基于群智能的永磁同步电机故障诊断

发布时间：2018-11-12 20:32

【摘要】：自动化和智能化是当今工业系统的发展趋势,而系统的稳定性是自动化和智能化实现的前提。永磁同步电机(Permanent Magnet Synchronous Motor,PMSM)同时具有高效率、高功率密度以及强鲁棒性等性能,现代化工业领域已离不开PMSM的应用,尤其是在精密控制领域的应用。当电机发生故障而未能及时发现处理时,轻则电机本身损伤,重则损坏整个电机设备造成巨大的经济损失,因此对PMSM故障诊断的研究十分有必要,具有重大意义。而PMSM故障中最为常见的故障分别是驱动系统开路和定子匝间短路故障。本文采用群智能优化算法对PMSM驱动系统开路进行和PMSM定子匝间短路故障进行诊断研究。首先,本文在矢量控制的基础上,建立PMSM静态坐标下数学模型和dq轴数学模型,介绍电机的矢量变换原理,然后分别分析PMSM驱动系统开路和定子匝间短路故障状态下的数学模型。然后针对PMSM驱动系统开路故障,提出一种基于自适应二阶粒子群算法(Self-adaptive SECond-order Particle Swarm Optimization,SASECPSO)的改进极限学习机(Improved Extreme Learning Machine,IELM)算法。该SASECPSO算法采用自适应惯性权重策略及线性变化认知系数方法,提高二阶粒子群算法(SECond-order Particle Swarm Optimization,SECPSO)的收敛速度和收敛精度。此外,运用SASECPSO算法同时对极限学习机的输入权值和隐含层阈值参数优化,可提高极限学习机算法在PMSM故障中的识别率。以电机转速和ABC相电流作为多源样本数据,多组实验证明IELM算法相对于其他算法具有较高的诊断精度。最后针对PMSM常见的匝间短路故障,利用能量频谱分析提取特征向量,采用自适应动态猫群算法(ADAptive dynamic Cat Swarm Optimization,ADACSO)优化SVM的惩罚因子和核函数参数,随后将优化后的SVM用于电机故障诊断。以小波能量频谱得到的特征向量作为SVM算法的样本数据来进行仿真实验,结果表明,相对于其他优化算法,采用ADACSO优化SVM参数能够使SVM在PMSM故障诊断中具有更高的诊断精度和准确率。
[Abstract]:Automation and intelligence are the development trend of industrial system, and the stability of system is the premise of automation and intelligent realization. Permanent magnet synchronous motor (Permanent Magnet Synchronous Motor,PMSM) has high efficiency, high power density and strong robustness at the same time. Modern industry has been inseparable from the application of PMSM, especially in the field of precision control. When the motor fails to find and deal with the fault in time, the light motor itself will be damaged, and the heavy motor equipment will be damaged. Therefore, the study of PMSM fault diagnosis is very necessary and has great significance. The most common faults in PMSM are open circuit fault of drive system and short circuit fault of stator turn. In this paper, the open circuit of PMSM drive system and the fault diagnosis of PMSM stator inter-turn short circuit are studied by using swarm intelligence optimization algorithm. Firstly, on the basis of vector control, the mathematical model and dq axis mathematical model under PMSM static coordinate are established, and the vector transformation principle of motor is introduced. Then the mathematical models of PMSM drive system under open circuit and stator interturn short circuit are analyzed respectively. Then an improved extreme learning machine (Improved Extreme Learning Machine,IELM) algorithm based on adaptive second-order particle swarm optimization (Self-adaptive SECond-order Particle Swarm Optimization,SASECPSO) is proposed for the open circuit fault of PMSM drive system. The SASECPSO algorithm adopts adaptive inertial weight strategy and linear varying cognitive coefficient method to improve the convergence speed and accuracy of the second-order particle swarm optimization (SECond-order Particle Swarm Optimization,SECPSO) algorithm. In addition, using SASECPSO algorithm to optimize the input weights of LLM and threshold parameters of hidden layer at the same time, the recognition rate of LLM algorithm in PMSM fault can be improved. The speed of motor and the phase current of ABC are used as multi-source sample data. Many experiments show that the IELM algorithm has higher diagnostic accuracy than other algorithms. Finally, for the common inter-turn short circuit faults in PMSM, the eigenvector is extracted by energy spectrum analysis, and the penalty factor and kernel function parameters of SVM are optimized by adaptive dynamic cat swarm algorithm (ADAptive dynamic Cat Swarm Optimization,ADACSO). Then the optimized SVM is used in motor fault diagnosis. Using the eigenvector obtained from wavelet energy spectrum as the sample data of SVM algorithm, the simulation results show that, compared with other optimization algorithms, Using ADACSO to optimize SVM parameters can make SVM have higher diagnostic accuracy and accuracy in PMSM fault diagnosis.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TM341

【参考文献】