基于改进LSTM-NN的安全性自动驾驶换道轨迹规划模型

为改善现有的自动驾驶换道轨迹规划模型产生的换道轨迹与真实的换道轨迹存在较大偏差的问题,提出了一种改进LSTM-NN的安全敏感性深度学习模型,该模型可以缓解当前自动驾驶轨迹规划存在的不足,输出轨迹既保证了较高的精度又提高了安全性。CarSim仿真软件模拟了本模型产生轨迹的可跟踪性,结果显示轨迹非常平滑,并且自动驾驶车辆可以高效、安全地完成换道。     

基于数据驱动的电动汽车动力电池SOC预测

为准确预测电动汽车动力电池的能耗,缓解驾驶者的里程焦虑,本文中提出一种基于数据驱动的电动汽车动力电池SOC预测模型.首先分析电动汽车能耗构成并提取能耗影响因素,接着基于某款电动出租车CAN总线采集的汽车运行数据,采用机器学习算法,提出基于温度分层的能耗模型,通过宏观数据与微观数据的融合减小误差,最后使用该模型对车载BM...     

基于深度学习的智能车辆视觉里程计技术发展综述

针对基于模型的视觉里程计在光照条件恶劣的情况下存在鲁棒性差、回环检测准确率低、动态场景中精度不够、无法对场景进行语义理解等问题,利用深度学习可以弥补其不足。首先,简略介绍了基于模型的里程计的研究现状,然后对比了常用的智能车数据集,将基于深度学习的视觉里程计分为有监督学习、无监督学习和模型法与深度学习结合3种,从网络结构、输入和输出特征、鲁棒性等方面进行分析,最后,讨论了基于深度学习的智能车辆视觉里程计研究热点,从视觉里程计在动态场景的鲁棒性优化、多传感器融合、场景语义分割3个方面对智能车辆视觉里程计技术的发展趋势进行了展望。     

正交异性组合桥面构造优化

正交异性组合桥面由于其较好的抗疲劳性能和静力承载能力、相对较轻的自重近年来在实际工程中得到较多应用。正交异性组合桥面通常由正交异性钢板和混凝土层组成,在加入混凝土层后结构刚度发生较大变化,若正交异性板仍采用传统正交异性钢桥面的构造形式将会造成不必要的钢材用量和焊接量增加。采用遗传算法,以单位面积造价最小为目标,结合现有规范考虑了构造约束和承载能力约束两类约束条件,采用两种应力控制方案,对不同桥面板跨度和混凝土厚度的正交异性组合桥面的构造进行优化。同时解决了构造改变导致荷载分配改变对承载能力分析造成的影响。优化结果表明,桥面跨度4.5m时,与未优化组合桥面相比,优化方案用钢量可降低1.7%~10.1%,焊接量可降低16.5%~31.8%。     

用遗传算法求空间点列的最小包容球

建立了求空间点列包容球的数学模型,将求最小包容球问题转化为函数优化问题,并用遗传算法解决了函数最优解的求解问题.     

A second order volume of fluid （VOF） scheme for numerical simulation of 2-D breaking waves

Among all environmental forces acting on ocean structures and marine vessels, those resulting from wave impacts are likely to yield the highest loads. Being highly nonlinear, transient and complex, a theoretical analysis of their impact would be impossible without numerical simulations. In this paper, a pressure-split two-stage numerical algorithm is proposed based on Volume Of Fluid (VOF) methodology. The algorithm is characterized by introduction of two pressures at each half and full cycle time step, and thus it is a second-order accurate algorithm in time. A simplified second-order Godunov-type solver is used for the continuity equations. The method is applied to simulation of breaking waves in a 2-D water tank, and a qualitative comparison with experimental photo observations is made. Quite consistent results are observed between simulations and experiments. Commercially available software and Boundary Integral Method (BIM) have also been used to simulate the same problem. The results from present code and BIM are in good agreement with respect to breaking location and timing, while the results obtained from the commercial software which is only first-order accurate in time has clearly showed a temporal and spatial lag, verifying the need to use a higher order numerical scheme.     

对ANN技术表达船体曲线算法的改进

应用人工神经元网络（ANN）技术表达船体曲线。根据问题性质,选用小波基作为前向单层神经网络的神经元激励函数,结合逐层学习（OHLO）算法对一艘3．6万吨散货船的后半体进行了表达。编程运算结果表明,该方法速度较传统的BP算法有较大提高。     

A two-phase model for dry port location with an application to the port of Dalian in China

This paper provides guidance for an optimal and reasonable dry port layout for the port of Dalian in China. We present a two-phase framework on the location of dry ports, which solves the selection of candidate inland cities and optimal dry port location choice, respectively. Fuzzy C-Means Clustering is applied to select alternative cities in the vast hinterland of the seaport of Dalian, with a view to identify evaluation factors that affect the location selection decision. A cost-minimisation linear programming solution is proposed, with the aid of a genetic algorithm, to choose the optimal location as well as capacity level among the candidate inland cities.     

铁路客车轮对踏面硌伤自动分析系统的研究与实现

为了解决客车轮对踏面硌伤人工分析效率低且质量不高的问题,基于大数据关联分析技术,构建了一套硌伤自动分析系统.从硌伤排查、自动分析、故障定位、统计分析4个方面分析系统的主要组成和功能,探讨了实现系统的关键技术.系统的应用提高了对客车轮对踏面硌伤分析和故障排查的效率.     

智慧校园环境下课程资源建设的实践研究 ——以汽修专业《汽车空调原理与检修》课程建设应用为例

为研究智慧环境下课程资源建设,文章以《汽车空调原理与检修》课程建设应用为例,通过建设该教学平台,实现线上线下教学的有机结合,可以有效提升教师的教学质量及教学效率,还可以为那些渴望更深层次知识的学生提供一个良好的学习平台。在“智慧校园”的建设和发展过程中,需要不断更新与教学相关的信息,因此建设《汽车空调原理与检修》课程教学资源是高等院校课程改革的需求。