基于L-M优化算法的喷头射程神经网络预测模型  

Prediction of Sprinkler Nozzle Range Based on L-M Optimized Algorithm

作　　者：王波雷[1] 马孝义[1] 郝晶晶[1] 

机构地区：[1]西北农林科技大学水利与建筑工程学院

出　　处：《农业机械学报》2008年第5期36-40,35,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金资助项目(项目编号:50479052);国家科技支撑计划项目(项目编号:2006BAD11B04);西北农林科技大学青年学术骨干计划资助项目

摘　　要：喷头射程受较多因素影响,各因素之间相互作用,是一个复杂的非线性系统。神经网络能有效地描述非线性模型多输入和不确定的特征。采用Levengerg-Marquardt优化算法对神经网络进行了改进,对获得的数据进行训练,建立了喷头射程预测中喷嘴仰角、喷嘴直径和工作压力的映射网络模型,并模拟分析了喷头射程与其影响因素之间的变化规律。结果表明,用基于L-M算法的人工神经网络预测喷头射程时,不需要建立具体的模型,设计方便、运算迅速、仿真性强、精确度高。Sprinkler nozzle range is affected by many factors which interact with each other ana are complex multi-variable and non-linearity systems. The artificial neural network can describe the indefinite multi-input and multi-output features of non-linearity model effectively. However, the training of artificial neural network is a complex problem by conventional back-propagation （BP） optimal algorithm, which has intrinsic weakness in slow convergence and local minimization. A prediction model for the sprinkler nozzle range was developed using Levengerg - Marquardt optimal algorithm, and the nonlinear model among the sprinkler nozzle diameter, nozzle elevation and working pressure was established. Subsequently, the relationships of nozzle range and its influence factors were simulated by Matlab. The results showed that L - M optimal algorithm is more effective and accurate than BP optimal algorithm, and it can be applied for prediction of,sprinkler range.

关 键 词：喷头射程 预测 神经网络 L—M优化算法 

分 类 号：S275.5[农业科学—农业水土工程]