# Fujitsu artificial intelligence calculates the geometry of magnetic materials

Modern methods of calculating the geometry of magnetic materials make it difficult to choose its optimal parameters due to the high nonlinearity index (the so-called magnetic hysteresis). Even after performing geometry modeling, errors occur when calculating magnetic losses, which can differ significantly from experimentally measured values. Fujitsu has developed an AI technology that automatically calculates the optimal geometry of magnetic materials. We will tell about this innovation in this article.

Materials that, when exposed to a magnetic field, act as a magnet, are used in various components and devices, including electric motors and inductors, which allow to store electricity in batteries. In this case, magnetism in itself causes a loss of energy. The level of magnetic losses seriously depends on the geometry of magnetic materials. As a result, it is directly related to the energy efficiency of a component or device. Therefore, to ensure high energy efficiency, it is very important to calculate the optimal geometry of materials taking into account magnetic losses.

Fujitsu has developed an AI technology that automatically calculates the geometry of magnetic materials in a virtual space to reduce energy loss. New development significantly increases the efficiency of design departments, allowing you to calculate the geometry of magnets for various applications, including power electronics and electric motors. Fujitsu technology shortens prototyping time from a few months to a few days.

With its help, you can accurately calculate the distribution of eddy currents that pass through the inductor. To do this, they must be represented as a formula for the dielectric effects of ferritic microstructures used as inductive materials. In the previously used estimation methods, there was a limitation in the accuracy of determining the size of the loss in eddy currents if the operating frequency of the inductor exceeded several tens of kilohertz. New development allows you to perform an assessment at a frequency of several megahertz.

By combining a new method for modeling magnetic losses with the genetic algorithm * Fujitsu created a formula for automatically searching for a set of geometric parameters. They have the Pareto-optimal form ** (dimensions for each part of the form of a magnetic material) and minimize the magnetic energy losses as much as possible. By 2020, Fujitsu plans to introduce design services on the market that will include the technology described above.

* Computational optimization method, working on the basis of the principles of biological evolution. For the existing generation of possible solutions, several copies are created, which are then crossed with each other and mutated. “Surviving” copies are selected to create the next generation of solutions. By repeating this process, get the best solutions.

** In the situation of minimizing several values that have a compromise ratio, and the absence of circumstances that would give smaller values for all variables, these parameters are called Pareto optimal. As a rule, there are several Pareto optimum, and the line or plane formed by these optima is called the Pareto optimal form.

Materials that, when exposed to a magnetic field, act as a magnet, are used in various components and devices, including electric motors and inductors, which allow to store electricity in batteries. In this case, magnetism in itself causes a loss of energy. The level of magnetic losses seriously depends on the geometry of magnetic materials. As a result, it is directly related to the energy efficiency of a component or device. Therefore, to ensure high energy efficiency, it is very important to calculate the optimal geometry of materials taking into account magnetic losses.

### The benefits of new technology

Fujitsu has developed an AI technology that automatically calculates the geometry of magnetic materials in a virtual space to reduce energy loss. New development significantly increases the efficiency of design departments, allowing you to calculate the geometry of magnets for various applications, including power electronics and electric motors. Fujitsu technology shortens prototyping time from a few months to a few days.

With its help, you can accurately calculate the distribution of eddy currents that pass through the inductor. To do this, they must be represented as a formula for the dielectric effects of ferritic microstructures used as inductive materials. In the previously used estimation methods, there was a limitation in the accuracy of determining the size of the loss in eddy currents if the operating frequency of the inductor exceeded several tens of kilohertz. New development allows you to perform an assessment at a frequency of several megahertz.

*On the left - simulation of inductor magnetic loss (distribution of magnetic flux density in a magnetic material). Right - comparison of experimental and simulated results.*### The practical benefits of innovation

*Results of inductor computer-aided design (each point corresponds to one of the variants of inductor geometry)*By combining a new method for modeling magnetic losses with the genetic algorithm * Fujitsu created a formula for automatically searching for a set of geometric parameters. They have the Pareto-optimal form ** (dimensions for each part of the form of a magnetic material) and minimize the magnetic energy losses as much as possible. By 2020, Fujitsu plans to introduce design services on the market that will include the technology described above.

* Computational optimization method, working on the basis of the principles of biological evolution. For the existing generation of possible solutions, several copies are created, which are then crossed with each other and mutated. “Surviving” copies are selected to create the next generation of solutions. By repeating this process, get the best solutions.

** In the situation of minimizing several values that have a compromise ratio, and the absence of circumstances that would give smaller values for all variables, these parameters are called Pareto optimal. As a rule, there are several Pareto optimum, and the line or plane formed by these optima is called the Pareto optimal form.