ASSESSMENT OF USEFULNESS OF THE ADAPTIVE PARAMETER SELECTION TECHNIQUES FOR SUPPORT MANUFACTURING PROCESS OF ALUMINiUM-CERAMIC FIBRES COMPOSITES
Jarosław Durak, Marek Wojtaszek
Quarterly No. 4, 2010 pages 301-306
DOI:
keywords: composites, aluminium powder, ceramic fibres powder metallurgy, metal forming, mixing process, tensile strength, adaptive neuro-fuzzy systems
abstract An attempt to assess the possibility of adaptive techniques to estimate the beneficial and economic parameters of the forming process of composite particles in the system aluminum alloy-ceramic particles was carried out. As the input parameters for the computer analysis, the results of measurements of tensile strength of composite plastic samples in the warp of aluminium reinforced with ceramic fibres were used. Samples with different vo-lume share of fibres were made using powder metallurgy technology and the forming processing. Blending pro-cess, pressing of powder and mixtures at room temperature, and hot extrusion of compacts realized in isother-mal conditions for this purpose were used. Extrusion was carried out at different values of extrusion ratio and temperature. Results of studies on the effects of chemical composition, temperature and extrusion ratio on tensi-le strength of materials were used as the database to carry out the test analysis, using statistical and artificial in-telligence methods. Dependence of Rm on the share volume of fibers was estimated using curvilinear regression and adaptive neurofuzzy inference system ANFIS. This approach arose from the fact that in modern control sys-tems to better examine the methods based on neural networks, due to the higher speed operation and flexibility. Therefore it was assumed that the proposed comparison will give information about the possibility of their use in place of classical solutions. Based on the results of the analysis, authors concluded that the data obtained from tests of tensile strength measurements, witch amount was 72, were proved to be sufficient to carry out a test analysis using adaptive neurofuzzy inference system. It has been found highly compatibility of results obta-ined based on the proposed ANFIS method, with the results of the classical statistical ana¬lysis, carried out using the regression curves, which confirms the usefulness of this solution for computer aided design of composite materials. Based on a test analysis using the adaptive neurofuzzy system, the Rm parameter achieves the highest values when the participation of fibres in composite were between 4.0 and 5.5% by volume. The results obtain-ed confirm the adaptive techniques allow for automated correction of predefined relationship between the input parameters such as time and speed of mixing, the mixture composition, extrusion temperature, and output such as strength or density of the sample after the bench press.