Wide-range coil-less magnetic sensor based on amorphous wire driven by deep learning

To further improve the performance of giant magnetoimpedance (GMI) sensors, extending their measurement range beyond the linear response region has become a key challenge in this field. This study investigates the coil-less GMI and fluxgate characteristics of a cobalt-based amorphous wire with composition Co_67.11Fe_4.51Si_12.79B_15.59 and a diameter of 97.5 μm, and proposes a method for decoding the wire's voltage signals. When the amorphous wire is excited by a sinusoidal current of sufficient magnitude, the voltage waveform across its terminals is modulated by an external DC magnetic field applied along the axial direction. Analyzing voltage waveforms under different magnetic fields yields extensive field-related information. A one-dimensional convolutional neural network (1D-CNN) optimized by the Harris Hawks Optimizer (HHO) can accurately determine magnetic field strength, achieving a mean absolute error (MAE) of 0.0911 Oe under conditions of a full-scale range of 50 Oe and a calibration step of 0.1 Oe. Direct acquisition of the voltage waveform eliminates the need for coils, lock-in amplifiers, and similar structures, establishing an end-to-end mapping from voltage waveform to magnetic field. This method provides a novel technical pathway for overcoming the performance bottleneck of GMI sensors through hardware simplification and range extension, laying the foundation for embedded and densely arrayed sensing applications.