| Multi-Stage procedures are developed to build up simulation experiments for estimating Cycle time-throughput (CT-TH) curves of manufacturing systems where the response of the curve could be the mean, higher moments, or percentiles of cycle time. Simulation experiments are designed in such a way that the overall variability of the response estimators of the CT-TH curve are minimized. The procedure keeps collecting data until a user-specified precision is achieved for the estimated curve.; To estimate the moment CT-TH curve, which quantifies the relationship of moment of cycle time to throughput rate, a nonlinear regression model was developed to represent the underlying CT-TH curve implied by a manufacturing simulation model. A straightforward extension of this allows us to estimate the first three moment curves (1st 2 nd, and 3rd moment of cycle time vs. throughput) simultaneously, based on which we proposed an indirect method for estimating CT-TH percentile curves.; A highly flexible distribution, the generalized gamma, is used to represent the underlying distribution of cycle time. To obtain CT-TH percentile curves, we fit metamodels for the first three CT-TH moment curves over the throughput range of interest, determine the parameters of the generalized gamma by matching moments, and obtain percentiles by inverting the distribution. Numerical results are presented to demonstrate the effectiveness of the approach. This method is free of the huge data storage and computational effort that is usually required for percentile estimation. |
