Explainable AI for Engineering Design: A Unified Approach of Systems Engineering and Component- Based Deep Learning Demonstrated by Energy- Efficient Building Design

Component-based AI models improve reliability beyond training data: The research demonstrates that breaking down complex systems into interpretable components significantly enhances a model's ability to make accurate predictions on new building designs it hasn't seen before (R² = 0.94 vs. R² = 0.71 for conventional approaches), which is critical for real-world deployment where designs often differ from training examples.

Explainability can be built into AI architecture from the start: Rather than treating explainability as an afterthought, this approach embeds interpretability directly into the model's structure by ensuring that connections between components represent actual engineering quantities that domain experts can understand and verify, making the AI's reasoning transparent to decision-makers.

Aligning AI with established engineering practices increases trustworthiness: By structuring machine learning models to mirror how systems engineers already think about problems (hierarchical component systems), the approach makes AI tools feel more familiar and credible to professionals, potentially increasing adoption of AI-assisted design in engineering fields.

Domain knowledge integration enhances both performance and usability: The method combines deep learning with systems engineering principles and domain expertise, suggesting that AI systems designed specifically for professional fields perform better and are more useful than generic machine learning approaches applied broadly.