Etching profile anomalies occur around large-scale 3-dimensional (3D) structures due to distortion in the ion sheath and ion trajectories. To solve this problem, a system to predict such etching anomalies was developed by combining on-wafer sheath shaped sensor and simulations based on a neural network and a database. The sensor could measure the sheath voltage and saturation ion current density and sheath thickness can be calculated from them. A database was built by using the results from sensor measurements and etching experiments with samples with large vertical steps, which enables prediction of etching shape anomalies from measured parameters. Finally, the system could predict etching shape anomalies around large vertical steps.