公开(公告)号：US11255719B2

公开(公告)日：2022-02-22

申请号：US16662688

申请日：2019-10-24

申请人： ANRITSU CORPORATION

发明人： Masaaki Fuse ,  Ken Shioiri ,  Takao Tanimoto ,  Hideyuki Sakamoto

IPC分类号： G01J1/02 ,  G01J1/04

摘要： A material property inspection apparatus includes a conveyance unit, a light source, an irradiation unit, a light receiving unit, a signal detection unit, a material property value input unit, an inspection set value input unit, and a processing unit. The processing unit calculates a relation equation between the material property value from the material property values of the plurality of test pieces and the light intensity of the transmitted light or the reflected light of respective test pieces, calculates the material property value of the inspected object from the light intensity of the transmitted light or the reflected light detected by the signal detection unit and the relation equation, compares the calculated material property value of the inspected object with the inspection set value inputted from the inspection set value input unit, and determines the quality of the inspected object.

公开(公告)号：US11481673B2

公开(公告)日：2022-10-25

申请号：US16433592

申请日：2019-06-06

申请人： Anritsu Corporation

发明人： Ken Shioiri

IPC分类号： G06N20/00 ,  G06T7/00

摘要： A highly reliable signal analysis is performed by suppressing over-learning and appropriately distinguishing an abnormal signal. A signal analysis device 1 includes a data storage unit 4 which stores digital data of a measured signal, a training data storage unit 11 which stores training data to which a label of normality or abnormality is attached, a plurality of machine learning classifiers 5 which identify whether the measured signal is normal or abnormal by comparing the digital data of the measured signal stored in the data storage unit 4 with an identification boundary value and output identification results according to machine learning by the training data, and a determination unit 6 which determines whether the measured signal is normal, abnormal, or uncertain by comparing the identification results of the plurality of machine learning classifiers 5 with a determination threshold value.