Study Of Numerical Analysis Approach For Wear Particles

Recently, the description and recognition of the status of tribological system has attracted more and more attention from the points of both research and industrial application. For a machine in operation, the properties of the tribological system are relevant to the machine condition. The study of the tribological system relies on the analysis of a large number of output parameters such as wear loss, the topography of surface, characterization of wear particle, etc. The wear mechanisms can be determined through the considerable, dynamic and scientific measurement, description and analysis of tribological systemic properties. Among these analytical approaches, it is an important aspect to conduct wear debris analysis.The thesis has first studied the moving video image of the wear particles in an oil pool. Then, a qualitative research on the movement, deposition, rolling and selfhold characterization of the wear particles has been investigated. To avoid the overlap of wear debris, the oil pool is considered to be laid vertically during the experiment.The dissertation mainly focuses on the digitalization of wear debris. First, through analyzing the process of image photographing, a set of software has been developed. The software includes image acquisition, image pretreatment, image filtering, image segmentation, image labeling and calculation of parameters. A summarization of the shape parameters is given after the analysis of the references from domestic and overseas. Benefited from these approaches, a comprehensive study of the outline, texture and color characters of wear debris is carried out. For the analysis of the outline character, a program has been developed which can automatically search the outline points of the wear particles in the image and store the coordinate of the outline points. Then, according to the character of the outline, the Fourier transforms of the outline and the power chart of different frequency are calculated. The relationship between different frequency and the outline shape is studied. Furthermore, the fractal description is used to analyze the character of the outline. Last, according to the curvature expression of continuum function, the properexpression of outline curvature is analyzed. A curvature expression model is proposed and the proper parameters of curvature are used to describe Wear particles, including kurtosis, skewness, average variance and average difference. The correlation among all these parameters of four different types of wear debris is conducted. In the analysis of surface texture, the maximum and minimum brightness point in texture normal line is examined firstly. Then, the differences among some typical wear particles are analyzed through analyzing the texture gray level co-occurrence Matrix, color principal component analysis of the wear particle and the direction of wear particle outline. Through the study of the principal component analysis of color, the ferromagnetic and non-ferromagnetic particles are obvious different and easy to distinguish.For the 3D surface topography of wear debris, the depth of field of microscope has a significant influence on the formation of image. A comparison method has been proposed to compare in focus part and out of focus part of the images. The method first uses one-order and two-order differential filter operators and wavelet image analysis to process the two-dimensional image sections. Then, a threshold is given according to the above pretreatment. Next, image segmentation is performed to separate the in-focus part of the images. Thus, logical operations are used to conduct the multi-focus image merging. The 3D surface topography of wear particles is gained. This method can overcome the disadvantages of scanning electronic microcopy (SEM) which can only get the gray images. Through applying the proposed method in this thesis, color surface texture of wear debris can be gained.For the quantitative analysis using ferrograph, the oil samples of the 1000-hours reliability test of EQD210-10 diesel engine have been studied. Benefited from the research of image processing, the binarization of ferrograph image is carried out. According to the character of microscope images, a new segmentation method based on the compensation of the background has been developed. Then a quantitative analysis of the wear debris cover area ratio is given to these images. The results are compared to the PQ index of the oil samples. Linear regress analysis has showed that they have preferable correlation.As it is impossible for the computer to fully recognize the wear type from justone particle character, it is necessary to analyze the relationship among the wear particles characters including shape, outline, texture, color, size and wear rate through information fusion. Each parameter has its own advantage in the description of a certain type of wear particle. Through information fusion analysis, a reasonable relationship between the numerical characters of wear particles and wear rate is established. In the analysis of wear particle classification, clustering analysis has been done based on the analysis of some numerical parameters. Moreover, the curvature parameters of the wear outline are analyzed through fuzzy mathematic method. A fuzzy reasoning approach is applied to these parameters and it gives a good classification result. Based on these studies, a set of software integrated the functions of wear particle image acquisition, pretreatment, segmentation, parameter calculation, texture analysis and recognition has been developed. It has demonstrated that it is a useful tool for further research of wear debris analysis.