Application of nuclearization free -noise automatic encoder in planetary gear fault diagnosis

Use the characteristics of nuclearization of similar data to each other and alien data away from each other to select the best noise rate of versatile automatic encoder. Secondly, the inertial weight and threshold of the hidden layer of the exfoliating automatic encoder is trained layer by layer to build a feature vector set. In terms of state recognition, some feature vectors are randomly selected as inputs. The state recognition model based on the nuclear limit learning machine algorithm is trained through training. Use the DDS power transmission fault diagnosis of the vibration data collected by the planetary gears collected in different states of the planetary gears to verify the effectiveness of the above methods.