Fault diagnosis of pumping unit based on lightweight convolutional neural network

In order to ensure the efficient operation of the pumping unit system and monitor its status in real time, rapid and accurate fault diagnosis and identification of the oil well operating curve (work diagram) is a key step to evaluate the working efficiency of the pumping unit. In response to this demand, a lightweight model based on the MobileNetV3 model is proposed, which integrates the lightweight convolutional block attention module (MLCA) and improves the accuracy of dynamometer diagnosis without increasing the computational load of the model. Experiments Expanded on a data set containing 7 types of different working conditions and a total of 9464 work map images, the results show that the model still achieves an average recognition accuracy of up to $\mathbf{98.44\%}$, surpassing the performance of baseline models such as ShuffleNetV2, EfficientNetB0, and MobileNetV3. These data strongly prove that the improved MLCA-MobileNetV3 model performs well on the dynamometer recognition task and has significant practical value and efficiency advantages.