基于征兆邻搜索优化聚类和自组织映射神经网络的多病害诊断  

Method for multi-disease diagnosis based on optimized symptom adjacent-searching clustering and SOM NN

作　　者：张可[1] 柴毅[1] 匡金骏[1] 

机构地区：[1]输配电装备及系统安全与新技术国家重点实验室,自动化学院,重庆大学,重庆400030

出　　处：《农业工程学报》2011年第1期215-222,共8页Transactions of the Chinese Society of Agricultural Engineering

基　　金：中央高校基本科研业务费No.CDJRC10170005与No.CDJZR11170005资助

摘　　要：复杂过程具有多样性的特点,常出现多种异常同时发生的情况。针对该问题,对异常过程中征兆的表现及其描述进行了分析,在已有自组织特征映射神经网络(SOM NN,Self-organizing Map Neural Networks)单一故障(病害)诊断的方法的基础上,提出了具有3级分析结构的SOM NN的多诊断模型。该模型以欧几里德距离作为主要判别条件对邻搜索方法进行优化和改进,在诊断过程中不用学习多病害样本。并在此基础上以农作物中具有代表性的番茄病害为例,提取病害征兆,建立病害与病害征兆之间的映射关系,完成了对病害征兆组合的分类,通过对实例的仿真,证明了该方法在多病害诊断上能获得良好的效果。Complex processes have the characteristic of multifarious,and simultaneity multi-abnormality is familiar in the area. Aimed at this problem,the representations and descriptions of symptom with abnormality were analyzed. Based on an existing mono-fault （mono-disease） diagnosis method by Self-Organizing Map Neural Networks （SOM NN）,a multi-fault （multi-disease） diagnosis model was developed. This proposed SOM NN-based model has three layers,it has no need to study multi-disease samples. According to the analysis,Euclidean distance was taken as the main discrimination,and the sufficiency and necessity of symptom adjacent-searching were analyzed. The adjacent-searching algorithm was optimized and improved. Taking tomato disease as an example,the disease symptoms were extracted,and the mapping relationship between disease and symptom were developed. Using the method,the correct cluster results of disease symptom combinations were obtained. This model can achieve an accurate diagnosis of multi-diseases. The simulation results show that the proposed model performs well and the proposed multi-disease diagnosis is effective.

关 键 词：多病害诊断 人工神经网络 自组织映射 邻搜索优化 聚类分析 番茄病害 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] S436.412.1[自动化与计算机技术—控制科学与工程]