基于集成经验模态分解与集成机器学习的锂离子电池剩余使用寿命预测方法

准确预测储能锂离子电池剩余使用寿命(remaining useful life, RUL)对于电力系统的安全性与可靠性至关重要。针对锂离子电池老化轨迹呈现非线性变化的问题,提出一种基于集成经验模态分解(ensemble empirical mode decomposition, EEMD)和集成机器学习的锂离子电池剩余使用寿命预测方法。首先,利用集成经验模态分解算法分解锂离子电池老化数据。其次,分别利用集成的长短时记忆神经网络与相关向量机对分解得到的残差数据序列和本征模态数据序列建模预测。最后,融合预测的残差数据序列和本征模态数据序列,综合计算锂离子电池未来寿命老化轨迹。采用储能锂离子电池老化数据进行验证,结果显示所提出的锂离子电池RUL预测方法具有更好的鲁棒性与非线性跟踪能力。     

基于DeepLab V3模型的图像语义分割速度优化研究

为了解决移动端视觉感知模块的内存资源和硬件条件不适应卷积神经网络的快速应用等问题。以经典的DeepLab V3模型为基础,在保证分割精度的前提下,以优化卷积神经网络模型的运行速度为目标,提出一种新的轻量化网络结构。所设计的网络结构将原有非线性激活函数替换成新的Swish激活函数进行精度补偿,采用改进后的轻量化MobileNet V2结构替代DeepLab V3原有的特征提取器。实验结果表明,改进的DeepLab V3网络模型和目前精度最高的DeepLab V3+算法相比,其在维持一定精度的前提下,参数量和计算复杂度大大减小,运行速度明显提升,模型内存占用率下降了近96%,综合性能更强,更适合对分割性能要求较高的快速分割网络。     

High-performance zirconia ceramic additively manufactured via NanoParticle Jetting

Additive manufacturing has received tremendous attention in the manufacturing and materials industry in the past three decades. Zirconia-based advanced ceramics have been the subject of substantial interest related to structural and functional ceramics. NanoParticle Jetting (NPJ), a novel material jetting process for selectively depositing nanoparticles, is capable of fabricating dense zirconia components with a highlydetailed surface, precisely controllable shrinkage, and remarkable mechanical properties. The use of NPJ greatly improves the 3D printing process and increases the printing accuracy. An investigation into the performance of NPJ-printed ceramic components evaluated the physical and mechanical properties and microstructure. The experimental results suggested that the NPJ-fabricated ZrO₂ cuboids exhibited a high relative density of 99.5%, a glossy surface with minimum roughness of 0.33 μm, a general linear shrinkage factor of 17.47%, acceptable hardness of 12.43 ± 0.09 GPa, outstanding fracture toughness of 7.52 ± 0.34 MPa m^1/2, comparable flexural strength of 699 ± 104 MPa, dense grain distribution of the microstructure, and representative features of the fracture. Subsequently, the exclusive printing scheme that achieved these favorable properties was analyzed. The innovative NanoParticle Jetting? system was shown to have significant potential for additive manufacturing.     

基于CRF和深度学习的病历实体识别的研究

随着电子病历数据量的快速增长,如何深层次、高效率地利用电子病历资源成为越来越迫切需要解决的问题. 从真实病历出发,研究电子病历的医学实体识别问题,为计算机更好地辅助医疗奠定基础. 通过人工标注的108份心血管科的真实病历数据与3类特征模板,运用条件随机场和双向长短时记忆网络联合条件随机场对心血管科电子病历疾病命名实体抽取的实验,并进行比较分析. 结果表明,结合合适的特征模版,条件随机场模型有更好的抽取性能,是一种较为适用的病历命名实体抽取方法.     

聚乙烯表面接枝改性及应用的研究进展

聚乙烯是常用的热塑性工程材料,具有良好的物理化学性能和加工性能,被广泛应用于各个领域。但是其独特的线性分子结构,使得聚乙烯呈现出低表面能、低润湿性、弱粘结等特点,限制了应用的扩展。表面接枝技术,能够在聚乙烯材料表面引入功能基团,改善表面性能进而拓展其功能及应用。本文介绍了用于聚乙烯材料表面接枝改性的常用技术,包括紫外光辐照接枝、电子束辐照接枝、等离子体接枝、高能γ射线辐照接枝和可逆失活自由基聚合等,并详细阐述了每种接枝技术的基本机理。介绍和讨论了接枝改性聚乙烯材料的多种性能,如摩擦学性能、生物相容性、抗菌性、抗冲击性能、离子导电性及其他性能。在此基础上,对改性聚乙烯材料的相关应用领域进行了综述。最后,提出接枝改性聚乙烯研究中存在的问题,并针对其应用前景进行展望。     

An ellipsoid algorithm for equality-constrained nonlinear programs

This paper describes an ellipsoid algorithm that solves convex problems having linear equality constraints with or without inequality constraints. Experimental results show that the new method is also effective for some problems that have nonlinear equality constraints or are otherwise nonconvex.Scope and purposeThe purpose of this paper is to present a variant of the ellipsoid algorithm that can be used with equalities. This is a significant improvement over the classical algorithm, which yields accurate solutions to convex and many nonconvex nonlinear programming problems but requires the feasible set to be of full dimension and therefore cannot be used with equality constraints.     

四足机器人对角小跑步态非线性控制方法仿真

针对现有的四足机器人对角小跑步态控制方法存在的机器人运动速度较慢、灵活性较差等问题,提出了一种基于虚拟模型的四足机器人对角小跑步态非线性控制方法。方法需要构建一个四足机器人模型,并在该模型的工作范围内建立一个平面直角坐标系,在不考虑机器人足端车轮滑动的情况下,将驱动四足机器人的运动方程转换成矩阵的形式,寻找有界输入平动线速度和转动角速度,使矩阵在其控制下产生的误差可以在大范围内保持稳定。求解该四足机器人在工作平面坐标系中姿态误差的微分方程,构造该微分方程的Lyapunov函数并对其求导,根据求导结果设计一个四足机器人驱动控制器,通过该驱动控制器实现对四足机器人的对角小跑步态非线性控制。仿真结果表明,所提方法能够在快速、灵活的情况下实现对四足机器人对角小跑步态的非线性控制,且鲁棒性较高,能够满足用户需求。     

Non-cooperative signal modulation recognition algorithm based on joint feature parameter extraction

Aiming at the problem that in the current electromagnetic environment,the modulation method is complicated,the frequency-consuming equipment increases,the spectrum is congested,and the electromagnetic environment interference increases,the algorithm of OFDM signal detection and subcarrier identification in the background of non-cooperative communication were deeply studied.Using the different distribution states of OFDM signals and single carrier signals in the time domain,a joint characteristic parameter was proposed to solve the existence problem of OFDM in the received signal.For the phase shift and frequency offset problems caused by the channel transmission to the signal,by using the periodic stability the blind parameter estimation was performed to obtain the signal prior information.On the basis of the obtained signal prior information,a multi-level classification and recognition method for non-cooperative OFDM signal sub-carrier signals was proposed.Therefore,a model based on non-cooperative communication system OFDM signal detection and subcarrier modulation identification was designed,and finally modulation identification of unknown signals was completed.Simulation experiments show that in non-cooperative communication systems,OFDM signals and single-carrier signals can be accurately identified,and ideal modulation recognition effects can be achieved on empty subcarriers,QPSK,and 16QAM in the receiver OFDM signal subcarriers,overcoming the channel transmission band The problems of phase shift and frequency offset have improved the accuracy of modulation mode identification.