Variable-Step Multiscale Lempel-Ziv Complexity And Its Application To Health Condition Identification And Damage Severity Assessment Of Rotating Machinery

Transportation vehicles,as an important part of transportation and industrial production fields,play an essential role in promoting the development of transportation and industry.Bearings and gears are the key rotating components in the transportation vehicles,whose health condition is crucial to ensure the efficient and safe operation of the transportation vehicles.However,most rotating components are prone to failure due to long-term harsh working conditions.Once a failure occurs,it can result in financial loss,vehicle breakdown,and even personal casualty,etc.Therefore,it is crucial to conduct the research on condition monitoring and fault diagnosis for key rotating components of transportation vehicles to guarantee the safety and stability.To realize the health condition identification and damage severity assessment for key rotating components of transportation vehicles,this dissertation focuses on solving the shortcomings of Lempel-Ziv complexity(LZC)in fault feature extraction and fault evolution characterization,etc.Given the LZC employs single-scale analysis leading to insuffcient mining capability,multiscale analysis-based LZC has inaccurate results at large scales,and the encoding method of LZC omits the amplitude modulation information,the strategies such as variable-step multiscale analysis and equiprobable space partitioning are studied and the more efficient ways to calculate and encode LZC are constructed.Therefore,the schemes of health condition identification and damage severity assessment are studied,which enables qualitative and quantitative diagnosis for key rotating components of transportation vehicles.The specific works are as follows:(1)A health condition identification method for rotating components based on variable-step multiscale Lempel-Ziv complexity(VSMLZC).To improve the capability of feature extraction and to solve the problem of insufficient fault information mining due to the single-scale analysis of original LZC and inaccurate LZC calculation results caused by significantly shortened sequence length using traditional multiscale analysis,the variable-step multiscale strategy is proposed by improving the coarse-grained process and VSMLZC is then constructed.Then,the grey absolute close relational degree is proposed to address the shortcomings that the tradional grey close relational degree is overestimated when processing the intersecting sequence.Based on the above work,the multiscale feature vectors extracted by VSMLZC are fed into the proposed relational model to identify the health condition of rotating components.The effectiveness and superiority of the proposed method is demonstrated by the analysis of experimental signals and the comparison with other methods.(2)A damage severity assessment method for rotating components using variable-step multiscale fusion Lempel-Ziv complexity(VSMFLZC).To visually describe the damage severity and fully utilize the multiscale complexity information,a feature fusion algorithm based on Laplacian score weighting is constructed.The features of each scale are evaluated using Laplacian scores and assigned the weights according to the importance.Thus,the VSMLZC sequence is transformed into a comprehensive VSMFLZC,which characterizes the complexity of the signal and realizes the damage severity assessment of rotating components.The effectiveness and superiority of the proposed method is verified by bearing single-point defect data,bearing run-to-failure data and gear wear data.(3)A damage severity assessment method for rotating components using equiprobable space partitioning-based variable-step multiscale fusion Lempel-Ziv complexity.Since the 0-1 encoding during the LZC calculation ignores the amplitude modulation information of signals.The encoding strategy of equiprobable space partitioning(ESP)is proposed for more comprehensive mining of fault features,aiming to retain more information related to the health condition of the rotating equipment and further improve the LZC’s the fault feature extraction capability and noise immunity.Therefore,ESP-based VSMFLZC is developed to achieve damage severity assessment of rotating components under more complex operating conditions.The effectiveness and noise immunity of the proposed indicator is velidated by a bearing dynamic model and exprimental data.Compared with other methods,the proposed ESP-based VSMFLZC obtained better results in damage severity assessment under stronger noise environment.In summary,this paper takes key rotating components of transportation vehicles as the research subjects,and takes LZC-based health condition identidication and fault severity assessment of rotating machinery as the research objectives.The ESP-based variable-step multiscale strategy and the Laplacian score-based feature fusion strategy are proposed in this dissertation,which solves the problems that LZC only adopts single-scale analysis with insufficient portrayal capability and multiscale analysis-based LZC has inaccurate results due to drastically shortened sequence length and overcomes the shortcomings that the 0-1 encoding method employed by LZC ignores modulation information.Based on these,novel complexity indicators are developed,which enhance the feature mining capability and signal complexity characterization of LZC.Eventually,qualitative and quantitative diagnosis for rotating components of transportation vehicles is implemented.