Recently, Data driven design is one of latest design method to take advantage of emerging AI technology, especially metamodeling has applied in various fields and driven its influence on many industries.
However, the main difficulty lies the selection of good metamodeling for design system and the definition of the hype-parameters without the expertise knowledge. Therefore, this study focuses on developing the AI system that automatically selects an appropriate metamodel and its hyper-parameters without the knowledge of design techniques.
AI system should be constructed into two stages: First, both the extraction of features and definition of labels are performed by using big data from the mathematical and engineering problems. Second, AI system should be trained by machine learning algorithm (EDT or MLP). In order to demonstrate the performance of AI system, the accuracy of AI system was calculated by applying several test problems.
In conclusion, AI system can be suggested the best metamodel with its hyper-parameter by only requiring input the data with no need of designer’s knowledge.