Knowledge-guided generative surrogate modeling for high-dimensional design optimization under scarce data
arXiv:2603.00052v1 Announce Type: new Abstract: Surrogate models are widely used in mechanical design and manufacturing process optimization, where high-fidelity computational models may be unavailable or prohibitively expensive. Their effectiveness, however, is often limited by data scarcity, as purely data-driven surrogates struggle to achieve high predictive accuracy in such situations. Subject matter experts (SMEs) frequently possess valuable domain knowledge about functional relationships, yet few surrogate modeling techniques can systematically integrate this information with limited data. We address this challenge with RBF-Gen, a knowledge-guided surrogate modeling framework that combines scarce data with domain knowledge. This method constructs a radial basis function (RBF) space with more centers than training samples and leverages the null space via a generator network, inspired by the principle of maximum information preservation. The introduced latent variables provide a p
arXiv:2603.00052v1 Announce Type: new Abstract: Surrogate models are widely used in mechanical design and manufacturing process optimization, where high-fidelity computational models may be unavailable or prohibitively expensive. Their effectiveness, however, is often limited by data scarcity, as purely data-driven surrogates struggle to achieve high predictive accuracy in such situations. Subject matter experts (SMEs) frequently possess valuable domain knowledge about functional relationships, yet few surrogate modeling techniques can systematically integrate this information with limited data. We address this challenge with RBF-Gen, a knowledge-guided surrogate modeling framework that combines scarce data with domain knowledge. This method constructs a radial basis function (RBF) space with more centers than training samples and leverages the null space via a generator network, inspired by the principle of maximum information preservation. The introduced latent variables provide a principled mechanism to encode structural relationships and distributional priors during training, thereby guiding the surrogate toward physically meaningful solutions. Numerical studies demonstrate that RBF-Gen significantly outperforms standard RBF surrogates on 1D and 2D structural optimization problems in data-scarce settings, and achieves superior predictive accuracy on a real-world semiconductor manufacturing dataset. These results highlight the potential of combining limited experimental data with domain expertise to enable accurate and practical surrogate modeling in mechanical and process design problems.
Executive Summary
The article presents RBF-Gen, a knowledge-guided surrogate modeling framework that integrates scarce data with domain knowledge to achieve high predictive accuracy in mechanical design and manufacturing process optimization. By leveraging the null space via a generator network, RBF-Gen encodes structural relationships and distributional priors during training, guiding the surrogate toward physically meaningful solutions. Numerical studies demonstrate RBF-Gen's superiority over standard RBF surrogates in data-scarce settings, including a real-world semiconductor manufacturing dataset. This work highlights the potential of combining limited experimental data with domain expertise, enabling accurate and practical surrogate modeling in complex design problems. The framework's ability to systematically integrate domain knowledge with limited data is a significant improvement over existing surrogate modeling techniques.
Key Points
- ▸ RBF-Gen integrates scarce data with domain knowledge for high predictive accuracy
- ▸ The framework utilizes a generator network to encode structural relationships and distributional priors
- ▸ Numerical studies demonstrate RBF-Gen's superiority over standard RBF surrogates in data-scarce settings
Merits
Strength in Handling Data Scarcity
RBF-Gen's ability to combine limited experimental data with domain expertise enables accurate surrogate modeling in data-scarce settings, a significant improvement over existing techniques.
Improved Predictive Accuracy
The framework's use of a generator network and encoding of structural relationships and distributional priors during training leads to superior predictive accuracy compared to standard RBF surrogates.
Demerits
Computational Complexity
The use of a generator network and encoding of structural relationships and distributional priors may increase computational complexity, potentially limiting the framework's applicability in certain scenarios.
Dependence on Domain Expertise
RBF-Gen's reliance on domain expertise may limit its applicability in domains where subject matter expertise is lacking or difficult to obtain.
Expert Commentary
The article presents a significant contribution to the field of surrogate modeling, addressing a critical challenge in mechanical design and manufacturing process optimization. RBF-Gen's ability to integrate scarce data with domain knowledge is a major improvement over existing techniques, and its application in various industries has the potential to revolutionize design optimization. However, the framework's reliance on domain expertise and potential computational complexity may limit its applicability in certain scenarios. Future research should focus on addressing these limitations and exploring the application of RBF-Gen in other domains. Additionally, the development of new policies and regulations regarding the use of artificial intelligence in design optimization is an important consideration.
Recommendations
- ✓ Future research should focus on developing more robust and accurate methods for encoding structural relationships and distributional priors in RBF-Gen.
- ✓ The application of RBF-Gen in various industries should be explored, including aerospace, automotive, and energy, to demonstrate its potential for improving design optimization and reducing the need for physical testing.