基于遗传算法和支持向量机的特征子集选择方法

在模式分类系统中,往往需要从大量的特征中选择最优的特征子集,人工选择特征的方法往往费时费力,本文采用遗传算法(GA)对支持向量机进行封装的方法选择特征子集。首先使用遗传算法随机产生若干特征子集,通过选择、交叉和变异操作产生新的特征子集,经过若干代之后,得到最优的特征子集。在遗传算法中最重要的是适应度的确定,本文用支持向量机(SVM)作为分类器,为了避免出现“过拟和”,把特征子集的5阶交叉验证分类准确率和特征数量的联合函数作为适应度函数。对UCI机器学习库中SONAR和LED数据集进行实验,结果表明本方法可以有效滤除无关特征并提高分类准确率。     

基于目的层预约的双子电梯群优化控制方法

针对双子电梯群控系统多目标性、不确定性的特点,提出了一种基于目的层预约的双子电梯群优化控制方法。在分析双子电梯结构特点的基础上,构造了适合双子电梯群的目标函数;在不同交通模式下,运用改进的遗传算法进行目的层预约系统下双子电梯群的优化控制仿真。仿真结果表明,控制方法能够适应双子电梯的运行特点,并能适应不同交通流模式,与传统电梯控制系统比较,平均候梯时间、长候梯时间发生率和电梯停靠次数明显减小,可为双子电梯群控系统的工程应用提供理论依据。     

双闭环复合控制在轮胎硫化机中的应用

以硫化过程为研究对象,提出了一种分别控制扰动和系统参数变化的新型复合控制方法,并用遗传算法对控制参数进行了优化。通过对其在Matlab/Simulink中进行仿真,与模型参考自适应控制相比较,结果表明该控制方案具有更优良的鲁棒性和可靠性,控制方案简单且可实现在线整定,便于在轮胎企业推广应用。     

基于多目标遗传算法的山药真空干燥工艺优化

为了得到山药切片真空干燥的最优工艺参数,以平均干燥速率Y₁、单位能耗Y₂、复水比Y₃和白度指数Y₄为试验指标,以干燥温度T、压强V和切片厚度L为影响因素进行进行Box-Behnken响应面优化试验,通过多元非线性回归分析建立各指标的数学模型,最后分别使用加权评分法和遗传算法进行多目标优化。结果表明:多目标遗传算法优化结果更合理,得到的山药切片真空干燥最优工艺参数为:干燥温度63.76℃、压强0.0532 MPa和切片厚度2.46 mm,所对应的指标值为:平均干燥速率0.00897 g/(g·min)、单位能耗14.205 kW·h/kg、复水比2.078和白度指数81.220。     

基于遗传神经网络的矿井通风机故障诊断研究

为了提高矿井通风机机械故障诊断的准确性,提出了一种基于遗传神经网络的矿井通风机故障诊断模型。利用BP神经网络的自学习、自适应、强容错性,并通过遗传算法优化BP神经网络的连接权重和阈值。弱化了故障诊断中的人为因素,提高了评价结果的准确性和权威性。仿真结果表明,该诊断方法具有准确度高、诊断速度快等优点,是一种实用的故障诊断方法。     

Development and long-term in vivo evaluation of a biodegradable urethane-doped polyester elastomer

We have recently reported upon the development of crosslinked urethane-doped polyester (CUPE) network elastomers, which was motivated by the desire to overcome the drawbacks presented by crosslinked network polyesters and biodegradable polyurethanes for soft tissue engineering applications. Although the effect of the isocyanate content and post-polymerization conditions on the material structure-property relationship was examined in detail, the ability of the diol component to modulate the material properties was only studied briefly. Herein, we present a detailed report on the development of CUPE polymers synthesized using diols 4, 6, 8, 10, or 12 methylene units in length in order to investigate what role the diol component plays on the resulting material's physical properties, and assess their long-term biological performance in vivo. An increase in the diol length was shown to affect the physical properties of the CUPE polymers primarily through lowered polymeric crosslinking densities and elevated material hydrophobicity. The use of longer chain diols resulted in CUPE polymers with increased molecular weights resulting in higher tensile strength and elasticity, while also increasing the material hydrophobicity to lower bulk swelling and prolong the polymer degradation rates. Although the number of methylene units largely affected the physical properties of CUPE, the choice of diol did not affect the overall polymer cell/tissue-compatibility both in vitro and in vivo. In conclusion, we have established the diol component as an important parameter in controlling the structure-property relationship of the polymer in addition to diisocyanate concentration and post-polymerization conditions. Expanding the family of CUPE polymers increases the choices of biodegradable elastomers for tissue engineering applications.     

求解一般车辆优化调度问题的一种改进遗传算法

文章在对一般车辆优化调度问题(VRP)进行分析的基础上,为之建立了数学模型;提出了一种求解该问题的改进遗传算法,并通过实例验证了该算法具有优良性能。     

基于遗传神经网络优化模型的交通量预测

实时、准确的交通量预测是实现动态交通流控制及诱导的前提和基础.为了更准确地对其进行预测,本文建立了遗传神经网络优化模型,该模型既利用遗传算法全局搜索、快速收敛的优点,又利用神经网络非线性描述、自学习自适应的优点.并以实际道路为例,给出了具体的应用方法,计算机仿真结果表明该模型精度较高、具有可行性.     

A genetic algorithm for sequencing type problems in engineering design

A genetic algorithm for engineering applications that involve sequencing of operations is proposed and demonstrated. Such applications are known as travelling salesman problems in operations research literature. The proposed algorithm uses some new operators that are different from those typically used in genetic algorithms. Some enhancements for improving performance of the algorithm are also described. Treatment of two salesmen in the problem is also discussed. Results for test problems, including a vehicle A-pillar subassembly welding sequence application, show performance of the proposed algorithm to be quite robust. © 1997 John Wiley & Sons, Ltd.