US 12,306,596 B2
Additional learning device, additional learning method, and storage medium
Yuki Ueyama, Kyoto (JP); Yoshiki Ito, Kizugawa (JP); Yasuaki Abe, Takatsuki (JP); Shuji Inamoto, Nara (JP); and Nobuyuki Sakatani, Otsu (JP)
Assigned to OMRON Corporation, Kyoto (JP)
Filed by OMRON Corporation, Kyoto (JP)
Filed on Oct. 13, 2020, as Appl. No. 17/068,829.
Claims priority of application No. 2019-210657 (JP), filed on Nov. 21, 2019.
Prior Publication US 2021/0157285 A1, May 27, 2021
Int. Cl. G06N 20/00 (2019.01); G05B 13/02 (2006.01)
CPC G05B 13/028 (2013.01) [G06N 20/00 (2019.01)] 4 Claims
OG exemplary drawing
 
1. An additional learning device, comprising one or more processors configured to:
acquire observation information based on data observed in a manufacturing device that is controlled by an instruction in accordance with an output from a previously learned model outputting a first objective variable with respect to a first explanatory variable, wherein the first objective variable and the first explanatory variable are a pair of learning data used to train parameters of the previously learned model, wherein the observation information comprises a rotation speed of the motor of the manufacturing device;
determine whether or not a second explanatory variable corresponding to the observation information is unknown in the previously learned model, wherein the second explanatory variable is a feature value extracted from the observation information and represents a current physical state of a motor of the manufacturing device;
acquire a second objective variable corresponding to the second explanatory variable by interpolating values of the first objective variable included in the previously learning data in which values of the first explanatory variable corresponding to the values of the objective variable are known, correlate the second explanatory variable and the second objective variable, and retrain the parameters of the previously learned model with additional learn data including a pair of the second explanatory variable determined to be unknown and the second objective variable when the second explanatory variable is determined to be unknown to generate an updated learned model, wherein the first and second objective variables represent a predicted value of a future physical state of the motor of the manufacturing device;
adjust the rotation speed the motor of the manufacturing device in accordance to the predicted value of the future physical state represented by the second objective variable generated by the updated learned model in response to the second explanatory variable that is not included in the learning data used to train the previously learned model.