GANを用いた高勾配型磁力発生時の時系列データによる微量常磁性体の磁化率評価

抄録

The cathode material of lithium-ion batteries contains magnetic foreign material, which is separated by magnetic separating. However, because the cathode material and magnetic foreign material are weakly magnetic, they are all adsorbed by a high magnetic force separator, resulting in inadequate separation. There is a method to separate them by adjusting the magnetic force by clarifying the difference in magnetic susceptibility, but it is difficult to measure the magnetic force because of the small quantity of samples. Vibration of the measuring instrument also caused a lot of variation in the measured values, resulting in a large error margin. A system was needed to measure a small amount of magnetism from the small quantity of samples, remove the noise, and correct it to derive a value with a small error. In this study, the magnetic field analysis by the finite element method is used to form a uniform high-gradient magnetic field to measure the slight weight change caused by the magnetic field, and the measurement accuracy is improved by suppressing the noise during the measurement of the electronic balance with Generative adversarial network (GAN) from the aging data to suppress the variation, and the magnetic susceptibility is evaluated from a small amount of sample.