Hybrid AI Models for Predicting Kerf Width in Abrasive Water Jet Machining (AWJM)

Kerf width is essential in determining the dimensional precision, material loss, and surface finish of Abrasive Water Jet Machining (AWJM). The highly nonlinear interaction among hydraulic, abrasive, and material parameters complicates prediction with classical models. This work proposes a hybrid Artificial Intelligence (AI) framework integrating Finite Element Method (FEM)-generated synthetic data, data-driven learning via Convolutional Neural Networks (CNNs), and Physics-Informed Neural Networks (PINNs) to enforce erosion mechanics constraints. The hybrid model achieves RMSE below 0.02 mm and R2 above 0.98, outperforming ANN- and CNN-based models by 56% and 31%, respectively. The model generalizes across materials and supports real-time deployment with inference latency under 60 ms, enabling intelligent AWJM process control under the Industry 4.0 paradigm.