公开(公告)号：US11778916B2

公开(公告)日：2023-10-03

申请号：US17012074

申请日：2020-09-04

申请人： Wenzhou University

发明人： Liang Shu ,  Yeheng Zhang ,  Bo Li ,  Dingfang Chen ,  Quanguo Lu ,  Marcelo J. Dapino ,  Menglun Tao ,  Yanchao Zhu

IPC分类号： H10N35/80 ,  H10N35/00 ,  H02N2/00 ,  H02N2/04 ,  G06N3/04 ,  G06N3/08

CPC分类号： H10N35/80 ,  G06N3/04 ,  G06N3/08 ,  H10N35/00

摘要： A matching control method for mechanical impedance of a magnetostrictive precision transducer includes developing a three-layer neural network model corresponding to a Young's modulus of a Terfenol-D material; acquiring sample data to form a training sample set and a testing sample set; training the model using a Bayesian regularization training algorithm, and optimizing connection weights and thresholds among layers of the tested model, so as to obtain a final three-layer neural network model; based on the final model, building an inverse model of mechanical impedance of the magnetostrictive precision transducer; using a current level of impedance of a load as an input of the inverse model to obtain a bias magnetic field, and changing a level of the bias magnetic field by changing a bias current in an excitation coil of the transducer, thereby achieving adaptive matching between the mechanical impedance of the transducer and the impedance of the load.

公开(公告)号：US20210159386A1

公开(公告)日：2021-05-27

申请号：US17012074

申请日：2020-09-04

申请人： Wenzhou University

发明人： Liang Shu ,  Yeheng Zhang ,  Bo Li ,  Dingfang Chen ,  Quanguo Lu ,  Marcelo J. Dapino ,  Menglun Tao ,  Yanchao Zhu

IPC分类号： H01L41/06 ,  H01L41/12 ,  G06N3/08 ,  G06N3/04

摘要： A matching control method for mechanical impedance of a magnetostrictive precision transducer includes developing a three-layer neural network model corresponding to a Young's modulus of a Terfenol-D material; acquiring sample data to form a training sample set and a testing sample set; training the model using a Bayesian regularization training algorithm, and optimizing connection weights and thresholds among layers of the tested model, so as to obtain a final three-layer neural network model; based on the final model, building an inverse model of mechanical impedance of the magnetostrictive precision transducer; using a current level of impedance of a load as an input of the inverse model to obtain a bias magnetic field, and changing a level of the bias magnetic field by changing a bias current in an excitation coil of the transducer, thereby achieving adaptive matching between the mechanical impedance of the transducer and the impedance of the load.