US 11,886,831 B2
Data sorting device and method, and monitoring and diagnosis device
Tamami Kurihara, Tokyo (JP); Kengo Iwashige, Tokyo (JP); Tetsuji Morita, Tokyo (JP); Shigeki Tounosu, Tokyo (JP); and Tadaaki Kakimoto, Tokyo (JP)
Assigned to Mitsubishi Heavy Industries, Ltd., Tokyo (JP)
Filed by Mitsubishi Heavy Industries, Ltd., Tokyo (JP)
Filed on Dec. 13, 2019, as Appl. No. 16/713,604.
Claims priority of application No. 2018-244277 (JP), filed on Dec. 27, 2018.
Prior Publication US 2020/0210144 A1, Jul. 2, 2020
Int. Cl. G06F 7/24 (2006.01); G05B 13/04 (2006.01); G06F 17/18 (2006.01); G06N 20/00 (2019.01); G05B 13/02 (2006.01)
CPC G06F 7/24 (2013.01) [G05B 13/0265 (2013.01); G05B 13/047 (2013.01); G06F 17/18 (2013.01); G06N 20/00 (2019.01)] 9 Claims
OG exemplary drawing
 
1. A computer-implemented method training a learning network to diagnose anomalies in plant equipment, the computer-implemented method comprising:
receiving operating data from sensors that monitor the plant equipment;
storing the operating data and a time when the operating data was collected by the sensors in a first database;
excluding abnormal values from the operating data to form filtered data;
training an electric characteristics learning unit using a first subset of the filtered data to generate first weighting factors;
training a vibratory characteristics learning unit using a second subset of the filtered data to generate second weighting factors;
training a thermal characteristics learning unit using a third subset of the filtered data to generate third weighting factors;
storing the first weighting factors, the second weighting factors and the third weighting factors in a second database;
creating a model for nominal operation using the first weighting factors, the second weighting factors and the third weighting factors that are stored in the second database;
receiving additional data from the sensors that monitor the plant equipment; and
diagnosing the anomalies in the plant equipment by inputting the additional data into the model for nominal operated created.