基于神经网络和最近邻相似度的动臂实例检索  

Case Retrieval Strategy Based on Neural Network and Nearest Neighbor Similarity

作　　者：闫二乐 高领付 林述温[1] 林航[1] YAN Erle;GAO Lingfu;LIN Shuwen;LIN Hang(School of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学机械工程及其自动化学院,福建福州350108

出　　处：《机械制造与自动化》2018年第5期167-171,共5页Machine Building & Automation

基　　金：国家自然科学基金资助项目(51175086)

摘　　要：为了参照已有的设计实例实现工程机械的快速设计,缩短产品的设计周期,提出基于神经网络和最近邻相似度的方案检索策略。先对过去历史的设计实例进行结构性分解,选择适当的属性特征;再运用概率神经网络对实例进行分类,缩小二次检索的范围;最后针对动臂结构参数多、最近邻法的权重系数设置难点,提出专家经验知识和客观数据相结合的方法,计算各个实例与目标实例的相似度,并以一个设计需求进行实例推理验证基于实例推理方法的准确性和有效性。To extract the experience from the existing design examples to rapidly design the engineering machinery and shorten the product design cycle,this paper puts forward a case retrieval strategy based on neural network and nearest neighbor similarity.Firstly,the structural examples of the past history are decomposed and the appropriate attribute characteristics are selected.Secondly,the probabilistic neural network is used to classify the examples and narrow the scope of the secondary retrieval.Finally,for the parameters of the boom structure and the difficult set point of the nearest neighbor weight coefficient,it puts forward a method that the expert experience is combined with the objective data,which is used to calculate the similarity between each instance and target instance.And an example is used to validate the accuracy and validity of the case-based reasoning method.

关 键 词：工程机械 快速设计 初始方案设计 基于实例推理 神经网络 最近邻距离法 相似度 

分 类 号：TH122[机械工程—机械设计及理论] TP183[自动化与计算机技术—控制理论与控制工程]