基于主元分析与动态时间弯曲的故障诊断方法及应用研究

轴承是工程实际中常用而又极易损坏的部件,特别是对其早期微弱响应的辨识,具有重要的社会价值和意义。为提高运转轴承的安全可靠性和可维护性,提出了基于主元分析与动态时间弯曲距离的故障诊断方法,它可以准确对早期微弱动态响应辨识、诊断。该方法首先将典型故障样本信号与待测信号小波去噪并EMD分解,并对若干固有模态分量主元分析求取主元,然后对主元分量进行分析,获得相关特征值组成特征向量,计算待测信号与已知故障样本信号特征向量的弯曲距离,弯曲距离越小表明两信号越相似,从而辨识故障。此外,还可将其应用于转子、碰磨、齿轮故障诊断中,工程应用实例表明该方法可以准确故障分类,高效故障诊断。     

基于仿生特征模式匹配的海洋动物声信号识别*

针对小数据量的海洋动物声信号混合识别,将声信号同态分析过程中的线性频率转换为Mel频率,得到模拟人耳听觉特性的Mel频率倒谱系数作为声信号的特征。按照声信号所属的物种建立特征模板,使用动态时间规整算法对待识别特征进行分类识别,并对特征库和识别算法进行优化。分别提取了6种鱼类、3种虾类、12种鲸类的Mel频率倒谱系数,为每个物种建立特征模板。分3次对3种、5种、6种鱼类进行识别,分别获得了100%、96.25%、94.68%的识别率。对6种鱼类、3种虾类、12种鲸类共21个物种进行混合识别,总识别率由87.56%提升至优化后的88.96%。实验结果表明,基于Mel频率倒谱系数和动态时间规整算法的海洋动物声信号混合识别能够在小数据量时获得较高的识别率,优化后的特征库和识别算法能够提升识别率。     

Passive IR field gradient detection of thermal objects in active Fresnel zones

A dual element Passive-Infrared (PIR) sensor is used with a rotating slit aperture to map a narrow scanning beam on the sensing elements through each lens of a Fresnel lens array. The stimuli generated due to each thermal object fall in the active Fresnel zones in a certain direction based on their locations and temperature variations on the surfaces of the sources. These signals are used to analyze the signatures of stationary thermal objects, their slight movements and thermal field gradient changes of the source surfaces provided the object projected area is less than the area of the active zone. Pattern matching is performed using Dynamic-Time-Warping (DTW) algorithm on STFT reduced length time vectors. The object space is divided into m-active zones that correspond to the Fresnel zones in a Fresnel lens array. Within passive IR region from each of the active zones, the system identifies not only the heat intensity changes but also detects the slight movement of the thermal source. The efficiency of the system is dependent on the number of active Fresnel zones and the angular separation between them. This single node PIR sensor system is designed to cover an angular view of ∼10° × 80° while horizontal Field of View (FOV) is divided into 4 active Fresnel zones. Generally costly Thermal-IR camera is used for thermal analysis. Our system is comparatively less costly and active coverage zones are easily configurable by increasing number of Fresnel lenses.