SPOT WELDING QUALITY FUZZY CONTROL SYSTEM BASED ON MULTISENSOR INFORMATION FUSION

The multisensor information fusion technology is adopted for real time measuring the four parameters which are connected closely with the weld nugget size（welding current, electrode displacement, dynamic resistance, welding time）, thus much more original information is obtained. In this way, the difficulty caused by measuring indirectly weld nugget size can be decreased in spot welding quality control, and the stability of spot welding quality can be improved. According to this method, two-dimensional fuzzy controllers are designed with the information fusion result as input and the thyristor control signal as output. The spot welding experimental results indicate that the spot welding quality intelligent control method based on multiscnsor information fusion technology can compensate the influence caused by variable factors in welding process and ensure the stability of welding quality.

Sensor Fusion with Square-Root Cubature Information Filtering

This paper derives a square-root information-type filtering algorithm for nonlinear multi-sensor fusion problems using the cubature Kalman filter theory. The resulting filter is called the square-root cubature Information filter (SCIF). The SCIF propagates the square-root information matrices derived from numerically stable matrix operations and is therefore numerically robust. The SCIF is applied to a highly maneuvering target tracking problem in a distributed sensor network with feedback. The SCIF’s performance is finally compared with the regular cubature information filter and the traditional extended information filter. The results, presented herein, indicate that the SCIF is the most reliable of all three filters and yields a more accurate estimate than the extended information filter.

Applications of state estimation in multi-sensor information fusion for the monitoring of open pit mine slope deformation

The traditional open pit mine slope deformation monitoring system can not use the monitoring information coming from many monitoring points at the same time, can only using the monitoring data coming from a key monitoring point,and that is to say it can only handle one-dimensional time series.Given this shortage in the monitoring, the multi-sensor information fusion in the state estimation techniques would be intro- duced to the slope deformation monitoring system,and by the dynamic characteristics of deformation slope,the open pit slope would be regarded as a dynamic goal,the condi- tion monitoring of which would be regarded as a dynamic target tracking.Distributed In- formation fusion technology with feedback was used to process the monitoring data and on this basis Klman filtering algorithms was introduced,and the simulation examples was used to prove its effectivenes.

AN INFORMATION FUSION METHOD FOR SENSOR DATA RECTIFICATION

In the applications of water regime monitoring, incompleteness, and inaccuracy of sensor data may directly affect the reliability of acquired monitoring information. Based on the spatial and temporal correlation of water regime monitoring information, this paper addresses this issue and proposes an information fusion method to implement data rectification. An improved Back Propagation (BP) neural network is used to perform data fusion on the hardware platform of a stantion unit, which takes Field-Programmable Gate Array (FPGA) as the core component. In order to verify the effectiveness, five measurements including water level, discharge and velocity are selected from three different points in a water regime monitoring station. The simulation results show that this method can recitify random errors as well as gross errors significantly.

Sensor management based on fisher information gain

Multi-sensor system is becoming increasingly important in a variety of military and civilian applications. In general, single sensor system can only provide partial information about environment while multi-sensor system provides a synergistic effect, which improves the quality and availability of information. Data fusion techniques can effectively combine this environmental information from similar and/or dissimilar sensors. Sensor management, aiming at improving data fusion performance by controlling sensor behavior, plays an important role in a data fusion process. This paper presents a method using fisher information gain based sensor effectiveness metric for sensor assignment in multi-sensor and multi-target tracking applications. The fisher information gain is computed for every sensor-target pairing on each scan. The advantage for this metric over other ones is that the fisher information gain for the target obtained by multi-sensors is equal to the sum of ones obtained by the individual sensor, so standard transportation problem formulation can be used to solve this problem without importing the concept of pseudo sensor. The simulation results show the effectiveness of the method.

Neural Network Based Algorithm and Simulation of Information Fusion in the Coal Mine

The concepts of information fusion and the basic principles of neural networks are introduced. Neural net-works were introduced as a way of building an information fusion model in a coal mine monitoring system. This assures the accurate transmission of the multi-sensor information that comes from the coal mine monitoring systems. The in-formation fusion mode was analyzed. An algorithm was designed based on this analysis and some simulation results were given. Finally,conclusions that could provide auxiliary decision making information to the coal mine dispatching officers were presented.

Altitude information fusion method and experiment for UAV

Altitude regulation is a fundamental problem in UAV(unmanned aerial vehicles) control to ensure hovering and autonomous navigation performance.However,data from altitude sensors may be unstable by interference.A digital-filter-based improved adaptive Kalman method is proposed to improve accuracy and reliability of the altitude measurement information.A unique sensor data fusion structure is designed to make different sensors switch automatically in different environment.Simulation and experimental results show that an improved Sage-Husa adaptive extended Kalman filter(SHAEKF) is adopted in altitude data fusion which means that altitude error is limited to 1.5m in high altitude and 1.2m near the ground.This method is proved feasible and effective through hovering flight test and three-dimensional track flight experiment.

基于多传感器信息融合的机床测量数据自动补偿系统

机床测量数据自动补偿能够减少加工误差,为提高机床加工精度,设计基于多传感器信息融合的机床测量数据自动补偿系统。首先设计温度传感器、单片机主控与补偿脉冲发生装置,获取机床运动数据,然后采用多传感器融合算法融合采集信息,对信息中误差处理,采用粗糙集理论寻找最佳补偿值,实现机床测量数据自动补偿。实验结果表明,该系统能够有效减少加工误差,可以满足实际应用要求。

传感器信息融合下新能源汽车动力电池信号故障检测方法

新能源汽车作为节能环保的新产品具有较好社会前景,内在的动力电池是新能源汽车的主动力源,但电池发动机是一个较为复杂的系统,在处于恶劣环境时,可能出现各种故障问题。新能源汽车的动力电池若发生故障,不但会使汽车的系统性能下降,还会造成灾难性的后果,为此,研究传感器信息融合下新能源汽车动力电池信号故障检测方法。通过一致性定律整理电池系统传感器数据,在近似概率和频率中估算新能源汽车动力电池信号;选择熵权重法理论对数据信号进行区分,以时刻内单体电压作为评价指标,在预处理后构建判断信号故障矩阵;通过故障判断矩阵确定异常信号,在传感器信息融合算法下修订权值,以最大误差范围检测信号输出,检测新能源汽车动力电池信号故障,完成检测方法设计。实验以四组不同类型的新能源汽车作为测试对象,对其动力电池的运动工况进行信号模拟,在不同的接口处获取故障电压信号并完成检测测试,设计的电池信号故障检测方法能够实现精准的故障信号跟踪,完成较为精准的故障信号检测,具有一定的应用价值。

改进的多传感器信息融合算法及其应用

针对量测噪声未知的问题,提出一种改进的多传感器信息融合方法,首先对量测噪声进行实时跟踪与估计,接着基于估计的量测噪声进行最优加权融合,解决常规加权法权值不是最优的问题,最后将融合的结果进行卡尔曼滤波,得到系统的状态估计。将改进的多传感器信息融合方法应用某型无人机(UAV)垂向高度信息融合系统中,经仿真验证,表明该方法具有一定的工程实用价值。

多传感器信息融合的轴承故障迁移诊断方法

在重型装备低速、重载、强噪声环境下,采用单一传感器难以全面获取轴承的故障诊断信息,导致故障识别率低、识别不稳定,致使变工况下轴承故障迁移诊断失效。针对以上问题,提出了一种多传感器信息融合的轴承故障迁移诊断方法。首先,结合传感器的通道数,构建了堆叠卷积神经网络(MCNNs)提取各个通道的故障特征;然后,在MCNNs中引入最小绝对收缩与选择算子(Lasso),并通过网络反向传播完成了特征权值的更新,从而获得了多通道特征的融合;最后,利用源域数据对模型进行了训练,提取了故障特征,并完成了特征融合,采用损失函数完成了模型参数的优化,将源域训练得到的模型结果作为目标域的初始模型,利用目标域样本对初始模型的参数进行了微调,从而完成了模型迁移;并进行了信息融合效果、方法对比以及传感器信息采集属性的性能实验。研究结果表明:传感器的安装位置对信息融合影响较大,MCNNs+Lasso方法具有较好的特征融合效果,平均迁移诊断精度为99.03%,部分精度可达99.97%,在多个变工况的迁移任务中表现出较高迁移精度和良好的泛化性能。

ADAS系统视觉与毫米波雷达分布式抗差卡尔曼滤波融合算法

自动驾驶车辆常常利用多传感器对周围目标进行检测和跟踪,但受限于传感器多源异构特性和复杂多变的驾驶环境,准确的多目标检测和跟踪仍是实现自动驾驶的一大困难和挑战。本文面向高级驾驶辅助系统(ADAS)的多目标检测与跟踪任务,采用了基于1个视觉传感器和5个毫米波雷达(1V5R)的传感器配置方案,且设计了基于分布式抗差卡尔曼滤波器的多传感器信息融合算法以实现对周围目标的准确感知。首先,针对不同传感器数据特征,采用不同的线性卡尔曼滤波器和扩展卡尔曼滤波器进行数据融合,并基于分布式卡尔曼滤波建立了1V5R多传感器信息融合框架。其次,为降低传感器动态误差对于融合精度的影响,在卡尔曼加权观测融合的基础上,引入抗差估计方法,实现了对传感器动态误差的实时估计和修正。最后,通过离线仿真和实车道路试验对所提出的基于分布式抗差卡尔曼滤波的多传感器融合算法进行了验证。试验结果表明,与单一传感器的测量值相比,所提出的算法能有效融合多个传感器的信息以提升目标的检测与跟踪精度,且鲁棒性较好。

基于MSP430的照明风扇智能控制系统设计

随着生活水平提高,消费者对电器产品的智能化需求与日俱增。针对市面上风扇功能单一、智能化程度不高的问题,设计一款照明风扇智能控制系统,以TI公司MSP430单片机为控制处理中心集成多种传感器实现对环境参数和用户状态的多维度检测,实时处理数据信息与智能精细化控制,同时具备定时提醒、信息显示、无线遥控等系列功能。该控制系统具有低成本、低功耗、高智能等特点,可满足照明风扇的应用场景多样化、不同功能融合控制、绿色健康环保等应用需求。

基于时域多参数信息融合与机器学习算法的井架钢结构损伤识别

为避免井架钢结构在作业过程中由结构损伤引发的安全隐患,将时域多参数信息融合与机器学习算法进行结合,基于井架钢结构位移响应信号提出了一种基于主成分分析和随机森林算法的井架钢结构损伤识别方法。利用加速度传感器提取在冲击载荷下井架钢结构的加速度响应信号,由加速度响应信号处理得到位移响应信号,并提取包含脉冲因子、裕度因子和峭度的3个时域特征;运用主成分分析法将3个特征融合成1个新的综合性时域特征,将原始信号中包含的大多数信息保留下来;将处理过的数据输入随机森林算法模型进行井架钢结构损伤识别,整个过程只需要采集损伤后的信号,不需要采集损伤前的信号。利用以上损伤识别模型对井架钢结构有限元模型进行仿真计算,同时结合ZJ70井架钢结构实验室模型的实验结果进行分析。分析结果表明:利用上述损伤识别方法可以准确高效地判断出井架钢结构单一或多处损伤,判断结果的正确率为90%以上。

基于自适应多尺度注意力机制的CNN-GRU矿用电动机健康状态评估

利用多传感器信息融合技术进行电动机健康状态评估时,矿用电动机监测数据中存在异常值和缺失值,而卷积神经网络和循环神经网络等深度学习模型在数据质量下降严重的情况下难以有效提取数据特征和更新网络权重,导致梯度消失或爆炸等问题。针对上述问题,提出了一种基于自适应多尺度注意力机制的CNN-GRU(CNN-GRU-AMSA)模型,用于评估矿用电动机健康状态。首先,对传感器采集的电动机运行数据进行填补、剔除和标准化处理,并以环境温度变化作为依据对矿用电动机运行数据进行工况划分。然后,根据马氏距离计算出电动机电流、电动机三相绕组温度、电动机前端轴承温度和电动机后端轴承温度等健康评估指标的健康指数(HI),采用Savitzky–Golay滤波器对指标HI进行降噪、平滑、归一化处理,并结合主成分分析法计算的不同指标对矿用电动机的贡献度,对指标HI进行加权融合得到矿用电动机HI。最后,将矿用电动机HI输入CNN-GRU-AMSA模型中,该模型通过动态调整注意力权重,实现对不同尺度特征的信息融合,从而准确输出电动机健康状态评估结果。实验结果表明,与其他常见的深度学习模型CNN,CNN-GRU,CNN-LSTM,CNN-LSTM-Attention相比,CNN-GRU-AMSA模型在均方根误差、平均绝对误差、准确率、Macro F1及Micro F1等评价指标上更优,且预测残差的波动范围更小,稳定性更优。

多传感器非线性系统序贯观测融合二阶扩展卡尔曼滤波器

对多传感器非线性系统提出了一种序贯观测融合二阶扩展卡尔曼滤波(Second-order Extended Kalman Filter,SOEKF)算法。该算法的中心思想是根据传感器观测数据到达融合中心的先后次序依次进行处理。研究表明,提出的序贯观测融合SOEKF算法比集中式观测融合EKF算法具有更高的精度;与集中式观测融合SOEKF算法精度相当,且具有更低的计算复杂度。目标跟踪系统的仿真验证了算法的有效性。

多传感器信息融合的机械臂避障系统设计

为实现更准确、可靠的环境感知和避障决策,基于多传感器信息融合设计了一种机械臂避障系统。首先利用多传感器采集机械臂工作环境信息,经过数据层、特征层以及决策层处理,建立了多传感器下的机械臂人工势场模型,并加入引力函数、斥力函数和限制函数;其次通过设置传感器检测权重,融合障碍因素综合信息,实现多传感器信息融合下机械臂的避障;最后通过仿真实验证明,采用多传感器信息融合方法能够增加障碍物特征识别的成功率,提升避障成功率和路径规划效率。

多传感器信息融合的天线促动器故障诊断

促动器作为天线主动面的唯一调整装置,是保障天线反射面精度的关键部件,因此对促动器健康状态的监测至关重要。文中针对单个传感器诊断存在数据维度有限且现有工程数据稀缺的问题,提出了一种连续小波变换与组归一化并行卷积神经网络(Continuous Wavelet Transform-Group Normalization Parallel Convolutional Neural Networks,CWT-GPCNN)的故障诊断方法。首先建立CWT-GPCNN的故障诊断模型,引用组归一化技术加快网络收敛速度并提高诊断精度;然后通过评估超参数对模型性能的影响,确定诊断的最佳模型;最后,采用促动器传动系统实验数据集对所提方法进行验证,实验结果表明所建模型具有较好的泛化能力及多传感器融合的优越性。文中对多传感器融合与单传感器的诊断性能进行了比较,结果证明了多传感器融合诊断的优越性。此外,还对CWT-GPCNN模型与其他3种信息融合模型进行了比较。CWT-GPCNN模型的准确率高达93%,表明它具有良好的诊断性能。

无线传感器网络的多信道信息融合方法设计

在无线传感器网络组成的物联网中,受并行通信的干扰,导致多信道信息融合效果差、时延高、信息传送能耗大。为此,提出了针对物联网中无线传感器网络的多信道信息融合方法。通过树型拓扑结构合理分配物联网中传感路由节点信道,抑制并行传感通信的干扰。根据信息熵理论,获取各传感信道信息之间的距离差,采用欧氏距离得出各信息单元的距离熵,进行归一化处理,计算出各传感信息单元属性与融合的权重,按照线性加权算法融合各部分的信息,完成多信道信息融合。仿真结果表明:所提方法的多信道信息融合时延和信息传送能耗始终低于0.69 s和59 J,证明所提方法在传感网络多信道信息融合的效果较好,能够有效降低传感信息融合时延,减小传感信息传送能耗。