ANN-based structure peciliaties evaluation of polymer composite reinforced with unidirectional carbon fiber

At the moment, there is a growing interest in composite materials with matrices based on thermoplastic polymers; these materials are superior to traditional carbon fiber plastics due to higher fracture toughness and impact strength, low smoke generation during combustion, and high thermal and chemical resistance. The deformation behavior of such composite materials due to the high plastic deformation of the matrix differs from the behavior of traditional composites, which must be considered when performing calculations. In our study, as a model object, the features of the microstructure of single carbon thread impregnated with polysulfone were studied in order to evaluate the distribution of misorientation angles of elementary fibers, the degree of their damage, the thickness of the polymer matrix interlayers between elementary fibers in threads impregnated with more viscous thermoplastic binders. Statistical analysis of a large array of micrographs was carried out using specially developed algorithms for computer processing of electron microscopic images of a composite material. Algorithms for processing arrays of images using machine vision algorithms are proposed. Based on the analysis of the data array, the distributions of carbon fibers impregnated with polysulfone were plotted over the misorientation angle of the filaments and over the interfilament distance in the longitudinal and cross sections of the fiber. Using the machine learning method, an algorithm for detecting violations of the filament structure was implemented. The data obtained can be used to refine the calculations of the strength and deformation characteristics of composite materials with thermoplastic matrices.