基于双尺度非均匀传声器阵列的大型装备声学故障诊断技术研究

Considering the special dimensions and complex structures of a large equipment, it is difficult to rebuild a high-resolution sound field for the equipment in a limited space under reasonable budget using well-known existing audio-visual technology.Based on the idea that the key and common components of a large equipment require different image resolutions, a fault diagnosis system for large equipments using two-scale non-uniform microphone array is proposed in this project. The proposed fault diagnosis system creatively introduces large intervals to ensure the key components of the equipment align in the axial direction of the sub-arrays to gain higher local resolutions..The diagnositc methodology proposed in the project is detailed as follows: Firstly, the key noise sources are located by the Minimum Variance Distortionless signal Response (MVDR) or Estimation Signal Parameters via Rotational Invariance Techniques (ESPRIT) using the signals from a two-scale non-uniform microphone array. .Secondly, the high-frequency sound field information near the key noise sources is obtained by the far-field beam forming using MVDR,and the low-frequency information near the key noise sources is computed by the Patch Near-field Acoustic Holography(Patch-NAH) method. The Hidden Markov Model (HMM), Support Vector Machine (SVM) and other intelligent diagnostic techniques are adopted to diagnose the faults after the feactures of the sound field have been extracted..Using the two-scale non-uniform microphone array for the large equipment fault diagnosis optimize the microphone resources configuration in a limited space and under reasonable budget. Futhermore, a high-resolution sound field of the key components can also be achieved, and it is expected that the proposed project will provide an advanced theory and a novel method of the acoustic fault diagnosis for large equipments.

大型设备尺寸庞大、结构复杂，已有声像技术无法在有限空间与成本下重建大型设备的高分辨率声场，导致难以进行准确的故障诊断。本申请考虑到大型设备关键部位与普通部位成像分辨率要求不同等特点，针对性地提出了双尺度非均匀传声器阵列并将其用于大型设备的故障诊断。该阵列通过引入大间隔使得设备多个关键部位位于各子阵的轴线方向，以获得更高的局部分辨率。采用最小方差信号无畸变响应法、旋转不变信号参数估计法，应用双尺度非均匀传声器阵列对设备关键声源进行定位；在获得关键声源后，通过远场波束成形及局部近场声全息技术分别获取设备关键声源附近的高频与低频信息，并重建声场进行特征提取，最终应用隐马尔可夫模型、支持向量机等智能诊断方法进行故障诊断。基于双尺度非均匀传声器阵列的大型装备故障诊断方法通过对传声器资源的优化配置，重建大型设备关键部位的高分辨率声场进行故障诊断，有望提供一种新型的适合大型设备的声学故障诊断理论与方法。

声学检测具备非接触式、不影响设备正常工作、易于集成、可获得故障的发生位置信息等优点。但是大型设备尺寸庞大、结构复杂，实际诊断过程中难以用阵列一次覆盖。本项目通过构建双尺度非均匀传声器阵列，利用大型设备关键位置等已知先验知识，首先采用定位技术初步判断故障位置；进而应用声全息技术获取反映故障特征的设备关键部位的局部声像信号；从而给出适合大型装备非接触式声学故障诊断系统的构建方法与诊断方法。在项目实施过程中，课题组对于双尺度非均匀传声器阵列构建、基于远场声源定位的噪声定位技术、基于近场声全息的声成像技术、特征提取以及对应的故障诊断技术均进行了系统深入的研究。在定位研究方面提出了敏感度二次型约束的对角加载技术、结合盲源信号分离的广义特征分解技术；在双尺度非均匀传声器阵列构建方面研究了支持双侧声源的混合波叠加法、以十字阵为基础的传声器删减影响分析，最终提出了法线水平与法线垂直的双尺度非均匀阵列布局；在进场声全息方面给出了基于数据外推与内插的联合局部近场声全息算法以及双尺度非均匀阵列重建算法；在机械设备声场的基本分布规律方面利用Actron软件对机械结构的流固声耦合进行了仿真；在特征提取方面提出了信号MCA与Hilbert谱分析相结合的特征提取技术；在故障诊断方面给出了HMM模型自组织训练及去除个别干扰的算法。最终项目组建了双尺度非均匀传声器阵列的声学故障诊断系统，能够实现远场噪声定位、近场双尺度成像以及主动激振实验以及相应的分析诊断功能，并通过仿真分析以及故障轴承、实验电机的声学诊断对系统进行了实验验证。此外，课题组在研究传声器阵列对机械设备进行故障诊断的同时，进一步将其拓展到超声领域，研究了压电陶瓷片超声成像技术，取得了很好的效果。项目总体完成情况优秀。