Development of Bolted Flange Design Tool Based on Finite Element Analysis and Artificial Neural Network

Abstract

In bolted flange connections, commonly utilized in aircraft engine designs, structural integrity and minimization of the weight are achieved by the optimum combination of the design parameters utilizing the outcome of many structural analyses. Bolt size, the number of bolts, bolt locations, casing thickness, flange thickness, bolt preload, and axial external force are some of the critical design parameters in bolted flange connections. Theoretical analysis and finite element analysis (FEA) are two main approaches to perform structural analysis of bolted flange connections. Theoretical approaches require the simplification of the geometry and are generally oversafe. In contrast, finite element analysis is more reliable but at the cost of high computational power. In this paper, a methodology is developed for iterative analyses of bolted flange that utilizes artificial neural network approximation of a database formed with more than ten thousand non-linear analyses with contact algorithm. In the design tool, a structural analysis database is created by taking permutations of the parametric variables. The number of intervals for each variable in the upper and lower range of the variables is determined with the parameters correlation study in which the significance of parameters are evaluated. The prediction of the ANN based design tool is then compared with FEA results and the theoretical approach of ESDU. The results show excellent agreement of the ANN based design tool with the actual non-linear finite element analysis results within the training limits of the ANN.