基于高频率分辨力谱估计技术与优化算法的异步电动机初发故障检测方法研究

The sensitive/reliable detection of incipient fault in induction motors has been of profound importance in improving system safety, reducing outage time and repair costs. Unfortunately, the current detection methods seriously suffer the following problem: they are all puzzled by the limit of Frequency Resolution and Measurement Time derived from Fourier Transform, on which they are generally based, and thus they always lose their sensitivity/reliability in much reality where unstable interference such as load fluctuation and random noise is inevitable. This has been the key problem in the detection of incipient fault in induction motors. Aiming to solve it, this proposal will deeply research High Frequency Resolution Spectrum Estimation Technique and Optimization Algorithm, and smoothly combine them to detect incipient fault in induction motors, sensitively and reliably. This objective will be implemented by the following work: (1) To present novel high Frequency Resolution Spectrum Estimation Techniques, which are suitable for extracting the features of incipient fault in induction motors, and the matching Optimization Algorithms; (2) Based on that, novel methods, sensitive/reliable for detecting incipient fault in induction motors, will be furthermore presented. It can be expected that the sensitive/reliable detection of incipient fault in induction motors will really become true through the above work, which is full of practical value and theoretical significance.

灵敏、可靠地实现异步电动机初发故障检测具有重要意义。但是，当前异步电动机初发故障检测方法广泛基于傅里叶频谱分析而存在局限性- - 无法摆脱频率分辨力/采样时长的限制，在负载波动、噪声等干扰情况下可能失效，"灵敏、可靠"无从谈起。这是当前亟待解决的难点问题、关键问题。为此，本项目将对高频率分辨力谱估计技术与优化算法展开研究并将其结合以实现异步电动机初发故障的灵敏、可靠检测。研究内容、意义如下：(1) 提出适于异步电动机初发故障特征提取的新型高频率分辨力谱估计技术以及与之匹配的新型优化算法。这具有工程应用价值，在信号处理领域/优化领域亦具科学意义。(2) 进而，提出基于高频率分辨力谱估计技术与优化算法的异步电动机初发故障检测新方法。将高频率分辨力谱估计技术与优化算法结合而应用于电机故障检测领域尚属首次，具有创新性。该方法可望真正实现异步电动机初发故障的灵敏、可靠检测，具有广阔工程应用前景。

将ESPRIT (Estimation of Signal Parameters via Rotational Invariance Technique)、MUSIC (Multiple Signal Classification)、APES (Amplitude and Phase Estimation of Signal)、MVDR (Minimum Variance Distortion Response)等高频率分辨力谱估计技术应用于异步电动机初发故障检测，结合异步电动机初发故障检测的特点与需求，通过仿真与实验，证明——就异步电动机初发故障检测而言，ESPRIT优于MUSIC、APES、MVDR。..进而，提出了有色噪声背景下的预滤波ESPRIT算法、四阶累积量ESPRIT算法，二者对背景噪声的免疫力优于常规的ESPRIT算法。..另一方面，将E-Prony (Extended-Prony)、PSA (Pattern Search Algorithm)、SAA (Simulated Annealing Algorithm)、粒子群算法 (Particle Swarm Algorithm)等优化算法应用于异步电动机初发故障检测，结合异步电动机初发故障检测的特点与需求，通过仿真与实验，证明——就异步电动机初发故障检测而言，E-Prony优于PSA、SAA以及粒子群算法。..在此基础之上，提出了基于高频率分辨力谱估计技术与优化算法的异步电动机转子与轴承故障检测新方法，并完成了实验室验证、现场实践检验。特别地，基于ESPRIT与E-Prony、通过拓展异步电动机转子故障与负荷波动之检测机理，首次提出了异步电动机转子故障与负荷波动协同识别方法，成功解决了异步电动机转子故障与负荷波动的灵敏、可靠检测问题。..上述基于高频率分辨力谱估计技术与优化算法的异步电动机转子与轴承故障检测方法业已集成于此前研制成功的异步电动机初发故障在线检测仪并应用于工程实际。..发表学术论文22篇，其中SCI检索1篇、EI检索11篇。业已获得发明专利授权6项。