In this paper, we present two different optimization procedures, developed and implemented to deal with economic and constructive problems of free form grid shells. This kind of structures is generally composed by a supporting grid that defines the geometry of a large number of cladding glass elements always different one...
[More] In this paper, we present two different optimization procedures, developed and implemented to deal with economic and constructive problems of free form grid shells. This kind of structures is generally composed by a supporting grid that defines the geometry of a large number of cladding glass elements always different one from another. From the constructive point of view it means that every single piece needs to be designed and produced “ad hoc”, then marked and laid with the aid of an assembling table. Moreover if the grid is defined by four or more sides elements the realization of curve glass slabs turns out to be very expensive because of the carving process, forcing very often designers to try to avoid it by triangulating the grid where the shape results more complex. In order to reduce the cost of complex glass grid shells by limiting the number of element typologies and, if we have quadrilateral elements, defining at the same time a plane panels configuration, two optimization procedures have been developed. Starting with an analytic approach, it is impossible to define a set of rules that are able to generate efficiently a mesh corresponding to the previously set goals. For this reason, gradient based methods, such as the Force Density Method, as well as evolutionary methods, such as Genetic Algorithms, have been applied separately to benchmark shapes and a real case study, in order to compare their efficiency. All the free form surfaces are defined and handled by means of a commercial NURBS based software. Instead, its implemented VB based programming language has allowed to develop all the presented optimization procedures. Due to the smoothness of the solution domain of these specific problems, gradient based procedures seem to be the most efficient in the rapidity of convergence to the optimal or sub-optimal solution.
[Less]