Abstract: With the advancement of machine learning, data-driven flow field prediction has become a research hotspot. Compared with conventional neural networks, which rely on large amounts of training data and have poor interpretability, physics-informed neural network (PINN) embed physical constraints directly into the loss function, enabling flow field prediction with only a small amount of training data. This study explores the reconstruction of shear flow around a circular cylinder under low Reynolds number conditions using a PINN-based approach. Numerical simulation data generated by the open-source CFD software OpenFOAM were used to construct the training dataset. The PINN model for circular cylinder flow with shear flow boundary conditions was developed, and velocity and pressure information from sampling points in the flow field were utilized to train the model to predict velocity and pressure distributions in different regions. By comparing the predictions of the PINN model with the numerical results, the performance of PINN in reconstructing complex flow fields was evaluated. Additionally, the effects of neural network parameters, such as the number of layers, nodes per layer, and monitoring point locations, on prediction accuracy were investigated, and the optimization mechanisms of these parameters were analyzed. The results indicate that PINN can effectively reconstruct the velocity and pressure distributions in shear flow fields around a cylinder, with prediction accuracy closely matching that of numerical simulations. This validates the accuracy of PINN in reconstructing shear flow around a circular cylinder. Comparative analysis further reveals that the location of monitoring points has a significant impact on prediction results, and selecting monitoring points appropriately can improve PINN prediction accuracy by 1 to 2 orders of magnitude.