Remaining useful life prediction of industrial robot RV reducer with multiple deep networks and multicore support vector data description

Abstract

The remaining useful life prediction of Industrial robot RV reducer is challenging due to the strong redundancy, unstable degradation initiation point, and environmental interference that may obscure the key state information during long-term operation. To address this problem, this paper proposes a novel remaining useful life prediction method for robot RV reducer with multi-depth network and multi-kernel support vector data description. Firstly, the degradation features are constructed by the hidden layer node of deep belief network to reduce the interference and redundancy. Secondly, the first predicting time node is determined by multi-kernel support vector data description to locate the stable degradation initiation node. Then, the temporal convolutional network is applied to predict the remaining useful life in the degradation stage, which improves the prediction accuracy. Finally, the effectiveness of the proposed method is verified by the accelerated fatigue experiment of a self-made robot.