Analysis of tail probability in stochastic RTM process simulation of FRP

Abstract

There is a growing need for molding process simulation of fiber reinforced plastics (FRP) in determining an appropriate set of process parameters, because a large number of process parameters exist and moreover those parameters have uncertainty or variability. Stochastic process simulations have been studied so far such as the Monte Carlo simulation (MCS), which provides us the expected value and standard deviation of the quantity of interest (QoI), considering the variability of input parameters. However, the results in the tail distribution were not highlighted except the authors' previous reports. This paper proposes a modified sampling scheme named stepwise limited sampling (SLS) to generate sampling points more efficiently and accurately in the multi-dimensional input parameter space, which lead to the tail distribution of QoI. The proposed method was applied to a resin transfer molding (RTM) process simulation considering 31 random parameters. Compared to the conventional MCS using 10,000 sampling points, it was demonstrated that enough number of cases in the tail distribution was analyzed by the modified method using only 1,700 sampling points.