MULTI-GRANULARITY PARALLEL CNN MODEL-BASED EMG SIGNAL-TORQUE MATCHING METHOD

WIPO Patent Application WO/2021/169036

A1

A multi-granularity parallel CNN model-based EMG signal-torque matching method: step 1: collecting a torque signal and an EMG signal when a bolt is tightened; step 2: dividing a sensor range according to at least two granularities to generate a plurality of torque intervals corresponding to the granularities, and adding a torque label to each torque interval; step 3: generating an EMG corresponding to each time window; step 4: determining a torque label of each time window under the division of each granularity according to the torque interval into which an average torque value falls; step 5: generating a sample set; step 6: constructing a multi-granularity parallel CNN model, and using the sample set to train each independent CNN model; and step 7: inputting an EMG signal in an actual assembly process of an operator into a trained multi-granularity parallel CNN model to identify an assembly torque. The EMG signal is inputted into the multi-granularity parallel CNN model to obtain a torque value, which conveniently and accurately monitors the tightening torque in real time.

CHEN CHENGJUN (CN)
HUANG KAI (CN)
LI DONGNIAN (CN)
ZHENG SHUAI (CN)
HONG JUN (CN)

PCT/CN2020/088876

September 02, 2021

May 07, 2020

Click for automatic bibliography generation  

UNIV QINGDAO TECHNOLOGICAL (CN)
UNIV XI AN JIAOTONG (CN)

G01L5/24; A61B5/22

CN110210366A	2019-09-06
CN110672312A	2020-01-10
CN109816049A	2019-05-28
CN110146213A	2019-08-20
US20140182393A1	2014-07-03

FUZHOU KEYANG PATENT FIRM (CN)

View/Download PDF      PDF Help

Previous Patent: APPLICATIONS OF TRANSFERRIN EXPRESSION-DETECTING REAGENT IN PREPARING DIAGNOSTIC REAGENT OR REAGENT ...

Next Patent: ELECTRIC VALVE