Research On Device Fault Diagnosis Method Based On Full Vector Sparse Coding

The safe and stable operation of machinery and equipment is not only related to the realistic benefit of many companies,but also closely related to the life safety of the majority of mechanical equipment operators.Therefore,whether it can accurately and comprehensively extract the fault features of the equipment and identify the fault types of the equipment with high efficiency and high accuracy,and then formulate targeted measures to ensure the safe and stable operation of the equipment is particularly significant.At present,the fault diagnosis technology based on vibration signal has the widest range of applications.The diagnostic principle is to use the sensor to obtain the vibration signal when the equipment is running,and then to extract the information that can reflect the fault characteristics from the vibration signal,and then diagnose the equipment fault.The traditional fault diagnosis technology based on vibration signal is mostly based on the signals collected from single channel.Therefore,there is a defect that information is not used comprehensively,which may easily cause misdiagnosis or missed diagnosis.The full vector spectrum technique is an information fusion method based on homologous two-channel signals,which can effectively improve the information missing from single-channel analysis.Therefore,it can improve the comprehensiveness of fault feature extraction to a large extent.This technology has been gradually applied to engineering practice.In the event of a failure of a rotary machine,its vibration signals tend to be nonlinear,non-stationary,and non-Gaussian.Conventional signal analysis methods have significant limitations in processing the above types of signals and therefore cannot effectively extract fault features.Sparse coding originates from compressed sensing theory.It is a linear combination of a series of basis functions under the effort of a group of highly redundant basis functions.It can realize the efficient and flexible representation of original signals with a small number of obvious components.The sparse coding method can adaptively match its impact characteristics according to the characteristics of the fault signal,providing a powerful tool for fault feature extraction and diagnosis based on the vibration signal.Based on the above description,this paper aims at the problems existing in the traditional mechanical equipment fault diagnosis methods and draws on the technical advantages of the homogenous information fusion as well as compressive sensing,combines the full vector spectrum technology with the signal sparse coding method,and proposes a fault diagnosis method for rotating machinery equipment based on full vector sparse coding.The main contents of this paper include the following aspects:(1)Aiming at the theoretical and practical problems faced in the development of fault feature extraction and fault type recognition methods for rotating machinery equipment,the research course and significance of homologous information fusion technology are described.Summing up the research status of fault feature extraction and intelligent fault identification methods for mechanical equipment in recent years,the advantage of signal sparse coding method applied in the field is derived,and the possibility and effectiveness of combining sparse coding with homologous information fusion technology are analyzed,and the research content of the paper is established.(2)The theoretical basis and numerical calculation process of the full vector spectrum technique are introduced in detail,and related application examples are given to illustrate the effectiveness of the method.The basic concepts of signal sparse coding methods,sparse coding correlation solving algorithms,and construction of overcomplete dictionaries are introduced in detail.Then resonance sparse decomposition method is introduced,and its theory,thoughts and application characteristics are comprehensively studied.The contents of the two chapters serve as the main theoretical support for the full text,laying the foundation for the follow-up specific research content of the paper.(3)A device fault feature extraction method is proposed by combining the full vector spectrum with sparse coding.First,homologous two-channel signals are each separated into a low-resonance component that reflects a transient impact component and a high-resonance component that reflects a periodical harmonic using resonance-based signal sparsity decomposition.Then,the low-resonance component sub-band reconstruction signal obtained by the sparse decomposition of the two channels is subjected to full-vector information fusion.Finally,the enveloped demodulation analysis is performed on the fused signal to extract the fault feature frequency.The effectiveness of the method is verified by simulation signals and experimental signals.(4)Aiming at the feature shift phenomenon of time domain signal sparse coding and the problem of single channel analysis often result in inadequate use of information,a new method for fault identification of rolling bearings was proposed by combining the full vector spectrum technique and sparse coding.Firstly,full vector information fusion of the homogeneous dual-channel signal of the rolling bearing in each state was carried out.Then,the main vibration vector signals obtained was applied to construct all kinds of redundant dictionaries.Finally,these dictionaries were employed to reconstruct the test samples,and the reconstruction residual was taken as the criterion to identify the status of these samples.Through converting the time-domain signals into main vector signals,the information contained in the training samples is more comprehensive and accurate.Besides,the feature extraction step can be eliminated,so it can reduce the effect of human factors.The test results demonstrated that the proposed method,which has high efficiency as well as good practicability,and can effectively identify the fault pattern of rolling bearing.