Enhancing the precision of magnetocrystalline anisotropy energy estimation from Barkhausen Noise using a deep neural network

Recent works in the literature have demonstrated that the angular dependence of magnetocrystalline anisotropy energy could be estimated from magnetic Barkhausen noise in carbon steels using the energy of a Barkhausen signal band associated with the magnetic domain nucleation. However, although the estimations of this magnitude using the Barkhausen noise signal have practical advantages with respect to X-ray method, it is not as precise as the latter. This work proposes a method that uses a deep neural network for significantly enhancing the estimation precision of the shape of the angular dependence of the magneto-crystalline anisotropy energy using Barkhausen signal. Additionally, the proposed algorithm allows to estimate for the first time the amplitude of this angular dependence using the Barkhausen signal. The results are also compared with the determination of magnetocrystalline anisotropy energy using a feature selection method. The most significant features of the Barkhausen signal for computing the magnetocrystalline anisotropy energy are analyzed.