Abrasive water jet cutting is widely accepted by the machine shop over the world and its advantages are well known.In order to perform precise machining,such as precise cutting,milling,drilling and grinding etc.,it is necessary to precisely predict the cutting depth of abrasive water jet.Based on the artificial neural network(ANN),a model for the abrasive water jet cutting process is created.Three layers network topology is adopted.The input variables are water pressure,diameter of orifice,diameter of abrasive particles,flow rate of abrasive and feed rate of cutting head respectively.The output variables are cutting depth of abrasive water jet.The sample data set is collected on JJ-I water jet machining and mild steel A3 is used as the cutting sample material.The network is then trained based on sample data set using improved BP algorithm.The trained network establishes nonlinear relationships among the parameters of abrasive water jet cutting process at given accuracy.Thereafter the cutting depth can be precisely predicted using this ANN model.The precise machining process can be achieved by integrated this model into computer numerical controller.