Intelligent Detection of Robotic Arm Roller Wear Based on Combined DWAE and GRUNN Models

Abstract

Abstract Long time and high intensity work makes the wear of the roller of the robotic arm increase gradually, which will affect the performance and service life if not regularly overhauled. In this paper, the sensor is used to obtain roller data from the mechanical arm, and the roller wear characteristics are extracted by combining them with the LSTM model. Based on the autoencoder, the wavelet function is introduced to establish a deep wavelet autoencoder, and the DWAE-GRUNN model is established by combining the gated recurrent unit neural network, which is used for the intelligent detection of the wear of the roller of the robotic arm. The model’s classification performance can be improved by using the CTC loss function, and the parameters can be optimized using the sparrow search algorithm. For the application of the DWAE-GRUNN model in the intelligent detection of roller wear of mechanical arms, this paper validates the model based on the aspects of performance, coefficient selection, and detection results. It is found that the average correct detection rate of the model in this paper is 95.97%, and the optimal accuracy is obtained when the weight attenuation coefficient of the DWAE structure is 0.005. The friction force changes inversely with the degree of smoothness after 3.1h of the detection time of the robotic arm roller, and the stiffness, cumulative large abrasive grains, and cumulative small abrasive grains also show significant changes. Using the DWAE-GRUNN model, changes in the wear of the robotic arm roller can be obtained, providing data support for timely maintenance.