The goals of the present paper are to develop and demonstrate an efficient
technique for the design/analysis of electronic packages through a novel de
composition procedure. The ultimate utility of these techniques is to enabl
e quick and accurate design decisions at system-level, during package devel
opment by enabling one to develop a reusable library of modules in a manner
analogous to the object-oriented programming paradigm of modern computer s
cience. The methodology allows simultaneous design as well as domain decomp
osition and is based on a nonlinear optimization procedure that ensures the
approximate satisfaction of the principle of virtual work. The developed p
rocedure is demonstrated on a 5x5 hypothetical arrayed package. It is shown
that with the use of the decomposed solution methodology, approximately 35
0 percent improvement in computational efficiency is achieved at an accurac
y loss of only 6 percent. A windows-based graphical program founded on an a
rtificial neural network model for predicting life given shear and axial de
formation of solder joints was also developed. This neural network encapsul
ates the results of finite element analyses and predicts life for a given l
oading in a fraction of a second. [S1043-73978(00)00201-2].