An improved adaptive unscented Kalman filter for estimating the states of in‐wheel‐motored electric vehicle

Vehicle state is essential for active safety stability control. However, the accurate measurement of some vehicle states is difficult to achieve without the use of expensive equipment. To improve estimation accuracy in real time, this paper proposes an estimator of vehicle velocity based on the adaptive unscented Kalman filter (AUKF) for an in‐wheel‐motored electric vehicle (IWMEV). Given the merits of an independent drive structure, the tire forces of the IWMEV can be directly calculated through a vehicle dynamic model. Additionally, by means of the normalized innovation square, the validity of vehicle velocity estimation can be detected, and the sliding window length can be adjusted adaptively; thus, the steady‐state error and the dynamic performance of the IWMEV are demonstrated to be simultaneously improved over an alternative approach in comparisons. Then, an adaptive adjustment strategy for the noise covariance matrices is introduced to overcome the impact of parameter uncertainties. The numerically simulated and experimental results prove that the proposed vehicle velocity estimator based on AUKF not only improves estimation accuracy but also possesses strong robustness against parameter uncertainties. The deployment of the estimation algorithm by using a single‐chip microcomputer verifies the strong real‐time performance and easy‐to‐implement characteristics of the proposed algorithm.     

基于广义二次互相关的低信噪比信号时延估计

针对时延估计精度受噪声影响,导致时延估计不准确的问题,在现有广义二次相关算法的基础上,提出了一种改进的广义二次相关算法,通过将二次相关函数先做指数运算,降低噪声干扰,再将傅里叶逆变换得到的相关函数做高次方运算,达到锐化峰值提高时延正确率的目的。仿真结果表明,信噪比(SNR)在0~10 dB时,改进算法的均方根误差明显优于广义二次相关算法,正确率相比于广义二次相关算法也显著提高,且在更低SNR的情况下仍然具有一定优势。     

基于卷积神经网络的单目深度估计

针对利用深度学习方法对街道图像进行深度估计，提出采用语义分割的方法解决深度图出现边界模糊等问题；估计深度通过左右视角图生成视差图进行无监督的训练。在网络模型中添加语义分割层，采取多个空洞卷积并行的结构增加感受野，同时减少了图像下采样的次数，降低了由于下采样带来的信息损失，使得的结果更加准确。这也是在深度估计中首次与空洞卷积相结合增加准确率。通过对KITTI街道数据集进行训练，与现有结果相比，除了增加检测准确性，降低错误率之外，使得效果图中的物体更加清晰，并且在效果图中还保留了一些原模型中被忽视掉的细节信息，将原始图像更加完整的表现出来。     

井下安全阀力学性能分析

井下安全阀对安全性和可靠性要求较高。以渤海油田常用油管携带式井下安全阀为分析对象,对其工作压力、地面控制压力、全开启和全关闭压力、最大下入深度和自平衡能力等关键力学性能进行分析。井下安全阀的工作压力值与地面液控压力值相等,其选取依据为井口压力; 全开启和全关闭液控压力值可通过压力曲线法得到; 最大下入深度取决于弹簧安装时的预紧力; 自平衡能力与活塞面积和自平衡孔面积的比值有关,比值越大,自平衡能力越强。     

Skin depth and detection ability of magneto-optical imaging for weld defects in alternating magnetic field

To detect and evaluate weld defects, the skin depth and detection ability of magneto-optical imaging (MOI) for weld defects in alternating magnetic field were studied, and the application scope of MOI in alternating magnetic field was determined. A model of magnetic flux leakage testing (MFL) for weld defects is established by finite element method, and the reliability of the model is determined by law analysis and MOI experimental verification. Comparing MFL and eddy current detection with simulation method, the skin effect mechanism and rules of MOI for weld defects in alternating magnetic field are analyzed and obtained. Considering the effect of skin effect, different MFL detection models are established, and the detection ability of MOI for weld defects with different frequencies under corresponding models is studied and obtained. Finally, the MOI experiment is used to verify the results, which are consistent with the simulations. The skin effect rule and detection ability of MOI for weld defects obtained by simulation method can be used as the theoretical basis for practical experiments.     

基于神经网络的电动汽车电池SOC估算研究

论文首先分析了当前电动汽车电池管理系统中存在的问题,特别是电池电压的精确测量和剩余电量的准确预测问题一直亟待突破,因此,论文在分析电池荷电状态(SOC)影响因素的基础上,进行了动力电池的充放电实验,建立了BP网络电池模型,通过对网络进行训练,应用神经网络模型进行SOC估算,实验表明:建立的BP网络具有较好的适应性,能有效预测锂离子动力电池电压、电流和放电容量间的映射关系。可以准确地对电动汽车电池进行SOC估算。     

基于径流污染控制目标为目的的分流制初期雨水调蓄池容积确定

鉴于初期雨水污染为地表水体的主要污染源之一,以儒乐湖初期雨水调蓄池容积确定为例,采用调蓄池工程措施进行截流,以V类水体标准为径流污染控制目标,在《室外排水设计规范》中控制径流污染的分流制初期雨水调蓄池计算公式的基础上,应用模型法模拟得出控制目标下的初期雨水径流累积深度及调蓄池容积。结果表明,截留径流累积深度为7mm,初期雨水调蓄池容积分别为9 250、5 850m^3时,研究区域内能够满足在管网重现期下排水水质达到V类水体标准。研究成果可为初期雨水截流工程实施提供建议及参考。     

Distributed adaptive state estimation and tracking by using active-passive sensor networks

Heterogeneous sensor networks (HSN) find a wide range of applications in the field of military and civilian environments, where sensor nodes are utilized to estimate the position of a target with both dynamics and control input being unknown for the purposes of tracking. In the HSN, nodes are considered active depending upon their ability to sense the target output while the others are taken passive. Accurate estimation requires local information exchange among the spatially located sensor nodes, so that the active nodes as well as the passive nodes converge simultaneously to the same value. The local information exchange among the nodes is dictated by a connected graph. By using the criterion of collective observability, a novel distributed adaptive estimation scheme is introduced via adaptive observer where the nodes are allowed to have different sensor modalities. Using the estimated information, a subset of active and passive nodes, referred to as mobile nodes, can track the moving target. By using a constant state feedback controller at each mobile node, the state and parameter estimation as well as the tracking errors are shown to be uniformly ultimately bounded. Simulation results verify theoretical claims.     

Relationship between adjustment of low water level and utilization of water depth in Shashi Reach in middle Yangtze River

Hydrological, sediment, and bathymetric data of the Shashi Reach in the middle Yangtze River for the period of 1975–2018 were collected, and the characteristics of low water level changes and their impacts on utilization of water depth for navigation were investigated. The results showed that, during the study period, the Shashi Reach riverbed was significantly scoured and incised, with cross-sectional profiles showing overall narrowing and deepening. This indicated a strong potential to improve the water depth of the channel. The analysis of the temporal variation of in-channel topographical features showed that the Taipingkou diara underwent siltation and erosion, with its head gradually scoured and relocated downstream after 2008, and the Sanbatan diara continued to shrink and migrate leftwards. Low water levels with the same flow rate over the study period decreased. For instance, from 2003 to 2020, the water level at the Shashi hydrological station decreased to 1.37 m with a flow rate of 6 000 m³/s. Furthermore, the designed minimum navigable water level of the Shashi Reach was approximately 2.11 m lower than the recommended level. In terms of utilization of the channel water depth, continuous scouring of the river channel is expected to result in a reduction in discharge at the Taipingkou mouth, which will improve the water depth conditions of the channel during the dry season in the Shashi Reach. With several channel regulation projects, the 3.5-m depth of the Shashi Reach would basically be unobstructed. This promotes utilization of the shipping route from the Taipingkou south branch to the Sanbatan north branch as the main navigation channel during the dry season. Considering the factors of current water depth and the clear width limitation of the navigation hole at the Jingzhou Yangtze River Bridge, this route can still be favored as the main navigation channel with a 4.5-m depth during the dry season.     

Multi-task learning for video anomaly detection

We propose a multi-task learning framework for video anomaly detection based on a novel pipeline. Our model contains two crossing streams, one stream employs the backbone of Attention-R2U-net for future frame prediction, while the other is designed based on an encoder–decoder network to reconstruct the input optical flow maps. In addition, the latent layers of the two streams are merged together and assigned with a Deep SVDD-based loss at each location individually. Through the combination of these three tasks, the two-stream-crossing pipeline can be trained end-to-end to provide a comprehensive evaluation for the anomaly targets. Experimental results on several popular benchmark datasets show that our model outperforms the state-of-the-art competing models, which can be applied to different types of anomalous targets and meanwhile achieves remarkable precision.