Electromagnetic simulation - Models and optimization

The design of a device has long been a question of the experience acquired by the engineer in charge of the job. While this experience is still an asset today, it is no longer sufficient to find a solution that matches the specifications, or what is more commonly referred to as "the specification". The main reason for this is the growing complexity of devices, which requires an increasing number of modifiable parameters (degrees of freedom) and very extensive specifications that can impose constraints on several output parameters. This, of course, stems from the demand for multi-performance systems or devices. For example, an antenna may require high gain, while at the same time requiring a side lobe level below a certain value for different pointing angles. In this case, a sufficient number of degrees of freedom will be required to achieve all objectives simultaneously. Moreover, the number of degrees of freedom will have to increase with the number of output quantities, i.e. those to be optimized. It would therefore be unthinkable to use a trial-and-error approach, varying input parameters on a purely intuitive basis. Even if the designer's experience enables him to start with a solution that delivers performance not too far from what is expected, the final stage, which involves adjusting the many parameters to achieve an optimum solution, could take several years, depending on the case. There are two reasons for this: the possibilities for adjustment remain numerous, and the implementation of a prototype for each test run would prove terribly costly. Even the use of a model (which should be as rigorous as possible), for performance verification rather than prototype production, does not remove the difficulty in finding a strategy for adjusting input quantities.

For example, digital electromagnetic calculation tools are analysis tools capable of taking into account coupling and electromagnetic radiation effects as rigorously as possible. However, their use implies a calculation time that is often exhaustive. It therefore seems illusory to insert them directly into an optimization process. This is one of the weak links in the computer-aided design (CAD) procedure, although many improvements have been made to make these analysis models faster. With certain constraints on the validity domain, alternative, faster approaches can replace the electromagnetic model. For example, equivalent device model techniques have been developed. A structure can advantageously be replaced by an equivalent electrical circuit, whose output parameters can be calculated relatively quickly. Alternatively, empirical formulas can be derived from measurement or previously simulated results. A more flexible model consists in building a database storing, in a multi-dimensional space, the values of the output variable...