—In order to provide fast, accurate, and low cost shock sensitivity calibration of acceleration sensors, a pneumatically operated piston shock calibrator based on ISO 16063-22 is designed. comparative calibration method is used to calibrate the shock sensitivity of tested sensor within a shock peak amplitude of 25 to 2000 g and a pulse duration of 0.1 to 3 ms. The pneumatically operated piston shock calibrator includes modules of mechanical shock, shock signal generation, and measurement control based on the NI data acquisition card and LabVIEW control program. Numerical analysis and calculation of the pneumatic ejection process by MATLAB is performed to analyze the influence factors and levels of the impact velocity. Moreover, experiments on key performance indexes and test data processing are carried out. The repeatability of shock peak amplitude and pulse duration are better than 1.24 and 2.19%, respectively, which reaches the level of the commercially available calibrator of the same type. Meanwhile, the calibrated uncertainty is 2.56% ( k = 2), which meets the ISO 16063-22 requirement with the accuracy of less than 5%.
刀具磨损监测是推动数控机床稳定运行的有效手段.为实现刀具健康实时监测,设计一种基于长短记忆网络的刀具磨损监测方法.将实时采集的切削力信号进行去噪、降维等预处理;用堆叠自编码器进行特征提取,获得影响刀具磨损的本质特征;用长短记忆网络构建刀具磨损监测模型,实现加工过程中刀具磨损监测.利用铣削实验数据进行实例验证,获得磨损量平均绝对误差为 0.060 80 的铣刀磨损量监测模型,验证了该方法的有效性.
采用飞秒激光对SiC/SiC复合材料进行环形扫描刻蚀去除试验,分析脉冲能量、光斑重合度及线重合度对加工区域材料残留及环形凹槽外轮廓的影响.结果表明:光斑重合度、线重合度和脉冲能量显著影响光束扫描面积内的能量密度,材料去除量随扫描能量密度的增大而增大.在较小光斑重合度和线重合度条件下,即能量密度较小时,刻蚀槽外轮廓呈现锯齿状,且扫描轨迹间有较多材料残留.随着光斑重合度和线重合度的增大,外轮廓及底面趋于光滑.
基于高精度遥操作设备和协作机器人搭建一种新型医疗超声检测平台,可用于远程超声检查等应用领域.研究基于双边末端的位姿映射方法、双边力反馈方法与运动保护机制以及动态冲击力反馈与实时感知规避的柔顺控制方法,实现应用于远程医疗超声平台的遥操作机器人随动控制与临场力感知效果,保证遥操作医疗超声平台的实时平稳操作.通过实验验证:该平台具有较强的可行性和稳定性,可为未来超声检测提供新的工作范式.
针对基于几何变换(GT)的异常检测忽略了特征紧凑性,难以准确区分高度相似的合格与缺陷字符的问题,定义秩正则损失鼓励紧凑特征表达,提出基于秩正则几何变换(RRGT)的字符缺陷检测算法.无监督异常检测仅使用正常样本(合格字符)作为训练集,RRGT对训练样本进行不同类型的几何变换.定义秩正则损失训练网络模型识别正常样本的变换类型.秩正则迫使同类变换的特征矩阵具有较小的秩,从而鼓励特征紧凑性.推理阶段使用网络预测施加在新样本上的变换类型.由于网络未使用异常样本(缺陷字符)训练,网络对异常样本的预测不确定性高,由不确定性可以检出异常样本.在旺旺牛奶盒喷点字符上的实验表明:相比GT,RRGT大幅提升了网络对缺陷字符预测不确定性的识别能力,缺陷检测的AUC值提高了 4.8％.
