.jpg)
In the field of design and analysis of shell structures with complex geometries, the integration between computer-aided design (CAD) and computer-aided engineering (CAE) has historically been a challenge. Traditional methods often require data transfer processes between systems, introducing inefficiencies and approximation errors. Against this backdrop, SplineNet emerges, an isogeometric deep learning approach that promises seamless and accurate integration. SplineNet uses watertight spline representations, such as unstructured analysis-suitable T-splines, and Bézier extraction to build the neural network architecture, where Bernstein polynomials function as nonlinear activation functions. This innovation allows the network itself to incorporate energy formulations based on the Kirchhoff-Love model to calculate mechanical behaviors of shells, eliminating the need for time- and resource-intensive data exchanges.
The ability to work both data-free and data-driven opens a range of possibilities. In data-free mode, losses are derived directly from the system's energy, fitting perfectly into CAE workflows. In data-driven mode, SplineNet can act as an interpretable backbone in DeepONet architectures, enabling immediate results for new inputs without retraining the network or repeating the entire analysis process. This advancement has direct implications for the efficiency of design and simulation cycles in industries such as automotive, aerospace, or architecture. Companies developing custom applications and custom software find here an opportunity to offer faster and more robust simulation solutions to their clients.
At Q2BSTUDIO, we understand that the convergence between artificial intelligence and classical numerical methods is key for the next generation of engineering tools. Our teams work on developing AI for businesses and implementing AI agents that can learn from complex physical simulations. Furthermore, the computational infrastructure needed to train models like SplineNet can be efficiently managed with cloud services aws and azure, ensuring scalability and security. And not only that: cybersecurity becomes critical when handling sensitive design data, so we also offer cybersecurity solutions to protect digital assets. Likewise, industrial process optimization benefits from business intelligence services and tools like power bi, which allow real-time visualization of simulation results and performance analysis of complex structures.
In short, SplineNet represents a qualitative leap in shell simulation, merging the rigor of isogeometric analysis with the flexibility of deep learning. For companies seeking to adopt these technologies, having technology partners like Q2BSTUDIO, specialized in process automation and comprehensive software solutions, makes the difference between staying at the forefront or falling behind in an increasingly demanding market.
