JOURNAL OF THE BRAZILIAN SOCIETY OF MECHANICAL SCIENCES AND ENGINEERING, cilt.40, 2018 (SCI İndekslerine Giren Dergi)
A new Mamdani-type fuzzy inference system with four input variables such as weld current, electrode force, a transformed variable of weld time + sheet thickness, and projection height was proposed for both DC 04 and DP 600 steel sheets to predict the degree of expulsion and spattering (conjugate) index (E) and the maximum push-out load of weld (weld strength) (F-k) in projection welding. In the computational analysis, trapezoidal membership functions were constructed for the fuzzy subsets of each model variable. The product (prod) and the center of gravity (centroid) methods were implemented as the built-in AND method and defuzzification method, respectively. Fuzzy logic outputs were compared with the predictions of multiple regression analysis-based models derived within the scope of this work. Model performances were quantified by means of various statistical performance parameters. Linear regressions between the outputs of the fuzzy logic model and the experimentally measured values yielded very high determination coefficients (R (2) = 0.967 - 0.994). The well-correlated results reveal applicability of the fuzzy logic model for predicting the expulsion and spattering (conjugate) index, and weld strength, as well as determining the optimal combination of the process parameters to conduct the projection welding operations resulted in desired weld quality.