基于耦合神经网络工业产品选材的实例检索模型

论述了耦合神经网络的实例推理技术,实例推理的典型实现框架、基于约束对象的实例检索模型及实例的三级层次组织.并在此基础上综合运用面向对象技术、基于实例推理技术和人工神经元网络模型等先进技术提出耦合神经网络工程用钢的实例检索模型以达到缩小检索范围,提高中标率的目的.     

基于位索引的多层次实例检索模型研究

通过对基于实例的推理技术、实例的检索策略及实例相似性的研究，提出了可应用于机械产品设计领域的多层次实例检索模型和基于位索引的检索算法。给出了实例检索模型的层次划分方法，分析了位索引的建立过程和通过相似度映射表来提高实例检索效率的原理，最后通过实例验证了该算法的简洁性和高效性。     

轴类零件耦合神经网络实例推理 CAPP 索引模型的研究

提出了一种基于神经网络的工艺设计实例推进索引模型。与现存大多数实例推理系统不同,该方法用神经网络实现实例的动态分类和索引。实例层次分类的三层结构和基于特征的聚类模板概念,为实现基于符号处理的实例推理求解模式向基于神经计算的模式识别求解模式映射提供了条件。该方法的优点在于实例的高速、有效检索,知识获取的简化以及基于神经网络的检索算法的鲁棒性。     

基于规则推理的最近邻实例检索模型研究

实例检索是基于实例产品设计的关键。本文用基于多粒度的方法统一表示实例和规则的索引 ,利用规则推理减小实例检索的范围 ,提高检索的质量 ,提出了以最近邻算法为基础的实例检索模型 ,支持索引词汇的动态扩充和特征权重与特征属性相似度的动态调整。开发的“基于混合推理的机床夹具CAD系统”已在某大型航空企业成功地应用于国家重点型号工程的研制 ,验证了本文算法的可行性和高效性。     

基于神经网络和最近邻相似度的实例检索算法

为了提高实例推理系统中实例检索的效率与质量,提出了基于神经网络的、和最近邻相似度相结合的实例检索算法。首先,利用概率神经网络实现对实例的动态分类,以缩小实例检索范围;其次,介绍了实例属性相似度计算方法,针对各属性相似度的权重问题,给出主观和客观赋权方法;最后,给出截止阀总体设计方案生成的算例,验证方法的可行性。     

一种基于知识和耦合神经网络实例混合推理的智能CAPP策略研究

提出一种新的耦合神经网络的实例与知识混合推理策略。采用面向对象的方法表达实例和知识,并将神经网络用于实例推理中,在此基础上实现了CAPP的变异设计。建立了基于零件及其工艺数据知识的分层分解表达模式与层次式优化推理机制,从而实现了CAPP的创成式设计。给出智能化CAPP系统的总体结构,对基于神经网络的实例相似性判定算法、知识表示模型和知识推理控制策略作了讨论。由于吸收了派生法的类比设计思想,又具有创成功能,整个CAPP系统的推理决策具有更高的效率和更好的质量。     

基于TRIZ知识库的实例检索研究

为弥补以TRIZ为核心的CAI在实例检索中的不足,通过对TRIZ、CAI及CBR的分析,提出了分析与计算当前问题与成功实例的相似度与匹配度算法,构建了基于当前设计问题与实例间匹配度的功能本体知识库实例检索机制。同时,构建了基于CBR的解题流程。     

最近邻实例检索相似度分析

针对最近邻实例检索中实例属性相似度和权重的计算问题,给出了区间值属性相似度的计算模型,并将各种属性类型的相似度计算方法加以统一,提出了基于相似度离差信息的客观赋权方法,并以组合权重计算实例相似度.该方法已应用于基于实例推理的导弹概念设计之中,结果表明,所提出的权重计算方法更有利于实例的检索.     

回转类零件工艺设计中实例表达和实例检索研究

针对回转类零件给出基于实例推理工艺设计的实例表达方法,构造一种相似性系数的计算公式,论述新零件与实例进行比较和匹配的策略和算法。在实例检索过程中,匹配信息被详细记录下来,使得后续的实例工艺筛选和工艺修正变得相对简单并易于实现。最后给出了系统运行的实例。     

摩托车智能设计实例推理系统的索引模型

为提高实例推理系统中实例检索的效率与质量,提出了基于人工神经网络的实例检索模型。该模型在实例层次组织的基础上,利用自适应共振网络实现对实例的动态分类,以缩小实例搜索的范围。采用前馈型神经网络记忆各实例的索引,以便在缩小的实例范围内快速地提取相似实例,提高检索效率与质量。最后给出了摩托车总体设计过程方案选择的算例,并将检索的结果与基于最近邻法的实例检索结果进行比较分析,结果证明了该模型的有效性。     

Hybrid neural network bushing model for vehicle dynamics simulation

Although the linear model was widely used for the bushing model in vehicle suspension systems, it could not express the nonlinear characteristics of bushing in terms of the amplitude and the frequency. An artificial neural network model was suggested to consider the hysteretic responses of bushings. This model, however, often diverges due to the uncertainties of the neural network under the unexpected excitation inputs. In this paper, a hybrid neural network bushing model combining linear and neural network is suggested. A linear model was employed to represent linear stiffness and damping effects, and the artificial neural network algorithm was adopted to take into account the hysteretic responses. A rubber test was performed to capture bushing characteristics, where sine excitation with different frequencies and amplitudes is applied. Random test results were used to update the weighting factors of the neural network model. It is proven that the proposed model has more robust characteristics than a simple neural network model under step excitation input. A full car simulation was carried out to verify the proposed bushing models. It was shown that the hybrid model results are almost identical to the linear model under several maneuvers. This paper was recommended for publication in revised form by Associate Editor Hong Hee Yoo Dr. Wan-Suk Yoo was born in 1954, and received B.S. degree from Seoul National University (1976), and got M.S. degree from KAIST (1978) and Ph.D. from the University of Iowa (1985). He is currently a full professor at the Pusan National University in Korea, where he joined since 1978. His major area is vehicle dynamics and flexible multibody dynamics. He became an ASME Fellow (2004), and currently serving as an associate editor for the ASME, J. of computational and nonlinear dynamics. He is also serving a contributing editor for the multibody system dynamics journal. He is serving as ISC chair for the ACMD2008, and a member at IFToMM TC for multibody dynamics. He is currently a vicepresident of the KSME (Korean Society of Mechanical Engineers).     

一类基于统计理论的神经网络模式识别方法

本文针对用人工神经网络进行模式识别时样本特征指标过多的问题,提出用统计理论的主成分分析方法对数据进行预处理,再选出几个主成分作为神经网络的输入节点,从而极大地简化人工神经网络,提高了模式识别的效果。     

A neural network application for reliability modelling and condition-based predictive maintenance

Traditionally, decisions on the use of machinery are based on previous experience, historical data and common sense. However, carrying out an effective predictive maintenance plan, information about current machine conditions must be made known to the decision-maker. In this paper, a new method of obtaining maintenance information has been proposed. By integrating traditional reliability modelling techniques with a real-time, online performance estimation model, machine reliability information such as hazard rate and mean time between failures can be calculated. Essentially, this paper presents an innovative method to synthesise low level information (such as vibration signals) with high level information (like reliability statistics) to form a rigorous theoretical base for better machine maintenance.     

Prediction of chip flow angle in orthogonal turning of mild steel by neural network approach

Improvement of chip control is a necessity for automated machining. Chip control is closely related to chip flow and it plays also a predominant role in the effective control of chip formation and chip breaking for the easy and safe disposal of chips, as well as for protecting the surface-integrity of the workpiece. Although several ways to predict the chip flow angle (CFA) have been subjected in some researches, a good approximation has not been achieved yet. In this study, using different indexable inserts and cutting conditions for turning of mild steel, the chip flow angles were measured and some of the collected data from this experimental study were used for training with a two hidden layered backpropagation neural network algorithm. A group was formed from randomly selected data for testing. The chip flow angle values found from multiple regression, neural network (NN) and studies of previous researchers under the same turning conditions of the present study were compared. It has been seen that the best prediction was obtained by neural network approach.     

综合集成网络在电镦机加热电流预报中的应用

选取镦粗缸活塞运动速度、砧子缸活塞运动速度、镦粗压力和加热时间作为输入参数,加热电流作为输出参数,并用数学模型、BP神经网络、加法网络、乘法网络以及神经网络与机理模型综合集成的五种方案来对加热电流进行预报。比较结果表明,综合集成模型将数学模型的知识集成到网络结构中,在“小样本”时,不仅能减少连接权值,而且能加快训练速度,提高泛化能力,在电镦机加热电流的预报中取得了良好效果。     

神经网络在火灾视频识别中的应用

利用早期火灾火焰的形体变化特性,结合火灾视频探测信号的特点,用图像处理方法提取火焰图像的面积、形状及纹理特征并形成火灾识别判据.设计BP神经网络结构,利用训练好的神经网络对有、无干扰源火灾环境进行火灾识别,验证结果表明该方法的正确识别报警率较高,能够减少火灾报警系统漏报率和误报率,是一种可靠的智能火灾探测方法.     

基于神经网络的线图数据处理模型

提出了一种能够一次处理一组线图数据的新方法，基于神经网络的基本原理，建立了能够处理线图数据的神经网络模型，并且详细的讨论了径向基函数神经网络的具体应用。成为处理线图数据尤其是多线图数据的有利工具。处理过程主要分成三步：(1)采集数据，建立输入、输出矩阵，(2)训练网络模型，(3)检验误差。文中给出了应用实例及误差分析。     

净水厂最佳投药量的神经网络控制系统

混凝剂的投加是净水厂水处理工艺的重要环节，本文通过对目前水厂实际运行过程中混凝剂投加量的影响因素的分析，利用神经网络预测理论，建立了净水厂混凝剂投加量的预测模型，并利用该模型对某水厂的实际运行情况进行了预测，对网络预测模型的性能进行了验证。结果表明，该模型具有很强的自学习性，自适应性和容错性，依靠网络的在线自学习，可使预测结果的准确度明显提高。     

船舶航迹保持的在线SIMO神经网络控制器研究

本文提出了一种船舶航迹保持的在线神经网络控制器.该控制器能解决精确的船舶动态模型难以建立问题,它能用舵角同时控制航迹偏差和航向偏差,它通过对控制精度的直接计算来自动在线训练学习,它不需离线训练学习过程.计算机仿真结果表明控制器训练方法的有效性和控制的鲁棒性.