欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在经典IWO杂草算法的基础上提出一种适用于神经网络优化的新算法.该算法将多种结构的神经网络权值阈值编码为不同维度的杂草种子,以神经网络均方误差作为种子适应度的统一评价标准,同时对多个维度的杂草种子进行排序筛选,实现了神经网络权值阈值与结构同时优化的目的.应用该方法于转子系统故障分类问题,实验结果表明该方法可以在结合BP算法优势的同时有效优化神经网络各参数,可以得到分类精度高、结构最简且泛化能力强的神经网络故障分类器.
英文摘要：
Based on the classical ecologically inspired meta- heuristic Invasive Weed Optimization （IWO） algorithm, ahybrid intelligent algorithm is proposed. In this method, the weights and thresholds of multi-structures of the neural networkare coded as weed seeds with different dimensions, and the MSE of the neural network is used as the uniform index forevaluation of the fitness of the weed seeds. The weed seeds with multi- dimensions are then arranged and optimized. Thesimultaneous optimization of the weights, thresholds and the structure of the neural network are realized. Then, this algorithmis applied to the fault classification of rotor systems. Results of the experiment show that this new method can reserve theadvantages of the BP algorithm while optimizing the neural network parameters, and obtain a fault classifier for the neuralnetwork with high accuracy, simple structure and strong generalization ability.
同期刊论文项目
同项目期刊论文
Fault feature extraction of a rotor system based on local mean decomposition and Teager energy kurto
期刊信息
位置：