针对传统声信号特征处理方法无法有效提取叶片声音特征、导致叶片故障诊断准确率低的问题,提出一种基于改进梅尔频率倒谱系数（MFCC）算法的风力机叶片故障诊断方法。首先采用快速傅里叶变换（FFT）分析不同风速下叶片声音信号和风噪的频率特性,明确叶片声音信号的频率分布区域,将全频段分为三部分;然后采用粒子群优化算法（PSO）对梅尔（Mel）函数在不同频段上的敏感度进行优化,在迭代过程中将MFCC算法提取的叶片声音特征进行聚类,以轮廓系数作为适应度函数;最后基于支持向量机（SVM）构建分类器,实现风力机叶片故障的准确识别。以华北某风电场的叶片声音采集数据为算例,考察该算法在不同风速工况下的适应性,验证该方法的有效性。
Abstract
The traditional acoustic signal processing methods cannot effectively extract the acoustic features of wind turbine blades and the fault diagnosis accuracy is insufficient. Therefore, a fault feature extraction method based on improved Mel frequency cepstrum coefficient （MFCC） algorithm is proposed. Fast Fourier transform is used to analyze the frequency characteristics of the acoustic signal from wind turbine blade and wind noise at different wind speeds, and the corresponding frequency distribution regions are obtained. Meanwhile, we divide the whole frequency band into three parts, and use particle swarm optimization （PSO） to optimize the sensitivity of Mel function in different frequency bands. In the iterative optimization process, the sound characteristics of turbine blades extracted by MFCC algorithm are clustered, and the Silhouette coefficient is taken as the fitness function. Taking the blade sound acquisition data of a wind farm in North China as an example, the adaptability of the algorithm under different wind speeds is investigated, and a classifier based on support vector machine （SVM） is constructed to achieve accurate identification of wind turbine blade faults, which verifies the effectiveness of the method.
关键词
风力机叶片 /
声信号处理 /
故障诊断 /
特征提取 /
梅尔频率倒谱系数
Key words
wind turbine blades /
acoustic signal processing /
fault diagnosis /
feature extraction /
Mel frequency cepstrum coefficient
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] SWENSON L, GAO L Y, HONG J R, et al.An efficacious model for predicting icing-induced energy loss for wind turbines[J]. Applied energy, 2022, 305: 117809.
[2] 张霆. 风力发电机叶片状态监测与故障诊断系统设计和实现[D]. 上海: 上海电机学院, 2016.
ZHANG T.Design and implementation of condition monitoring and bearing fault diagnosis system for wind turbine blades[D]. Shanghai: Shanghai Dianji University, 2016.
[3] XU D, LIU P F, CHEN Z P.Damage mode identification and singular signal detection of composite wind turbine blade using acoustic emission[J]. Composite structures, 2021, 255: 112954.
[4] XU D, LIU P F, CHEN Z P, et al.Achieving robust damage mode identification of adhesive composite joints for wind turbine blade using acoustic emission and machine learning[J]. Composite structures, 2020, 236: 111840.
[5] KIM H I, HAN J H, BANG H J.Real-time deformed shape estimation of a wind turbine blade using distributed fiber Bragg grating sensors[J]. Wind energy, 2014, 17(9): 1455-1467.
[6] SUN S L, WANG T Y, YANG H X, et al.Condition monitoring of wind turbine blades based on self-supervised health representation learning: a conducive technique to effective and reliable utilization of wind energy[J]. Applied energy, 2022, 313: 118882.
[7] REDDY A, INDRAGANDHI V, RAVI L, et al.Detection of Cracks and damage in wind turbine blades using artificial intelligence-based image analytics[J]. Measurement, 2019, 147: 106823.
[8] 杨婧. 基于谐波结构的乐器音色特征提取方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2018.
YANG J.Research on timbre feature extraction of musical instrument through harmonic structure[D]. Harbin: Harbin Institute of Technology, 2018.
[9] SOARES B S, LUZ J S, DE MACÊDO V F, et al. MFCC-based descriptor for bee queen presence detection[J]. Expert systems with applications, 2022, 201: 117104.
[10] BAWA P, KADYAN V.Noise robust in-domain children speech enhancement for automatic Punjabi recognition system under mismatched conditions[J]. Applied acoustics, 2021, 175: 107810.
[11] 覃光华. 人工神经网络技术及其应用[D]. 成都: 四川大学, 2003.
QIN G H.Artificial neural networks and its applications[D]. Chengdu: Sichuan University, 2003.
[12] 杨志凌, 姚治业, 王瑞明. 基于HMM的风电机组齿轮箱故障诊断研究[J]. 太阳能学报, 2017, 38(9): 2574-2581.
YANG Z L, YAO Z Y, WANG R M.Study on fault diagnosis of gearbox based on HMM[J]. Acta energiae solaris sinica, 2017, 38(9): 2574-2581.
[13] 平源. 基于支持向量机的聚类及文本分类研究[D]. 北京: 北京邮电大学, 2012.
PING Y.Research on clustering and text categorization based on support vector machine[D]. Beijing: Beijing University of Posts and Telecommunications, 2012.
[14] 郑楠, 于广亮, 刘娟楠, 等. 一种基于SVDD和PCA的风机故障诊断方法[J]. 自动化技术与应用, 2019, 38(11): 143-146.
ZHENG N, YU G L, LIU J N, et al.A fault diagnosis method for fan based on SVDD and PCA[J]. Techniques of automation and applications, 2019, 38(11): 143-146.
[15] WANG Y X, TAN J X, QIN X L, et al.Wind turbine blade fault diagnosis based on MFCC feature optimization strategy[J]. IOP conference series: earth and environmental science, 2021, 675(1): 012074.
[16] 王宇星. 基于MFCC与原型聚类的风机叶片故障诊断研究[D]. 天津: 河北工业大学, 2020.
WANG Y X.Research on wind turbine blade fault diagnosis based on MFCC and prototype clustering[D]. Tianjin: Hebei University of Technology, 2020.
[17] 彭晓庆. 基于K-Means与原型网络的两阶段聚类算法及应用[D]. 太原: 山西大学, 2020.
PENG X Q.Two-stage clustering algorithm based on K-means and prototype network and its application[D]. Taiyuan: Shanxi University, 2020.
基金
河北省自然科学基金创新群体项目（E2020202142）
PDF(2051 KB)
