Antenna Design Surrogate Model Using Transformer Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for designing antennas are computationally intensive and time-consuming, especially for complex devices like artificial or virtual reality systems, due to the need for sequential simulations and high-resolution mesh representations, which are costly and slow to test multiple design iterations.
Innovation Solution
The use of machine-learning methods to derive a surrogate model that automates the generation of antenna designs by training on image representations, applying convolution and softmax functions, and leveraging transformer network architecture to predict frequency responses and handle non-linear relationships between antenna topology and resonances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sequential simulation methods are used for antenna design, then measurement precision and reliability are improved, but productivity and speed are significantly worsened
Solution Approach 1:
The patent creates a surrogate model that copies the behavior of the complex electromagnetic simulation system. This surrogate model is trained on simulation data to predict frequency responses, replacing the need for repeated full-scale electromagnetic simulations. The copying approach maintains prediction accuracy while dramatically reducing computation time, enabling rapid design iteration.
Solution Approach 2:
The patent performs preliminary action by pre-computing a training dataset through electromagnetic simulations before the actual design optimization process. This pre-computed data is used to train the surrogate model, which then handles subsequent predictions. By doing the heavy computational lifting upfront, the system enables fast iterative design without repeating full simulations.
2Measurement precision
If high-resolution mesh representations are used for accurate antenna modeling, then measurement precision is improved, but use of energy and computational cost are significantly increased
Solution Approach 1:
The patent applies partial action by using simplified geometric representations (such as boundary element models or reduced-order models) instead of full high-resolution mesh representations for the surrogate model inputs. This partial representation captures the essential electromagnetic behavior while requiring significantly less computational energy to process and store.
3Manufacturing precision
If multiple design iterations are tested through sequential simulations, then manufacturing precision is improved, but loss of time and productivity are significantly worsened
Solution Approach 1:
The patent creates a surrogate model that copies the electromagnetic simulation functionality, enabling rapid evaluation of multiple design iterations. This copied model can be queried thousands of times in the time it takes to run a single full simulation, allowing extensive design exploration and optimization without proportional time investment.
Solution Approach 2:
The patent performs preliminary action by pre-computing comprehensive training data that covers a wide range of antenna configurations. This pre-computed knowledge base enables the surrogate model to quickly evaluate new designs by comparing them against the training data patterns, rather than performing new full simulations for each iteration.
Data Source
AI summary
The disclosed computer-implemented method may include generating, using a machine-learning model of a computing device, a set of antenna designs. The method may also include tokenizing, by the computing device, each antenna design in the generated set of antenna designs. Additionally, the method may include predicting, by the machine-learning model of the computing device, a frequency response for each tokenized antenna design. Furthermore, the method may include comparing, by the computing device, the frequency response for each tokenized antenna design. Finally, the method may include selecting, by the computing device based on the comparison, an antenna design that meets a performance threshold for the frequency response. Various other methods, systems, and computer-readable media are also disclosed.