Wright Williams & Kelly
Factory Explorer® v2.3
Multiple Replications and
Summary: For simulation runs made with multiple replications, queue delay estimates on the cycle time contribution chart are biased high for replications after the first. To avoid this problem, do not rely on cycle time contribution chart output created from runs with multiple replications. Instead, use Factory Explorer® 's batch analysis capability to make multiple runs with different starting random number streams. This error is fixed in Factory Explorer® v2.3a and Factory Explorer® v2.4, beta 1.
Details: To estimate queue delay contribution for the cycle time contribution chart, Factory Explorer® keeps track of the number of lots in queue of each product type at each tool group in the factory. At the end of each analysis period, the average queue delay is calculated from the average queue length using Little's law. At the beginning of each replication, the number in queue variable should be reset to zero. In Factory Explorer® v2.3, however, this reset was not properly completed. The net result is that number in queue estimates (and hence queue delay estimates) are biased high for subsequent replications when multiple replications are performed. This product-specific queue delay estimate is only used by the cycle time contribution chart, so other output charts and worksheets should not be affected.
Workaround: Avoid using the cycle time contribution chart when multiple replications are performed. To mimic the effect of multiple replications, use Factory Explorer® 's batch analysis capability to make multiple runs, each with a single replication, but vary the starting random number stream (using the -FirstStream option). WWK is testing a maintenance release of Factory Explorer® , v2.3a, that will be shipped to customers with maintenance contracts within the next 2 weeks.