Probabilistic simulation of multi-scale composite behavior

A methodology is developed to computationally assess the probabilistic composite behavior at all composite scales (from micro to structural) due to the uncertainties in the constituent (fiber and matrix) properties, in the fabrication process and in structural variables (primitive variables). The methodology is computationally efficient for simulating the probability distributions of composite behavior, such as material properties, laminate and structural responses. Byproducts of the methodology are probabilistic sensitivities of the composite primitive variables. The methodology has been implemented into the computer codes: Probabilistic Integrated Composite ANalyzer (PICAN) and Integrated Probabilistic Assessment of Composite Structures (IPACS). The accuracy and efficiency of this methodology are demonstrated by simulating the uncertainties in composite typical laminates and comparing the results with the Monte Carlo simulation method. Available experimental data of composite laminate behavior at all scales fall within the scatters predicted by PICAN. Multi-scaling is extended to simulate probabilistic thermo-mechanical fatigue and to simulate the probabilistic design of a composite redome in order to illustrate its versatility. Results show that probabilistic fatigue can be simulated for different temperature amplitudes and for different cyclic stress magnitudes. Results also show that laminate configurations can be selected to increase the redome reliability by several orders of magnitude without increasing the laminate thickness-a unique feature of structural composites. (C) 2003 Elsevier Ltd. All rights reserved.