改进型遗传算法在木材干燥过程建模中的应用研究
|
摘要
全球森林资源不断减少带来的环保和生态问题与人们对木材日益增长的需求构成了一对难以调和的矛盾。因此,如何有效利用有限的木材资源显得至关重要。降低木材资源消耗及提高木制品质量引起了人们的广泛关注。尤其像我国这样一个少林国家,如何更好地改善木材性能并提高其利用率已成为木材科学工作者研究的重点。木材干燥是改善木材物理性能、减少木材降等损失、提高木材利用率的重要技术措施,也是保证木制品质量的关键之一。
干燥基准模型是描述温、湿度与木材含水率之问关系的模型,此模型的建立将会实现干燥基准的数学模型化及干燥过程含水率预测,为优化干燥基准提供了有力的依据。因木材是一种多孔性渗水和吸湿的物质,且水分存在于其中的形式又是多种多样,所以木材干燥过程是复杂非线性的,使得建立理想的、符合实际的干燥模型变得很困难。
本文是基于神经网络理论和改进型遗传算法建立木材干燥模型。首先,采用BP神经网络建立木材干燥基准模型。设计了BP神经网络的结构并选取了适当的训练算法,通过实验数据进行模型的训练与验证,仿真结果表明所建立的模型有效,但该模型还存在着一些缺陷,如训练速度慢、全局搜索能力弱、网络的收敛性受初始值选择的影响较大等。
其次,采用改进型遗传算法优化已建立的BP神经网络木材干燥基准模型。云模型是李德毅院士提出的一种定性定量转换模型,它在知识表达中具有不确定中带有确定性、稳定之中又有变化的特点,体现出了自然界物种进化的基本原理。基于云模型的改进型遗传算法就是利用云模型的优良特性,结合遗传算法的基本原理,得到的一种自适应、高精度、快速随机搜索的方法。该算法能很好的避免遗传算法易陷入局部最优解和早熟收敛等问题,并为解决BP网络存在的缺陷提供了一种新方法。本文对用改进型遗传算法优化后的模型也进行仿真研究,结果表明,该模型解决了BP网络中不同初始值训练出的网络存在较大差异的问题,并且较BP网络模型具有更好的收敛速度和误差精度,进而证明了基于云模型的改进型遗传算法应用于木材干燥的可行性和优越性。
本论文的研究为木材干燥系统建模提出了一种新思路,为木材干燥系统的建立提供了一些有效的理论方法。
Problems of environmental protection and ecology causing by global forest resources reducing constantly and people's increasing demand for wood constitute a pair of irreconciable contradictions. Therefore, how to use the limited wood resources effectively is critical. Reducing consumption of wood resources and improving the quality of wood products have been aroused extensive attention. How to better the wood performance and raise its utilization ratio has become the focus of the study for wood scientists as our country has fewer forest resources. Wood drying is an important technical measure for improving the wood physical performance, reducing loss of lower grade and raising wood utilization ratio, and also one of key points for guaranteeing the quality of wood products.
Drying schedule model describes the relationship between temperature, humidity and wood moisture content. Establishment of this model will realize mathematic model of drying schedule and predict moisture content of drying process, provide effective basis for optimizing drying schedule. Because wood is a porous water seepage substance and hygroscopic material, and the presenting form of moisture in which is various, wood drying process is complex and nonlinear, which makes the establishment of the ideal and realistic drying model become very difficult.
This dissertation built wood drying models based on neural network theory and improved genetic algorithm. First, using BP neural network established wood drying schedule model. This paper designed the BP neural network structure and selected the suitable training algorithm, in addition, trained and verified the model through experimental data. The simulation result showed that the model we built is effective, but this model also has some flaws, such as low training speed, weak overall situation search ability, big influence of initial value choice on network astringency and so on.
Second, using the improved genetic algorithm optimized the BP neural network wood drying schedule model which had been established. Cloud model is a qualitative and quantitative conversion model proposed by Li Deyi. It has characteristics of uncertainty with certainty, stability with variability in knowledge representation, demonstrates basic principles of species evolution of natural. Improved genetic algorithm based on cloud model is a auto-adapted, high accuracy, fast random searching method combining the excellent characteristic of cloud model with the basic principle of genetic algorithm. This algorithm can avoid local optimal solution and premature convergence problem causing by genetic algorithms, and provide a new method for solving shortcomings in BP network.This paper did simulation study of the model which was optimized by improved genetic algorithm, and its results showed that BP network model solved problems of network discrepancy creating by different initial value and it has better convergence speed and precision comparing to BP network, which proved that genetic algorithm has feasibility and superiority in wood drying.
This article proposed an new idea for wood drying system modeling, and provided a number of effective theoretical methods for the establishment of wood drying system.
干燥基准模型是描述温、湿度与木材含水率之问关系的模型,此模型的建立将会实现干燥基准的数学模型化及干燥过程含水率预测,为优化干燥基准提供了有力的依据。因木材是一种多孔性渗水和吸湿的物质,且水分存在于其中的形式又是多种多样,所以木材干燥过程是复杂非线性的,使得建立理想的、符合实际的干燥模型变得很困难。
本文是基于神经网络理论和改进型遗传算法建立木材干燥模型。首先,采用BP神经网络建立木材干燥基准模型。设计了BP神经网络的结构并选取了适当的训练算法,通过实验数据进行模型的训练与验证,仿真结果表明所建立的模型有效,但该模型还存在着一些缺陷,如训练速度慢、全局搜索能力弱、网络的收敛性受初始值选择的影响较大等。
其次,采用改进型遗传算法优化已建立的BP神经网络木材干燥基准模型。云模型是李德毅院士提出的一种定性定量转换模型,它在知识表达中具有不确定中带有确定性、稳定之中又有变化的特点,体现出了自然界物种进化的基本原理。基于云模型的改进型遗传算法就是利用云模型的优良特性,结合遗传算法的基本原理,得到的一种自适应、高精度、快速随机搜索的方法。该算法能很好的避免遗传算法易陷入局部最优解和早熟收敛等问题,并为解决BP网络存在的缺陷提供了一种新方法。本文对用改进型遗传算法优化后的模型也进行仿真研究,结果表明,该模型解决了BP网络中不同初始值训练出的网络存在较大差异的问题,并且较BP网络模型具有更好的收敛速度和误差精度,进而证明了基于云模型的改进型遗传算法应用于木材干燥的可行性和优越性。
本论文的研究为木材干燥系统建模提出了一种新思路,为木材干燥系统的建立提供了一些有效的理论方法。
Problems of environmental protection and ecology causing by global forest resources reducing constantly and people's increasing demand for wood constitute a pair of irreconciable contradictions. Therefore, how to use the limited wood resources effectively is critical. Reducing consumption of wood resources and improving the quality of wood products have been aroused extensive attention. How to better the wood performance and raise its utilization ratio has become the focus of the study for wood scientists as our country has fewer forest resources. Wood drying is an important technical measure for improving the wood physical performance, reducing loss of lower grade and raising wood utilization ratio, and also one of key points for guaranteeing the quality of wood products.
Drying schedule model describes the relationship between temperature, humidity and wood moisture content. Establishment of this model will realize mathematic model of drying schedule and predict moisture content of drying process, provide effective basis for optimizing drying schedule. Because wood is a porous water seepage substance and hygroscopic material, and the presenting form of moisture in which is various, wood drying process is complex and nonlinear, which makes the establishment of the ideal and realistic drying model become very difficult.
This dissertation built wood drying models based on neural network theory and improved genetic algorithm. First, using BP neural network established wood drying schedule model. This paper designed the BP neural network structure and selected the suitable training algorithm, in addition, trained and verified the model through experimental data. The simulation result showed that the model we built is effective, but this model also has some flaws, such as low training speed, weak overall situation search ability, big influence of initial value choice on network astringency and so on.
Second, using the improved genetic algorithm optimized the BP neural network wood drying schedule model which had been established. Cloud model is a qualitative and quantitative conversion model proposed by Li Deyi. It has characteristics of uncertainty with certainty, stability with variability in knowledge representation, demonstrates basic principles of species evolution of natural. Improved genetic algorithm based on cloud model is a auto-adapted, high accuracy, fast random searching method combining the excellent characteristic of cloud model with the basic principle of genetic algorithm. This algorithm can avoid local optimal solution and premature convergence problem causing by genetic algorithms, and provide a new method for solving shortcomings in BP network.This paper did simulation study of the model which was optimized by improved genetic algorithm, and its results showed that BP network model solved problems of network discrepancy creating by different initial value and it has better convergence speed and precision comparing to BP network, which proved that genetic algorithm has feasibility and superiority in wood drying.
This article proposed an new idea for wood drying system modeling, and provided a number of effective theoretical methods for the establishment of wood drying system.
引文
[1]郝华涛.木材干燥技术.北京：高等教育出版社[M],2002：1-3 19-47 104-140
[2]张璧光,李延军.热泵干燥木材的技术现状与发展趋势.干燥技术与设备[J],2003,1：6-8
[3]张璧光.我国木材干燥技术现状与国内外发展趋势.北京林业大学学报[J],2002,24(5／6)：262-266
[4]黄月瑞,严华洪.木材干燥技术问答,第一版.北京：中国林业出版社[M],1985：10-60 90-120
[5]严平,钱尚源.木材干燥机理和应用的研究.林业机械与木工设备[J],2002,30(6)：7-9
[6]张璧光.实用木材干燥技术.化学工业出版社[M],2005：200-600
[7]顾炼百.锯材干燥基准的分析和选用(木材干燥第2讲).林产工业[J],2002,29(3)：48-50
[8]李梁,张璧光,伊松林.木材干燥基准的研究现状及展望.干燥技术与设备[J],2007,5(1)：22-23
[9]顾炼百,杜国兴.中国木材干燥工业的现状与展望.南京林业大学学报(自然科学版)[J],2001,25(6)：1-5
[10]解析：我国木材干燥工业现状与科技需求,中国化工机械网,2009.8.5
[11]常建民.木材对流干燥过程热值传递规律及其湿迁移特性.东北林业大学博士论文[D].1994：1-11
[12]Mounji H,M.EL Kouali.Modeling of the drying process of Wood in 3-Dimensions.Drying Technology[J],1991,9(5):1259-1314
[13]Siau J.F.Nonisothermal diffusion model based on irreversible thermodynamics.Wood Science Technoogy[J],1992,26(5):325-328
[14]Collignan A,Nadeau J.P,etal.Description and analysis of timber drying kinetics.Drying Technology[J],1993,11(3):489-506
[15]CloutierA,Fortin Y,etal.A wood drying finite element model based on the water Potential concept.Drying Technology[J],1992,10(5):1151-1181
[16]Cloutier A,Fortin Y.A model of moisture movement in wood based on water potential and the determination of the effective water conductivity. Wood Science Technology[J],1993,27(2):95-114
[17]Tarasiewicz S,Leger F.Industrial lumber drying and its internal model conception for control system design.Porceedings of 5th International IUFRO Wood Drying Conference,Quebec City,Canada,1996:213-220
[18]P.Perre,I.W.Turner.The use of numerical simulation as a cognitive tool for studying the microwave drying of softwood in an over-sized waveguide. Wood Science Technology[J],1999,33(6):445-464
[19]P.Wiberg,T.J.Moren.Moisture flux determination in wood during drying above fibre saturation point using CT-scanning and digital image processing.Holz als Roh-und Werkstoff,1999,57(2):137-144
[20]Zhao H,Turner IW.The use of a coupled computational model for studying the microwave heating of wood.Applied Mathematical Modeling[J],2000,24(3):183-197
[21]Jarl-Gunnar Salin.Analysis and optimization of the conditioning phase in timber drying.Drying Technology[J],2001,19(8):1711-1724
[22]Aleksandar Dj.Dedic,Arun S.Mujumdar,Dimitrije K.Voronjec.A three dimensional model for heat and mass transfer in convective wood drying.Drying Technology[J],2003,21(1):1-15
[23]H.S.F.Awadalla,A.F.EI-Dib,M.A.Mohamad etal.Mathematical modeling and Experimental verification of wood drying process.Enegry Conversion&Management[J],2004,45(2):197-207
[24]张建华,常建民等.木材对流干燥质量传递经验模型.东北林业大学学报[J],1994,22(3)：101-104
[25]常建民.木材对流干燥热质传递模型的研究.林产工业[J],1996,23(1)：15-17
[26]杨庆贤.木材干燥过程中热质迁移交互作用的研究.福建林学院学报[J],1999,19(2)：101-104
[27]伊松林,张璧光,常建民.木材真空浮压干燥过程热质传递的数学模型,北京林业大学学报[J].2003,25(2)：68-71
[28]肖辉.高频真空干燥过程中木材的热质传递特性.东北林业大学硕士论文[D].2009
[29]P Carlsson,J Arfvidsson.Optimized wood drying.Drying Technology[J],2000,18(8):1779-1796
[30]Peter Carlsson,Mats Tinnsten.Optimization of drying schedules adapted for a mixture of boards with distribution of sapwood and heartwood.Drying Technology[J],2002,20(2):403-418
[31]K Cronin,P Baucour,K Abodayeh etal.Probabilistic analysis of timber drying schedules.Drying Technology[J],2003,21 (8):1433-1456
[32]Y Fortin,M Defo,M Nabhani etal.A Simulation tool for the optimization of lumber drying schedules.Drying Technology[J],2004,22(5):963-953
[33]张璧光等.木材干燥的国内外现状与发展趋势.干燥技术与设备[J],2006,4
[34]Farkas I,Remenyi P,Biro A.Modeling aspects of grain drying with a neural network.Computer Electronics in Agriculture[J],2000(29):99-113
[35]方建军,曹崇文.利用人工神经网络建立谷物干燥模型.中国农业大学学报[J],1997,2(6)P：35-38
[36]郑文利等.基于神经网络的真空冷冻干燥过程建模研究.真空[J],1998(4)：8-11
[37]吴涛,刘登瀛,许晓鸣等.利用神经网络外推预测干燥过程降水率.上海交通大学学报[J],1999,3(5)：597-599
[38]张冬妍.木材干燥神经网络建模与智能控制研究.东北林业大学博士论文[D].2005
[39]刘德胜.基于多模型数据融合算法的木材干燥动态建模研究.东北林业大学硕士论文[D].2007
[40]神经网络理论与MATLAB7实现.电子工业出版社[M],2005.3
[41]阎平凡,张长水.人工神经网络与模拟进化计算.清华大学出版社[M],2000
[42]杨仁付.基于遗传算法的BP神经网络在建模预测中的应用.铜凌学院学报[J],2007.5
[43]张文修,梁怡.遗传算法的数学基础.西安交通大学出版社[M],2000.5
[44]遗传算法理论.应用与软件实现.西安交通大学出版社[M],2002.1
[45]Srinivas,M.,and Patnaik,L.M.Adaptive probabilities of crossover and mutation in genetic algorithms.IEEE Trasaction on Systems,Man,and Cybernetics,1994,24(4):656-667
[46]He Xiaorong,Chen Bingzhen.Journal of chemical Industry and Engineering(China),1994,45(5):573-579
[47]周祥,陈炳珍等.一种用于BP神经网络训练的改进遗传算法.化工学报[J],2001.10
[48]戴朝华,朱云芳等.云遗传算法及其应用.电子学报[J],2007.7
[49]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[M],2005
[50]王树良.基于数据场与云模型的空间数据挖掘和知识发现.武汉大学硕士论文[D].2002
[51]刘桂花.基于云模型的关联规则的研究.山东师范大学硕士论文[D].2007
[52]杨海彦.基于云模型的算法改进及其在土石坝变形分析和预测中的应用.长安大学硕士论文[D].2009
[53]刘桂花,宋承祥,刘弘.云发生器的软件实现,计算机应用研究[J],2007.1
[54]蒋洪涛,李海军.结合遗传算法的BP神经网络训练方法研究,北方交通[J],2006,8
[2]张璧光,李延军.热泵干燥木材的技术现状与发展趋势.干燥技术与设备[J],2003,1：6-8
[3]张璧光.我国木材干燥技术现状与国内外发展趋势.北京林业大学学报[J],2002,24(5／6)：262-266
[4]黄月瑞,严华洪.木材干燥技术问答,第一版.北京：中国林业出版社[M],1985：10-60 90-120
[5]严平,钱尚源.木材干燥机理和应用的研究.林业机械与木工设备[J],2002,30(6)：7-9
[6]张璧光.实用木材干燥技术.化学工业出版社[M],2005：200-600
[7]顾炼百.锯材干燥基准的分析和选用(木材干燥第2讲).林产工业[J],2002,29(3)：48-50
[8]李梁,张璧光,伊松林.木材干燥基准的研究现状及展望.干燥技术与设备[J],2007,5(1)：22-23
[9]顾炼百,杜国兴.中国木材干燥工业的现状与展望.南京林业大学学报(自然科学版)[J],2001,25(6)：1-5
[10]解析：我国木材干燥工业现状与科技需求,中国化工机械网,2009.8.5
[11]常建民.木材对流干燥过程热值传递规律及其湿迁移特性.东北林业大学博士论文[D].1994：1-11
[12]Mounji H,M.EL Kouali.Modeling of the drying process of Wood in 3-Dimensions.Drying Technology[J],1991,9(5):1259-1314
[13]Siau J.F.Nonisothermal diffusion model based on irreversible thermodynamics.Wood Science Technoogy[J],1992,26(5):325-328
[14]Collignan A,Nadeau J.P,etal.Description and analysis of timber drying kinetics.Drying Technology[J],1993,11(3):489-506
[15]CloutierA,Fortin Y,etal.A wood drying finite element model based on the water Potential concept.Drying Technology[J],1992,10(5):1151-1181
[16]Cloutier A,Fortin Y.A model of moisture movement in wood based on water potential and the determination of the effective water conductivity. Wood Science Technology[J],1993,27(2):95-114
[17]Tarasiewicz S,Leger F.Industrial lumber drying and its internal model conception for control system design.Porceedings of 5th International IUFRO Wood Drying Conference,Quebec City,Canada,1996:213-220
[18]P.Perre,I.W.Turner.The use of numerical simulation as a cognitive tool for studying the microwave drying of softwood in an over-sized waveguide. Wood Science Technology[J],1999,33(6):445-464
[19]P.Wiberg,T.J.Moren.Moisture flux determination in wood during drying above fibre saturation point using CT-scanning and digital image processing.Holz als Roh-und Werkstoff,1999,57(2):137-144
[20]Zhao H,Turner IW.The use of a coupled computational model for studying the microwave heating of wood.Applied Mathematical Modeling[J],2000,24(3):183-197
[21]Jarl-Gunnar Salin.Analysis and optimization of the conditioning phase in timber drying.Drying Technology[J],2001,19(8):1711-1724
[22]Aleksandar Dj.Dedic,Arun S.Mujumdar,Dimitrije K.Voronjec.A three dimensional model for heat and mass transfer in convective wood drying.Drying Technology[J],2003,21(1):1-15
[23]H.S.F.Awadalla,A.F.EI-Dib,M.A.Mohamad etal.Mathematical modeling and Experimental verification of wood drying process.Enegry Conversion&Management[J],2004,45(2):197-207
[24]张建华,常建民等.木材对流干燥质量传递经验模型.东北林业大学学报[J],1994,22(3)：101-104
[25]常建民.木材对流干燥热质传递模型的研究.林产工业[J],1996,23(1)：15-17
[26]杨庆贤.木材干燥过程中热质迁移交互作用的研究.福建林学院学报[J],1999,19(2)：101-104
[27]伊松林,张璧光,常建民.木材真空浮压干燥过程热质传递的数学模型,北京林业大学学报[J].2003,25(2)：68-71
[28]肖辉.高频真空干燥过程中木材的热质传递特性.东北林业大学硕士论文[D].2009
[29]P Carlsson,J Arfvidsson.Optimized wood drying.Drying Technology[J],2000,18(8):1779-1796
[30]Peter Carlsson,Mats Tinnsten.Optimization of drying schedules adapted for a mixture of boards with distribution of sapwood and heartwood.Drying Technology[J],2002,20(2):403-418
[31]K Cronin,P Baucour,K Abodayeh etal.Probabilistic analysis of timber drying schedules.Drying Technology[J],2003,21 (8):1433-1456
[32]Y Fortin,M Defo,M Nabhani etal.A Simulation tool for the optimization of lumber drying schedules.Drying Technology[J],2004,22(5):963-953
[33]张璧光等.木材干燥的国内外现状与发展趋势.干燥技术与设备[J],2006,4
[34]Farkas I,Remenyi P,Biro A.Modeling aspects of grain drying with a neural network.Computer Electronics in Agriculture[J],2000(29):99-113
[35]方建军,曹崇文.利用人工神经网络建立谷物干燥模型.中国农业大学学报[J],1997,2(6)P：35-38
[36]郑文利等.基于神经网络的真空冷冻干燥过程建模研究.真空[J],1998(4)：8-11
[37]吴涛,刘登瀛,许晓鸣等.利用神经网络外推预测干燥过程降水率.上海交通大学学报[J],1999,3(5)：597-599
[38]张冬妍.木材干燥神经网络建模与智能控制研究.东北林业大学博士论文[D].2005
[39]刘德胜.基于多模型数据融合算法的木材干燥动态建模研究.东北林业大学硕士论文[D].2007
[40]神经网络理论与MATLAB7实现.电子工业出版社[M],2005.3
[41]阎平凡,张长水.人工神经网络与模拟进化计算.清华大学出版社[M],2000
[42]杨仁付.基于遗传算法的BP神经网络在建模预测中的应用.铜凌学院学报[J],2007.5
[43]张文修,梁怡.遗传算法的数学基础.西安交通大学出版社[M],2000.5
[44]遗传算法理论.应用与软件实现.西安交通大学出版社[M],2002.1
[45]Srinivas,M.,and Patnaik,L.M.Adaptive probabilities of crossover and mutation in genetic algorithms.IEEE Trasaction on Systems,Man,and Cybernetics,1994,24(4):656-667
[46]He Xiaorong,Chen Bingzhen.Journal of chemical Industry and Engineering(China),1994,45(5):573-579
[47]周祥,陈炳珍等.一种用于BP神经网络训练的改进遗传算法.化工学报[J],2001.10
[48]戴朝华,朱云芳等.云遗传算法及其应用.电子学报[J],2007.7
[49]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[M],2005
[50]王树良.基于数据场与云模型的空间数据挖掘和知识发现.武汉大学硕士论文[D].2002
[51]刘桂花.基于云模型的关联规则的研究.山东师范大学硕士论文[D].2007
[52]杨海彦.基于云模型的算法改进及其在土石坝变形分析和预测中的应用.长安大学硕士论文[D].2009
[53]刘桂花,宋承祥,刘弘.云发生器的软件实现,计算机应用研究[J],2007.1
[54]蒋洪涛,李海军.结合遗传算法的BP神经网络训练方法研究,北方交通[J],2006,8