzed in detail. It is found that the mutation rate, crossover rate and numerical constant have great influence on the performance of GEP algorithm. However, the traditional algorithm still uses the fixed crossover rate and mutation rate, and the numerical constant processing method is complex, which leads to the existence of local convergence and the low accuracy of the operation results. In this paper, an improved algorithm combining cloud model with GEP is proposed, which is called cloud adaptive GEP algorithm (CAGEP),). The cloud crossover operator strategy and population effective crossover strategy are used to improve the traditional GEP algorithm. By dynamically adjusting the mutation and crossover probability of fitness, the GEP algorithm can jump out of the local optimum and improve the convergence speed and accuracy. Thus the search performance of the algorithm is optimized. The validity of the cloud adaptive GEP algorithm is verified by the comparison of the function discovery experiments. The cloud adaptive GEP algorithm is applied to the construction of railway subgrade settlement prediction model, and the roadbed settlement prediction model based on cloud adaptive GEP is developed by programming language. Using traditional GEP algorithm, cloud adaptive GEP algorithm, BP neural network and gray scale algorithm respectively to forecast and compare the sample data of subgrade settlement deformation, the experimental results show that the cloud adaptive GEP algorithm has the highest prediction accuracy. More suitable for subgrade settlement deformation prediction, has a good use value.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U213.157;TP18

【相似文献】