Estimating simulation workload in cloud manufacturing using a classifying artificial neural network ensemble approach

Cloud manufacturing (CMfg) is an extension of cloud computing in the manufacturing sector. The CMfg concept of simulating a factory online by using Web services is a topic of interest. To distribute a simulation workload evenly among simulation clouds, a simulation task is typically decomposed into small parts that are simultaneously processed. Therefore, the time required to complete a simulation task must be estimated in advance. However, this topic is seldom discussed. In this paper, a classifying artificial neural network (ANN) ensemble approach is proposed for estimating the required time for a simulation task. In the proposed methodology, simulation tasks are classified using k-means before their simulation times are estimated. Subsequently, for each task category, an ANN is constructed to estimate the required task time in the category. However, to reduce the impact of ANN overfitting, the required time for each simulation task is estimated using the ANNs of all categories, and the estimation results are then weighted and summed. Thus, the ANNs form an ensemble. In addition to the proposed methodology, six statistical and soft computing methods were applied in real tasks. According to the experimental results, compared with the six existing methods, the proposed methodology reduced the estimation time considerably. In addition, this advantage was statistically significant according to the results of the paired t test.