The invention discloses a
gear pump cavitation state recognition method based on a
physical information neural network, and belongs to the technical field of fluid mechanical state monitoring. Comprising
test data acquisition, feature analysis, theoretical flow calculation and
cavitation identification. The method comprises the following steps: acquiring
gear pump driving wheel rotation angle and flow data under multiple working conditions, and analyzing flow pulsation, time-
frequency spectrum and flow pulsation and rotation angle information
coupling characteristics under different
cavitation states; theoretical flow is calculated, and the cavitation state is divided according to the actual-theoretical flow pulsation
rate ratio; constructing a
physical information neural network through a cavitation-
free state, and embedding theoretical flow into a
loss function to enhance the physical
interpretability of the model; and after training is completed, the features of different working conditions are input into a neural network to calculate the predicted flow, residual curve features between the predicted flow and the actual flow are obtained, and the residual curve features are input into a
support vector machine to achieve efficient classification of non-cavitation, primary cavitation, critical cavitation and serious cavitation. The method is simple in step, convenient to use and high in recognition precision.