Abstract:Since the establishment of Maxwell's equations in the 19th century, computational electromagnetics has undergone a century of stable development [...]
engineering, electrical & electronic,computer science, information systems,physics, applied
What problem does this paper attempt to address?
This paper aims to explore the applications of computational electromagnetics techniques and artificial intelligence in the engineering field, and specifically solves the following problems:
1. **Limitations of traditional computational electromagnetics algorithms**: Traditional algorithms are unable to effectively utilize historical data to mine underlying physical laws when facing design - optimization scenarios that require a large number of repetitive calculations, resulting in slow iteration speeds; at the same time, a large amount of intermediate calculation results and cache data are generated during large - scale simulation tasks, wasting computing resources.
2. **Improving simulation efficiency and intelligence level**: By introducing artificial intelligence algorithms, non - linear mapping relationships are automatically extracted from electromagnetic simulation parameters to the required electromagnetic responses, making full use of historical data information, improving simulation efficiency, and promoting the development of intelligent electromagnetic computing.
3. **Specific challenges in optimal design**:
- **Filter design**: A new method combining multi - objective evolutionary algorithm (MOEA/D) and one - dimensional convolutional auto - encoding network structure (1D - CAE) is proposed to reduce the time of filter design.
- **Routing protocol optimization**: A hybrid location - service strategy is designed, including forwarding strategy selection, handover design, etc., as well as the data packet transmission process based on the reactive routing protocol, which solves the routing problems in small unmanned aerial vehicle systems (UAS), especially in flying ad - hoc networks (FANETs).
- **Calculation of radiation characteristics of antenna arrays**: An improved method combining the method of moments (MoM) and the uniform theory of diffraction (UTD) is proposed, which avoids the problem of unreliable UTD results when the equivalent phase center of the antenna is blocked, and significantly reduces the root - mean - square error.
- **Ground - loss resistance of the VLF umbrella - shaped antenna grounding system**: The influence of introducing a current - collecting ring on the ground - loss resistance of the VLF umbrella - shaped antenna grounding system is studied. By establishing a geometric decomposition model and using Feko software for simulation, the effectiveness of the current - collecting ring structure is verified.
- **Improvement of the diode SPICE model**: Based on measurement and non - linear fitting stochastic optimization algorithm, an improved diode SPICE model is proposed, which improves the accuracy of the model and can truly reflect the electromagnetic interference characteristics.
- **High computational cost in filter optimization**: An adaptive online - updated one - dimensional convolutional auto - encoding (AOU - 1D - CAE) surrogate model is proposed, combined with the particle swarm optimization (PSO) algorithm, which improves the data collection efficiency and prediction performance, avoids falling into local optimal solutions, and improves the optimization efficiency.
- **The influence of gate - coupling on electronic devices**: The influence of front - door and back - door coupling on electronic devices is comprehensively analyzed by using the finite - difference time - domain method (FDTD). An improved equivalent circuit model based on the physical model is introduced, and protection suggestions are proposed.
- **Shielding effectiveness test of irregular cavities**: A method based on the electromagnetic topology model is proposed to comprehensively characterize and test the shielding effectiveness of irregular cavities.
- **Design of multi - band planar antennas**: A new type of fork - shaped triple - band planar antenna (NTTPA) is introduced, which is suitable for multiple wireless communication frequency bands, has a compact size and a stable radiation pattern.
- **Pre - design of multi - band planar antennas**: Artificial neural networks are used to quickly select the geometric shapes suitable for the required working frequency bands, simplifying and accelerating the antenna design process.
- **Design of balanced band - pass filters**: Two designs of balanced band - pass filters based on different resonators are proposed, which have good frequency selectivity, high common - mode rejection ability and wide stop - band characteristics.
These contributions jointly promote the applications and development of computational electromagnetics and artificial intelligence in the engineering field.