The long-standing problem of finding an economical method of structural reanalysis is considered Recent developments in neural computing technology, coupled with substantial increases in processing speed and memory capacity, have provided a suitable background from which to develop an iterative method using a simple neural network in which, as the structure is changed, processing continues from the point at which the required target error was reached in the previous state of the structure.

There are several advantages in this approach when compared with the usual processing of the banded stiffness matrix, the introduction of additional structural joints is achieved simply by adding new nodes to the network, conversely the removal of existing joints is achieved by a simple device whereby the corresponding node in the network is disabled - this allows a subsequent re-insertion of the joint if desired. Thus the expansion and repacking of a stiffness matrix is avoided since the new nodes are simply appended to the end of the neural network layers.

Options are provided for the editing of member properties and for the addition or removal of members and the addition or removal of applied loads Similarly modifications to support conditions including the insertion or removal of supports is easily carried out.

The reanalysis of linear elastic structures is carried out in a straightforward way. for non-linear structures options are provided for large-displacement analysis and for structures containing cables with significant sag.

The purpose is to provide a method of reanalysis capable of direct integration into the structural design procedure and so remove the need to detach the two processes.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £75 +P&P)