深度连续卷积神经网络模型构建与性能分析

为了提升卷积神经网络特征提取能力,设计了一种基于连续卷积的深度卷积神经网络模型.该模型采用小尺度的卷积核来更细致地提取局部特征,并借助连续的两个卷积层增加模型的非线性表达能力,结合Dropout技术降低神经元之间的相互依赖,利用抑制网络过拟合对模型进行优化.人脸表情、手写数字字符和彩色图像的目标识别实验表明,在图像较为复杂时,该模型在识别的准确性和泛化性能上比手工特征提取方法及一般的2、3层卷积结构具有明显的优势.     

Design of hydraulic shift mechanism for AMT vehicles based on the collaborative optimization method

Based on multidisciplinary design optimization (MDO), a new design method is put forward for hydraulic shift mechanism of heavy-duty vehicle automated manual transmission (AMT). Taking a shift cylinder for example, the collaborative optimization (CO) method for the design problem of a cylinder is devided into one system level design optimization problem and three subsystem level design optimization problems. The system level is an economic model and the subsystem level is mechanics, kinetics, and a reliability model. Application of the multidisciplinary design optimization software iSIGHT modeling and solving, optimal solution of the shifting cylinder CO model is obtained. According to the optimal solution, oil cylinders are machined out and installed on the gearbox of an AMT system for the bench cycle shift test. The results show that the output force and action speed of the optimized mechanism can meet requirements very well. In addition, the optimized mechanism has a better performance compared to the structure of the traditional design method, which indicates that the CO method can optimize the design of hydraulic transmission.     

基于多因素改进型PSOSVM算法的中长期负荷预测

中长期负荷预测作为电力规划与调度中的重要一环,其影响因素有着多样性和不确定性等特点.选取支持向量机作为中长期负荷预测的核心算法,筛选多种区域宏观经济因素,利用粒子群(PSO)寻优与循环寻优的改进型算法对支持向量机(SVM)的参数进行优化及负荷预测.仿真结果显示,改进型PSOSVM算法有着较高的预测精度.     

贵州杨柳溪水库灌区渠系建筑物优化设计

针对贵州杨柳溪水库灌区地质条件较复杂,沿线部分渠段边坡稳定性较差、易形成高陡边坡等问题,在现场踏勘的基础上,综合考虑工程占地、施工难度及工程投资等因素,规划出渠系整体布置方案,优选了渠系建筑物结构形式。     

基于漂木分离效果的鱼脊型水石分离结构参数优化

山洪、泥石流中往往携带有大量漂木,漂木的堵塞作用和冲击作用往往会加剧灾害,并对山洪、泥石流防治工程造成影响。鱼脊型泥石流水石分离结构是一种新型的水石分离结构,对泥石流中粗大颗粒具有良好的分离作用,但该结构对漂木的分离作用和漂木对结构的影响尚未进行研究。本文通过模型实验,研究了引流口的漂木过流能力和水石分离格栅对漂木的分离效果,探讨了该结构对漂木的减灾效果及相关影响因素。实验结果表明水石分离格栅结构参数以及流速对漂木的分离效果有重要影响。根据实验结果提出了基于漂木分离效果的鱼脊型水石分离结构主要参数的优化方法。     

面向大曲率组合曲面的喷涂机器人喷枪轨迹组合与连接问题研究

摘要：针对喷涂机器人在大曲率组合曲面的面片上喷枪轨迹间的组合与连接问题,结合曲面拟合理论与喷枪建模方法,建立喷枪轨迹优化模型,提出一种满足大曲率组合曲面上喷枪轨迹首尾衔接的算法程序。首先,基于曲面拟合理论将大曲率组合曲面转换为自然二次曲面面片组合,建立涂层增长模型,对组合曲面上不同类型自然二次曲面面片交界处的喷枪轨迹进行优化;其次,基于该模型运用交换子和交换序列的思想,设计出实现喷涂大曲率组合曲面的喷枪轨迹组合与连接的新算法,并编写相应的算法程序,实现喷涂轨迹的全局优化;最后,通过某一种汽车外壳的喷涂实验,验证了模型的有效性和算法程序的可行性,实现了对大曲率组合曲面的连续性喷涂。     

An automatic parameter calibration method for the SWAT model in runoff simulation

Runoff simulation is highly significant for hydrological monitoring, flood peak simulation, water resource management, and basin protection. Runoff simulation by distributed hydrological models, such as the soil and water assessment tool (SWAT) model which is the most widely used, is becoming a hotspot for hydrological forecasting research. However, parameter calibration is inefficient and inaccurate for the SWAT model. An automatic parameter calibration (APC) method of the SWAT model was developed by hybrid of the genetic algorithm (GA) and particle swarm optimization (PSO). Multi‐station and multi‐period runoff simulation and accuracy analysis were conducted in the basin of the Zhangjiang River on the basis of this hybrid algorithm. For example, in the Yaoxiaba Station, the calibration results produced an R² of 0.87 and Nash Sutcliffe efficiency (NSE) index of 0.85, while verification results revealed an R² of 0.83 and NSE of 0.83. Results of this study show that the proposed method can effectively improve the efficiency and simulation accuracy of the model parameters. It can be concluded that the feasibility and applicability of GA‐PSO as an APC method for the SWAT model were confirmed via case studies. The proposed method can provide theoretical guidance for many hydrological research fields, such as hydrological simulation, flood prevention, and forecasting.     

斜板沉淀池前配水渠的数值模拟及结构优化

针对配水渠出水优化措施单一、优化效果欠佳的问题,运用CFD软件,采用Realizable k-ε湍流模型对其流速场进行数值模拟,探究了进口宽度（b）和配水渠宽度（B）变化时出水均匀度最优的配水渠长度（L）,并优化了挡墙在双段配水渠中的最佳位置。结果表明：配水渠进口集中主流在池内流域空间的扩散程度是出水均匀度的主要影响因素;当b在1 200~1 600 mm范围时,L取1 750~2 250 mm时的出水均匀度最优,若b大于1 600 mm,最优的L值须适当增加,但不应大于3 000 mm;当B为3 900 mm时,L最佳取值为1 750~2 250 mm,当B为5 300或6 700 mm时,最佳的L值应增加500 mm左右;采用双段配水渠时,挡墙位置应设在进口侧,而不应在配水渠的中间位置。该研究可为配水渠的设计提供一定的理论指导。     

全局优化的蝴蝶优化算法

针对基本蝴蝶优化算法中存在的易陷入局部最优值、收敛速度慢等问题,提出一种全局优化的蝴蝶算法,引入limit阈值来限定蝴蝶优化算法陷入局部最优解的次数,从而改变算法易陷入早熟的问题,结合单纯形策略优化迭代后期位置较差的蝴蝶使种群能够较快地找到全局最优解;将正弦余弦算法作为局部算子融入BOA中,改善迭代后期种群多样性下降的缺陷,加快算法跳出局部最优。在仿真模拟实验中与多个算法进行对比,结果表明改进算法的寻优性能更好。     

基于改进加权空间平滑的凸优化协方差估计

针对相干信号受到非均匀噪声的干扰,在低信噪比环境中常规DOA估计存在估计效果较差甚至失效的情况,基于改进加权空间平滑,提出一种使用凸优化构造最优权重矩阵的方法。改进加权空间平滑算法解相干的同时构造权重矩阵,再用凸优化重构无噪声权重矩阵,将平滑过的协方差矩阵加权,并用MUSIC算法进行DOA估计。仿真结果证实,所提方法相对于空间平滑（spatial smoothing,SS）、基于特征空间MUSIC的空间平滑估计（spatial smoothing and eigen space based MUSIC,SS-ESMUSIC）以及接收信号协方差矩阵秩最小化（spatial smoothing based covariance rank minimization,SS-CRM）算法能更好地抑制非均匀噪声和解相干,且减少了低信噪比的干扰,展现出更优良的分辨力和准确性。