基于无线传感器网络的温湿度监测系统设计与实现

给出了以CC2430芯片为核心的环境监测无线传感器网络实现方案.设计中充分考虑到节点的剩余能量、链路质量和节点负载等因素,提出了一种能耗均衡树算法,使网络节点的负载逐渐趋于均衡,解决了因负载分配不均衡,使得网络节点出现过早失效、缩短网络寿命问题.实验表明,算法可均衡节点负载,延长网络的生存时间,实现网络的长效运行.     

水果表面缺陷自动检测系统中的人工智能方法研究

通过计算机视觉技术获取了带有各种表面缺陷的苹果图像并进行预处理，采用自适应特征聚类（ＳＡＦＣ）神经网络与模糊加权决策树（ＦＷＤＴ）相结合的方法实现了缺陷区域的准确检测和详细分类。实验结果表明，用人工智能方法进行表面缺陷检测，具有良好的抗噪容错能力并能有效地克服传统图像分割方法适应性差的缺点，提高判别准确率和分类精度。     

基于多层感知器神经网络对遥感融合图像和TM影像进行土地覆盖分类的研究

８０年代以来，人工神经元网络（ＡＮＮ）技术的应用不断向广度和深度发展．用不同空间分辨率的ＴＭ和ＩＲＳ遥感图像进行融合，综合了不同传感器数据所提供的信息，增强了图像的清晰度，改善了解译效果。用多层感知器神经网络对遥感融合图像进行分类，分类精度达９５％，比用多层感知器神经网络对ＴＭ图像进行分类（分类精度达７３％）效果要好．     

工程车辆并行自适应神经网络自动换挡控制

为了解决工程车辆在重载作业时传动系统效率大幅下降的问题,以工程车辆"两参数"换挡规律为依据,提出一种并行自适应神经网络自动换挡控制的方法。结构上由神经网络控制、自适应神经网络模型、网络评价和运行监控模型组成。仿真结果表明,该控制方法提高了工程车辆液力机械传动系统效率,有效地克服了神经网络控制实时性差,难以在工程实际中应用的问题,实现了换挡控制的智能化。     

离心输油泵故障诊断专家系统的研究与实现

提出了一种新的对基本神经网络进行组合的思路，将基本神经网络分层联接，每一层对应对象的某一属性，并根据属性分类。以这一思路为指导，组建了BP网络与竞争网络组织相结合的神经网络，并以此为核心，搭建了离心输油泵故障诊断专家系统的框架。     

PLD在小水电流接地选线装置中的

根据复杂可编程逻辑器件（CPLD）的特点及其在继电保护的应用背景，探讨了CPLD在配电网单相接地故障选线装置中的设计与使用方法。应用CPLD不仅可以大大简化电路、提高装置的可靠性；而且装置逻辑功能的改变可以通过编程实现，使得装置的调试更加方便。     

From on-farm solid-set sprinkler irrigation design to collective irrigation network design in windy areas

In this paper, a contribution to the design of collective pressurised irrigation networks in solid-set sprinkler-irrigated windy areas is presented. The methodology is based on guaranteeing minimum on-farm performance, using a historical hourly wind speed database and a ballistic solid-set irrigation simulation model. The proposed method was applied to the Montesnegros Irrigation District (central Ebro basin, Spain). The district irrigates an area of 3493 ha using an on-demand schedule. The average wind speed in the area is 2.8 m s⁻¹. An analysis of district water records showed that farmers often reduce water demand when the wind speed is high, but their irrigation decision making is limited by the capacity of the irrigation network and by the unpredictable character of local winds. Simulations were performed for 11 irrigation seasons, 2 triangular sprinkler spacings (18 m × 18 m and 18 m × 15 m), and 2 sprinkler models. The percentage of monthly suitable time for irrigation was determined for four management strategies. The first one was based on a wind speed threshold (3 m s⁻¹), while the other three were based on three levels (standard, relaxed and restrictive) of two irrigation performance parameters: the Christiansen Uniformity Coefficient (CU) and the Wind Drift and Evaporation Losses (WDEL). The standard strategy classified the time as suitable for irrigation when CU ≥ 84% and WDEL ≤ 20%. The thresholds limits of the irrigation parameters for the relaxed strategy were CU ≥ 80% and WDEL ≤ 25%. Finally, the restrictive strategy used thresholds of CU ≥ 90% and WDEL ≤ 15%. The suitable time for the first strategy (56%) was always lower than for the standard and the relaxed strategies (with respective average values of 75 and 86%), and higher than for the restrictive strategy (30%). In order to design the collective network, the hydrant operating time was equalled to the suitable time for irrigation. The differences in the cost of the collective network plus the on-farm equipment were particularly relevant between the restrictive strategy and the other three. Differences in suitable operating time were clear between sprinkler spacings, and less evident between sprinkler models. The application of the proposed methodology may be limited by the availability of historical wind speed records and CU estimates for different combinations of sprinkler models, sprinkler spacings and wind speed.     

电子差速履带车辆转向转矩神经网络PID控制

根据电子差速履带车辆转向动力学和运动学分析,提出一种电子差速履带车辆转向转矩模拟神经网络PID(ANNPID)控制策略,由双电动机转向转矩协调控制、ANNPID控制和感应电动机转矩控制组成.通过建立双感应电动机独立驱动履带车辆电子差速转向控制系统,实现基于ANNPID控制的转向转矩协调分配和基于模型参考自适应控制(MRAC)的感应电动机间接磁场定向(IFOC)转矩控制.采用该策略,在不同转向半径的行驶转向工况、0.5B半径转向工况和中心转向工况下的实车试验结果表明,低速转向具有较好的操控性能.     

基于计算机视觉和神经网络的牛肉颜色自动分

将采集的牛胴体眼肌切面图像人工确定其颜色等级,然后通过计算机图像处理方法,分割出肌肉区域并提取出其在RGB和HIS颜色空间的颜色特征参数.设计一个以牛肉的颜色特征参数为输入、牛肉的颜色等级为输出的BP神经网络模型,通过训练,确定模型的结构参数,用测试样本对该模型进行验证.结果显示,用该模型进行牛肉颜色等级预测的正确率可达95%,耗时仅0.25s.表明利用所设计的模型可以对牛肉的颜色等级进行快速、准确的自动判定.     

参考作物腾发量主成分神经网络预测模型

为解决采用神经网络模型预测参考作物蒸发蒸腾最Eto研究中预测能力不足的问题,将气象因子包括最高、最低和日平均温度、日照时数、气压、水汽压、相对湿度和风速进行主成分分析,提取主成分,建立了基于主成分的三层BP神经网络模型.选取崇川水利科学试验站2001年到2004年的旬气象资料,采用Matlab神经网络工具箱进行模型训练与预测,并以传统BP网络模犁作为对照.结果表明,主成分网络模型能够很好地反映诸多影响因子与Eto之间的关系,尤其对训练样本以外的验证样本,主成分网络模型具有显著优于传统BP网络模型的识别能力,取得更为可靠的预测结果.