Initialising occupancy for online simulation models with unknown service start times
针对在线仿真中服务开始时间未知的情况，推导了仅基于总服务时间分布和过去服务开始时间密度的部分服务时间分布公式，并通过实例验证其有效性，避免高达30%的仿真误差。
Initialisation concerns the setting of conditions at the start of a simulation. In a typical discrete event simulation process involving arrivals and services, setting current service occupancy is important, and requires the time until service end to be estimated for all subjects in service at the initialisation time. This can easily be performed through conditionally sampling from the total service time distribution – provided the times since service start are known for each subject. However, this cannot be assumed and there are many practical reasons why these partial service times may not be available at the point of initialisation. In this article, we provide the appropriate formulae for the partial service time distribution, based only on the total service time distribution and the temporal density of past service starts. In the case of time-homogeneous service starts, this depends only on the total service time distribution, with the result equivalent to the remaining service time distribution. Through illustrative examples, we demonstrate the validity of the formulae, and how naïve alternatives can produce simulation errors in excess of 30%. This article will be of particular interest to readers concerned with initialising simulation models for online use (also known as real-time or symbiotic simulation).