A Numerical Thermoplastic Analysis of Line Heating Processes for Saddle-type Shells with the Application of an Artificial Neural Network

Abstract

Line heating (LH) is a process for forming compound-curved shells of a ship's hull, and is carried out by skilled workers. The accuracy of final shape and the production time are based solely on the experience and intuition of the workers. Many attempts have been tried to analyze the LH mechanism theoretically and experimentally, in order to achieve productivity. The nature of the LH process involves a three-dimensional transient thermal conduction phenomenon, followed by temperature-induced permanent plastic deformation. Due to the complexity of the physical problem, a theoretical analysis is not presently available. Previous studies have been limited to simplified models or two-dimensional analyses, which are inadequate for applications in current shipyard practices. In addition, a final manufactured shape is dependent on many factors involved in the process, such as torch speed and position, type of heat, cooling method (air or water), as well as plate dimensions. The effect of each factor on a final deformed shape con not be obtained using simplified modeling. With a practical application in mind, a numerical approach was employed, to simulate the LH process. Based on the mechanics of LH, temperature and stress fields are uncoupled, and each field is solved using a general finite element program. Heat flux for the heating torch and convection condition for the cooling hose are modeled for temperature analysis. The plate to be fabricated is descretized using three-dimensional solid elements. In order to verify the validity of the present model, a temperature distribution is obtained for a flat plate problem and compared with published data. The results are in good agreement with the published data. A numerical simulation is then carried out for forming saddle-type shells according to current shipyard practices. Currently, the line heating is applied to a cylindrically curved shell to produce a doubly curved shape, that is saddle-type shell. Thus, singly curved shells are modeled to simulate the forming process of saddle-type shells and temperature distribution, permanent plastic deformation, and residual stress are calculated for the obtained final shells. Parametric studies are given and discussed relative to the effects of forming parameters, such as torch speed, cooling method, and plate dimensions. It should be noted that each piece of shell of a ship's hull is not identical. In addition, each ship is different. This means that every single piece of flat plate is fabricated using different combination of forming parameters. Thus, it is necessary to generate new parameters from the calculated results for the automation of the LH process. An artificial neural network (ANN) algorithm is applied to generate new parameters, and is verified with several actual examples.