Optimizing Parameters Contributing to Riveting Quality Using Imperialist Competitive Algorithm and Predicting Objective Function via Three Models MLR, RBF, and ANN-GA

Metal sheets play an important role in the mechanical design, particularly in the aerospace structures. Various parameters affect the quality of this operation. In this paper, optimization of the parameters contributing to the riveting quality in order to minimize the value of the maximum tangential stress in metal sheets is addressed. hence, the tolerance of the hole diameter in top and bottom sheets, the friction coefficient, and the tolerance of the rivet diameter and the rivet length were considered as the parameters influencing the riveting quality. A total of 64 models were obtained by permutations of the parameters, two at a time. The outputs were determined using finite element method. The objective function for optimization is the maximum tangential stress for which there is no analytical relation. Thus, three methods including the multivariable linear regression (MLR), the artificial neural network model of the radial basis function (RBF) type, and the hybrid model of the artificial neural network and the genetic algorithm (ANN-GA) were employed to model this function. Further, the performance of the three models was compared and the most suitable one was selected to model the objective function. The regression model was used to model the values of the height and the diameter after riveting. The imperialist competitive algorithm is utilized to solve this optimization problem. The obtained value for the maximum tangential stress using the imperialist competitive algorithm is 16368 pounds per square inches. After modification, this value increased to 23440 pounds per square inches using the finite element method. After riveting, the height and diameter of the rivet were respectively measured to be 0.07689 inches and 0.18524 inches.