基于人工神经网络的路面使用性能预测
|
摘要
在我国公路养护管理过程中,传统的经验决策仍占据重要地位,已有的使用性能预测模型方法明确简单,但精度低,局限性大,不可避免地造成养护维修决策的盲目性和片面性。而许多先建的重要道路的使用性能已经衰减至低谷,建立在传统理论模式下的路面管理系统迫切需要进一步完善以适应现代社会对交通的需求。随着路面检测技术的不断进步,路面性能的历史数据已经积累到一定程度,研究开发一种能够使用海量数据和兼顾多种因素的复杂的路面使用性能预测系统成为可能。随着计算机技术和人工智能技术的发展,人工神经网络(ANN)、遗传算法等理论开始用于路面使用性能衰变规律方面的研究。
人工神经网络(ANN)因其优越性能而广泛地应用于各个学科的研究领域。其中以BP网络(Back-Propagation Neural Network)的应用技术最为显著,它超强的非线性映射能力能够逼近任意函数,为解决非线性问题提供了有力的途径,可以模拟复杂的非线性动态系统或过程,实现系统的模式识别等。它能够从不同角度对不同的路面性能对象、应用不同的方法建立相应的预估模型,在一定程度上克服了传统方法的固有缺陷,能够较为客观地反映路面使用性能衰减规律。
国内对沥青路面使用性能的评价有若干指标,这些指标的发展趋势将直接影响路面使用性能的优劣。本文在MATLAB软件平台上,利用神经网络工具箱,结合道路使用性能的评价方法建立了基于BP网络的预测模型,结合整体路网内路面使用状况的历年实测数据,选取了若干路面使用性能指标进行了预测分析,为提高预测精度修正了预测模型,并采用修正后的模型对实测数据进行了预测分析,为公路管养部门提供了养护建议。
The traditional experience still plays an important role in highway maintenance management in our country. The old performance prediction model are clear and simple, but with low accuracy and limitations, maintenance decision-making will be blindness and one-sided. As the pavement performance of many important roads has been attenuated to a low ebb, Pavement Management System based on the traditional models need to be improved in order to adapt to modern society's transport needs. With the development of road detection technology, pavement performance data has accumulated to a certain extent. It is possible to develop a pavement performance prediction system which uses a huge amount of data and considers a variety of factors. With the development of computer technology and artificial intelligence technology, artificial neural network (ANN), genetic algorithms are used in studies of the decay law of pavement performance.
Artificial Neural Network (ANN) is widely used in various research disciplines because of its superior performance. BP network (Back-Propagation Neural Network)‘s application is the most significant. Its non-linear mapping ability is able to approximate any function to solve nonlinear problems. BP net work can simulate complex non-linear dynamic system or process to achieve the pattern recognition. It can establish different prediction models from different angles for objects. BP net work can overcome the inherent shortcomings of traditional models and reflect the attenuation law of pavement performance more objectively.
Domestic asphalt pavement performance evaluation has a number of indicators that can directly affect the pavement performance. In this paper, prediction models based on BP neural network are established by neural network toolbox on the MATLAB software platform. Combining with factual data of the whole road network conditions, some indicators are selected to predictive analyze. The prediction model is amended to improve the prediction accuracy. Using the revised prediction model, this paper analyzes the prediction results and gives a piece of advice to highway department.
人工神经网络(ANN)因其优越性能而广泛地应用于各个学科的研究领域。其中以BP网络(Back-Propagation Neural Network)的应用技术最为显著,它超强的非线性映射能力能够逼近任意函数,为解决非线性问题提供了有力的途径,可以模拟复杂的非线性动态系统或过程,实现系统的模式识别等。它能够从不同角度对不同的路面性能对象、应用不同的方法建立相应的预估模型,在一定程度上克服了传统方法的固有缺陷,能够较为客观地反映路面使用性能衰减规律。
国内对沥青路面使用性能的评价有若干指标,这些指标的发展趋势将直接影响路面使用性能的优劣。本文在MATLAB软件平台上,利用神经网络工具箱,结合道路使用性能的评价方法建立了基于BP网络的预测模型,结合整体路网内路面使用状况的历年实测数据,选取了若干路面使用性能指标进行了预测分析,为提高预测精度修正了预测模型,并采用修正后的模型对实测数据进行了预测分析,为公路管养部门提供了养护建议。
The traditional experience still plays an important role in highway maintenance management in our country. The old performance prediction model are clear and simple, but with low accuracy and limitations, maintenance decision-making will be blindness and one-sided. As the pavement performance of many important roads has been attenuated to a low ebb, Pavement Management System based on the traditional models need to be improved in order to adapt to modern society's transport needs. With the development of road detection technology, pavement performance data has accumulated to a certain extent. It is possible to develop a pavement performance prediction system which uses a huge amount of data and considers a variety of factors. With the development of computer technology and artificial intelligence technology, artificial neural network (ANN), genetic algorithms are used in studies of the decay law of pavement performance.
Artificial Neural Network (ANN) is widely used in various research disciplines because of its superior performance. BP network (Back-Propagation Neural Network)‘s application is the most significant. Its non-linear mapping ability is able to approximate any function to solve nonlinear problems. BP net work can simulate complex non-linear dynamic system or process to achieve the pattern recognition. It can establish different prediction models from different angles for objects. BP net work can overcome the inherent shortcomings of traditional models and reflect the attenuation law of pavement performance more objectively.
Domestic asphalt pavement performance evaluation has a number of indicators that can directly affect the pavement performance. In this paper, prediction models based on BP neural network are established by neural network toolbox on the MATLAB software platform. Combining with factual data of the whole road network conditions, some indicators are selected to predictive analyze. The prediction model is amended to improve the prediction accuracy. Using the revised prediction model, this paper analyzes the prediction results and gives a piece of advice to highway department.
引文
1李盛霖部长在全国公路养护管理工作会议上的讲话. http://www.mot.gov.cn/zhuzhan/buzhangwangye/lishenglin/zhongyaojianghua/200709/t20070926_408886.html
2潘玉利.路面管理系统原理.人民交通出版社. 1998:97~102
3姚祖康.路面管理系统.人民交通出版社. 1993:55~68
4 Hadley W. O, Copeland C. and Rowshan S.Strategic Highway Research Program-Long-Term Pavement Performance Information Management System, Transportation Research Record 1435,TRB,National Academy of Science, Washington D. C ,U.S.A.,1994
5 W.O.Hadley. SHRP-LTPP Overview: Five~Year RePort.SHRP-P-416,1994
6 AASHTO, Pavement Management Guide: Executive Summary Report. America Association of State Highway and Transportation Officials, 2001
7 AASHTO guidelines for pavement management systems, America Association of State Highway and Transportation officials, 444 North Capital Street, N.W.Suite 225, Washington, D.C.20001, USA, 1990
8 Mustafa Birkan. BAYRAK.USE OF ARTIFICIAL NEURAL NETWORKS FORPREDICT-ING RIGID PAVEMENT ROUGHNESS. Midwest Transportation Consortium, Fall Student Conference, Ames, Iowa, November 19,2004
9 J.D.Lea. Neural network for performance prediction on unsealed roads. Road & Transportant Research. Mar 1999
10 Florida Department of Transportation. Road surface crack condition forecasting neural network models. Summary of Final Report, WPI# 0510816. October 1999
111 J.J.Lu, J.Yang, and M.Gunaratne. Pavement surface performance forecasting by neuron network models. Department of Civil & Environmental Engineering University of South Florida
12孙立军,姚祖康.北京市公路路面结构状况的预估分析.同济大学学报. 1991,19: (4):455~463
13许志军,孙立军.沥青路面结构组合对使用性能的影响.同济大学学报. 1996, 24(5):515~519
14李家绪.浅析国内外路面结构性能预测模型.交通科技. Apr.2005:No.2
15曾沛霖等.路面管理系统的研究开发与推广应用.公路交通科技. 1993, 10(2):1~7.
16孙立军,刘喜平.路面使用性能的标准衰变方程.同济大学学报. 1995,23(5): 512~518
17杨文山.路面状况指数的BP网络预测.哈尔滨建筑大学学报. 2000,33(3):107~110
18周文献,李明利,孙立军.基于改进神经网络的水泥路面使用性能预测模型.同济大学学报(自然科学版). 2006,Vol.34 No.9
19周文献等.水泥混凝土路面调查及网级使用性能预测模型.同济大学学报(自然科学版). 2006,Vol.34 No.5:624~628
20于江霞等.基于神经网络的公路网规模预测.长安大学学报(自然科学版), 2006,Vo1.26 No.1:75~78
21谢旭飞.基于BP网络的路面使用性能预测与影响因素分析.哈尔滨工业大学硕士学位论文. 2004:22~27
22蒋宗礼.人工神经网络导论.高等教育出版社. 2001
23焦李成.神经网络系统理论.西安电子科技大学出版社. 1990:1~41
24葛哲学,孙志强.神经网络理论与MATLABR2007实现.电子工业出版社2007:2~24
25 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency. In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96, 1996:857~882
26 Hagen M.T, Menhaj M.B.Training feedforward net-works with the Levenberg Mar-quardt algorithm. IEEE Trans on Neural Networks.1994, 5(6):989～993
27 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation, IEEE Transactions on Neural Networks, 1996, 7(5):1220~1230
28闻新,周露,王丹力等. MATLAB神经网络应用设计.科学出版社. 2002:207~244
29 Marco Gori, Alberto Tesi. On the problem of local minima in back-propagation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1992, 14(1):76~85
30牛国宏.基于神经网络的交通事故预测.长安大学硕士学位论文. 2006:30~31
31沈金安等.沥青及沥青材料的路面性能.人民交通出版社. 2001
32沙庆林.高速公路沥青路面早期破坏现象及预防.人民交通出版社. 2001
33沙庆林.高等级半刚性基层沥青路面.人民交通出版社. 1997
34中华人民共和国交通部. JTJ 073.2—2001.公路沥青路面养护技术规范.人民交通出版社. 2001
35杜二鹏.基于灰色理论和神经网络的路网养护决策技术研究.同济大学博士论文. 2009
36杭飞.半刚性沥青路面交通荷载适应能力分析.哈尔滨工业大学硕士学位论文. 2007:22~23
37李德仁,王树良,李德毅.空间数据挖掘理论与应用.科学出版社. 2006
38柯郑林,样本观测值数据筛选标准的分析.科学技术与工程. Vo.l8 No. 20 Oct. 2008
39 Martin Hagan, Howard Demuth. Permission to include various problems, demonstrations, and other material from Neural Network Design, Jan.1996
40 Hagen M.T, Menhaj M.B.Training feedforward net-works with the Levenberg Mar-quardt algorithm. IEEE Trans on Neural Networks.1994, 5(6):989～993
41 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation, IEEE Transactions on Neural Networks, 1996, 7(5):1220~1230
42 Rumelhart D.E, Hinton G..E, Williams R.J. Learning representations by back-propagating errors.Nature,1986,323:533~536.
43 Hagen M.T, Menhaj M.B. Training feedforward net-works with the Levenberg Mar-quardt algorithm .IEEE Trans on Neural Networks. 1994, 5(6):989～993.
44 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation. IEEE Transactions on Neural Networks. 1996, 7(5):1220~1230.
45 Hugar L.H, Powell B.A, Kamens S.N, Adaptive networks for fault diagnosis and process control. Computer Chem Eng.1990,14(4):561~572.
46 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency. In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96,1996:857~882.
47徐春晖等.前馈型神经网络新学习算法的研究.清华大学学报. 1999, 39(3):1~3.
48高仁祥等.基于神经网络的变量选择方法.系统工程学报. 1998,13(2):32~37
49 Hugar L.H, Powell B.A, Kamens S.N, Adaptive networks for fault diagnosis and process control. Computer Chem Eng,1990,14(4):561~572
50 Monica Bianchini, Paola Frasconi, Marco Gori. Learning winth out local minima in radial basis function networks. IEEE Transactions on Neural Networks. 1995, 6(3):749~755
51吴翊,李永乐,胡军庆.应用数理统计.国防科技大学出版社, 1995:135~195
52 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency.In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96, 1996:857~882
2潘玉利.路面管理系统原理.人民交通出版社. 1998:97~102
3姚祖康.路面管理系统.人民交通出版社. 1993:55~68
4 Hadley W. O, Copeland C. and Rowshan S.Strategic Highway Research Program-Long-Term Pavement Performance Information Management System, Transportation Research Record 1435,TRB,National Academy of Science, Washington D. C ,U.S.A.,1994
5 W.O.Hadley. SHRP-LTPP Overview: Five~Year RePort.SHRP-P-416,1994
6 AASHTO, Pavement Management Guide: Executive Summary Report. America Association of State Highway and Transportation Officials, 2001
7 AASHTO guidelines for pavement management systems, America Association of State Highway and Transportation officials, 444 North Capital Street, N.W.Suite 225, Washington, D.C.20001, USA, 1990
8 Mustafa Birkan. BAYRAK.USE OF ARTIFICIAL NEURAL NETWORKS FORPREDICT-ING RIGID PAVEMENT ROUGHNESS. Midwest Transportation Consortium, Fall Student Conference, Ames, Iowa, November 19,2004
9 J.D.Lea. Neural network for performance prediction on unsealed roads. Road & Transportant Research. Mar 1999
10 Florida Department of Transportation. Road surface crack condition forecasting neural network models. Summary of Final Report, WPI# 0510816. October 1999
111 J.J.Lu, J.Yang, and M.Gunaratne. Pavement surface performance forecasting by neuron network models. Department of Civil & Environmental Engineering University of South Florida
12孙立军,姚祖康.北京市公路路面结构状况的预估分析.同济大学学报. 1991,19: (4):455~463
13许志军,孙立军.沥青路面结构组合对使用性能的影响.同济大学学报. 1996, 24(5):515~519
14李家绪.浅析国内外路面结构性能预测模型.交通科技. Apr.2005:No.2
15曾沛霖等.路面管理系统的研究开发与推广应用.公路交通科技. 1993, 10(2):1~7.
16孙立军,刘喜平.路面使用性能的标准衰变方程.同济大学学报. 1995,23(5): 512~518
17杨文山.路面状况指数的BP网络预测.哈尔滨建筑大学学报. 2000,33(3):107~110
18周文献,李明利,孙立军.基于改进神经网络的水泥路面使用性能预测模型.同济大学学报(自然科学版). 2006,Vol.34 No.9
19周文献等.水泥混凝土路面调查及网级使用性能预测模型.同济大学学报(自然科学版). 2006,Vol.34 No.5:624~628
20于江霞等.基于神经网络的公路网规模预测.长安大学学报(自然科学版), 2006,Vo1.26 No.1:75~78
21谢旭飞.基于BP网络的路面使用性能预测与影响因素分析.哈尔滨工业大学硕士学位论文. 2004:22~27
22蒋宗礼.人工神经网络导论.高等教育出版社. 2001
23焦李成.神经网络系统理论.西安电子科技大学出版社. 1990:1~41
24葛哲学,孙志强.神经网络理论与MATLABR2007实现.电子工业出版社2007:2~24
25 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency. In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96, 1996:857~882
26 Hagen M.T, Menhaj M.B.Training feedforward net-works with the Levenberg Mar-quardt algorithm. IEEE Trans on Neural Networks.1994, 5(6):989～993
27 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation, IEEE Transactions on Neural Networks, 1996, 7(5):1220~1230
28闻新,周露,王丹力等. MATLAB神经网络应用设计.科学出版社. 2002:207~244
29 Marco Gori, Alberto Tesi. On the problem of local minima in back-propagation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1992, 14(1):76~85
30牛国宏.基于神经网络的交通事故预测.长安大学硕士学位论文. 2006:30~31
31沈金安等.沥青及沥青材料的路面性能.人民交通出版社. 2001
32沙庆林.高速公路沥青路面早期破坏现象及预防.人民交通出版社. 2001
33沙庆林.高等级半刚性基层沥青路面.人民交通出版社. 1997
34中华人民共和国交通部. JTJ 073.2—2001.公路沥青路面养护技术规范.人民交通出版社. 2001
35杜二鹏.基于灰色理论和神经网络的路网养护决策技术研究.同济大学博士论文. 2009
36杭飞.半刚性沥青路面交通荷载适应能力分析.哈尔滨工业大学硕士学位论文. 2007:22~23
37李德仁,王树良,李德毅.空间数据挖掘理论与应用.科学出版社. 2006
38柯郑林,样本观测值数据筛选标准的分析.科学技术与工程. Vo.l8 No. 20 Oct. 2008
39 Martin Hagan, Howard Demuth. Permission to include various problems, demonstrations, and other material from Neural Network Design, Jan.1996
40 Hagen M.T, Menhaj M.B.Training feedforward net-works with the Levenberg Mar-quardt algorithm. IEEE Trans on Neural Networks.1994, 5(6):989～993
41 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation, IEEE Transactions on Neural Networks, 1996, 7(5):1220~1230
42 Rumelhart D.E, Hinton G..E, Williams R.J. Learning representations by back-propagating errors.Nature,1986,323:533~536.
43 Hagen M.T, Menhaj M.B. Training feedforward net-works with the Levenberg Mar-quardt algorithm .IEEE Trans on Neural Networks. 1994, 5(6):989～993.
44 Philip Chen C.L.A rapid supervised learning neural network for function interpolation and approximation. IEEE Transactions on Neural Networks. 1996, 7(5):1220~1230.
45 Hugar L.H, Powell B.A, Kamens S.N, Adaptive networks for fault diagnosis and process control. Computer Chem Eng.1990,14(4):561~572.
46 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency. In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96,1996:857~882.
47徐春晖等.前馈型神经网络新学习算法的研究.清华大学学报. 1999, 39(3):1~3.
48高仁祥等.基于神经网络的变量选择方法.系统工程学报. 1998,13(2):32~37
49 Hugar L.H, Powell B.A, Kamens S.N, Adaptive networks for fault diagnosis and process control. Computer Chem Eng,1990,14(4):561~572
50 Monica Bianchini, Paola Frasconi, Marco Gori. Learning winth out local minima in radial basis function networks. IEEE Transactions on Neural Networks. 1995, 6(3):749~755
51吴翊,李永乐,胡军庆.应用数理统计.国防科技大学出版社, 1995:135~195
52 Zhou G, Si J. Improving neural network training based on Jacobian rank deficiency.In Lecture Notes in Computer Science. Artificial Neural Networks ICANN96, 1996:857~882
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |