Zhi Qiu, Zeyu Xu, Min Luo, Simon X. Yang, and Tao Chen
AGV fault diagnosis; convolution neural network; cloud platform
Automated-guided vehicles (AGVs) are widely used in the manufacturing and logistics industry, so it is crucial to meet the demand for intelligent, efficient, and accurate fault diagnosis of AGVs. This study proposes a fault diagnosis method based on deep convolutional neural networks with a wide first-layer kernel (WDCNN), which enables a more complete extraction of fault features. A cloud platform trolley data acquisition system and an experimental AGV platform are established. Compared to traditional machine learning methods, WDCNN achieves more accurate differentiation and localisation of faults in the same part and obtains a higher diagnostic accuracy of over 99%. With the proposed method, the results of cloud-based diagnosis exceed 98% accuracy, micro-F1-scores surpass 0.98, and the prediction time is within 0.009 s. The proposed method is suitable for fault diagnosis of AGV trolleys under remote and unmanned monitoring.
Important Links:
Go Back
