Session: 01-05-02: Applications of Artificial Intelligence/Machine Learning for Aerospace Structures 2

Paper Number: 152678

152678 - Composites Ai: An Ai-Empowered Composites Expert System 

This research introduces Composites AI, a novel computational tool that integrates the constitutive modeling capabilities of SwiftComp for aerospace structures and materials made of composites with OpenAI’s GPT API, available across iOS, Android, and web platforms developed by Flutter. By utilizing GPT’s function-calling capabilities, Composites AI offers an intuitive chat interface that allows users to engage with complex topics in composite analysis, including lamina and laminate properties, such as engineering constants and stress/strain distributions. Upon receiving a user query, Composites AI automatically activates SwiftComp to conduct precise calculations, generating structured outputs tailored to the specific needs of aerospace composite applications, thereby enhancing both the accuracy and efficiency of composites design and analysis.

To minimize the number of APIs and to maximize the reusability of APIs, Composites AI adopts a bio-inspired design. As the nature used hundreds of elements form the varieties substance of our universe, we break the functionalities of SwiftComp into the very basic functional APIs, may be in the order of hundreds, then GPT can be trained to link these basic units to form complex workflow to solve more and more complex problems.

Composites AI also incorporates a feedback mechanism to continuously refine GPT’s responses based on user input. This iterative process enables the AI model to progressively align with user preferences and application-specific requirements, creating a personalized and adaptive simulation tool based on generative AI.

There is a unique design of Composites AI that it also allows training by real experts in addition to normal users. This is achieved by two steps. First, the Composites AI is released to experts for them to train Composites AI using human language. This training will include providing expert answers to typical questions and training Composites AI to link basic simulation elements to form complex simulation workflows to solve more complex problems. Second, we will also implement a group chat feature to Composites AI so that real composites experts can team up with GPT to answer questions from users. This feature will serve at least two purposes: acquiring more data for fine tuning of GPT model and harvest intelligence of real composites experts, which are not accessible from the publications of these experts, and digitalize them in Composites AI so that Composites AI will becoming more and more intelligent in addressing questions and providing answers to users.

By combining artificial intelligence with advanced engineering simulation tool, Composites AI provides a cross-platform solution that significantly streamlines traditionally complex composite material analysis workflows. This tool underscores the potential of AI-enhanced systems to optimize engineering processes, supporting more informed decision-making in aerospace composites design.

Presenting Author: Banghua Zhao Purdue University

Presenting Author Biography: Currently working as a senior iOS Engineer at MEGA The Privacy Company
Ph.D. Candidate in AAE at Purdue University in Professor Wenbin Yu's group