Machine Learning-Based Evaluation of the Contribution Effectiveness in SoS Missions

This article presents an efficient and comprehensive evaluation method for aircraft contribution rate to system-of-systems (CSS) by using machine learning (ML). It is based on the whole design space rather than some typical points. After collecting data from mission simulation or analysis, this method simply divides the dataset into two groups (success and failure) and then efficiently trains ML models to do the CSS evaluation. Related literature and novelty are first discussed. Then, after introducing theory fundamentals about mission success and success index space, the proposed CSS evaluation approach is given step by step. Furthermore, two system-of-systems (SoS) simulation cases are provided for the method validation: SoS combat penetration mission (military application) and helicopter SoS rescue mission (civil application). In each case, two ML methods, neural network pattern recognition and support vector machine, are utilized to evaluate CSS. New samples were also collected to validate the trained models. Both methods can train ML models for SIS generation in a very short time. Finally, the comparison with other CSS methods is summarized.