Blasting crater prediction is critical for optimizing charge design and ensuring safety in engineering practice; however, traditional theoretical models suffer from idealized assumptions, numerical simulations are hindered by empirical parameter calibration and high computational cost, and physical experiments are constrained by scalability and repeatability. To address these challenges, a novel Wavelet-Augmented Physics-Informed Neural Networks (WA-PINNs) framework for accurately and efficiently predicting the diameter and volume of blasting craters induced by cylindrical charges is proposed. By leveraging Starfield's superposition method to equivalently model cylindrical charges as a series of spherical sources, a partial differential equation (PDE) system governing the tensile stress field responsible for crater formation was established. Comprehensive validation against high-fidelity numerical simulations and field experiments at Baima Iron Mine demonstrated that the WA-PINNs (Dog) achieves superior accuracy in predicting crater diameter and volume compared to conventional PINNs, with relative errors as low as 0.01 and markedly reduced training time. The results confirmed that the proposed WA-PINNs (Dog) is a robust, efficient, and physics-consistent intelligent solution for complex blasting problems, offering significant potential for real-world blasting design optimization.