Research On Ferrograph Image Segmentation And Wear Particle Identification

With the development of modern industrial equipment toward large-scale, continuous and high reliability, the maintenance methods of equipment also change dramatically. Regular maintenance is transformed into condition maintenance, making increasingly high demand for condition monitoring and fault diagnosis of various types of machinery and equipment. In recent years, ferrography has been proven to be one of an effective technical means, and used in condition monitoring and fault diagnosis through qualitative and quantitative analysis for wear particles(wear debris). With the development of computer image processing,artificial intelligent and pattern recognition, ferrography which based on computer image processing and analysis will be further in-depth applications. Therefore, in order to improve the level of automation and intelligence of ferrography, we studied the ferrograph image processing and analysis methods, and proposed appropriate algorithms for wear particle edge detection, wear particle segmentation, wear particle recognition respectively. All algorithms have been validated and comparative analysis with the actual ferrograph images.To facilitate ferrograph image processing and analysis, we have developed the ferrograph image analysis system. The system consists of hardware devices and software platforms, of which software platform includes ferrograph image acquisition, image processing, image segmentation, and wear particle recognition modules and etc.Two new methods are proposed for the edge extraction of wear particles in ferrograph images. The first method is O&G which integrate Otsu algorithm with Grey relational analysis. In this method, Otsu algorithm is applied to separate wear particles from image background. Grey relational degree is used to detect edges of wear particles. The second method is IACO, which is based on improved ant colony optimization algorithm, color information is introduced into the ant colony algorithm. Therefore, both gradient values of pixels and color difference of neighboring regions are used to guide the moving of ants to search the edges of wear particles.Color image contains more information than grayscale image, we propose two methods which can be used to subtract background from image by using the color information. The first method extracts the background of the image based on the threshold of the ratio of color components. Because the ratio of color components of the background in ferrograph images is relative to fix, and has difference with that of wear particles, therefore, by optimizing the ratio threshold of background and wear particles, which will achieve the separation of them. This method is more applicable in forrograph images that the background is almost uniform. The second method is based on K-means color clustering, by using the K-means clustering of two dimensional color components in CIELAB color space, the wear particles could be segmented directly from the background of ferrograph images. This method can automatically select the cluster centers according to the background color, which meets the requirements of processing ferrograph images automatically.Although O&G and IACO algorithm can extract most of the wear particles edges, but there may be false edges and non-closed edges for some wear particles, and for deposited chains, it is difficult to accurately detect the internal contours. Therefore, in order to achieve accurate segmentation of wear particles, two integrated algorithms based on region segmentation and clustering are proposed in this paper. The first method is CWACA algorithm which combined marker-watershed and intelligent ant colony algorithm. In this method, marker-watershed is adopted firstly to obtain the initial separation of wear particles. The improved ant colony clustering algorithm is later used to merge the over-segmented regions. Next, the fiber ratio, one of the parameters of wear particles, is used for the discrimination and correction of clustering to fulfill the segmentation of wear particles. The second method is CMWGC algorithm which combining marker-watershed and an improved grey clustering algorithm. First, the marker watershed is applied to ferrograph images to efficiently obtain the initial segmentation of wear particles. Next, an improved grey clustering algorithm utilizing color characteristics and relative position information is applied to merge the over-segmented regions after watershed, the fibre ratio is used for the discrimination and correction of clustering to obtain the segmentation result accurately and correctly. The experimental results show that the two methods described above can be used to segment wear particles of different types, various sizes and blurred boundaries, such as deposited chains and abnormal large wear particles, in particularly, the CMWGC method has the advantages such as better segmentation result and fast calculation speed.In order to improve the accuracy and speed of debris identification, this paper proposes a new algorithm that combines principal component analysis and grey relational analysis(CPGA). First, principal component analysis is used to optimize the characteristic parameters of wear particles. Then, an improved grey relational analysis is used to discriminate between similar types of wear particles, such as severe sliding and fatigue particles. The experimental results indicate that the CPGA algorithm can successfully solve the information redundancy problem resulting from multiple parameters and proves to be a practical method to identify wear particles quickly and accurately.The results obtained in the research improve automation and intelligent level of ferrography, and provide a reference for its further extensive application.