High-efficiency abrasive water jet milling of aspheric RB-SiC surface based on BP neural network depth control models

For large processing allowance of large diameter RB-SiC mirror blanks and low efficiency of grinding, an abrasive water jet milling is used to quickly remove the processing allowance. In this article, a single kerf profile processed by abrasive water jet milling was effectively fitted to the Gaussian curve. By superimposing Gaussian curves linearly, the surface waviness of superimposed curve was gradually reduced as step-over distance decreased. The surface waviness induced by abrasive water jet milling can be effectively reduced when step-over distance is controlled to less than 1.8 sigma. BP neural network models between step-over distance, traverse speed, and milling depth were established. The prediction error of milling depth can be controlled at about 5% of the total depth, with a maximum error less than 7%. The aspherical RB-SiC surface was generated by abrasive water jet milling with a processing path composed of 20 spiral segments. Different milling depths were obtained by setting different levels of traverse speed and step-over distance for each spiral segment. The processed aspherical surface was highly fitted to the design aspherical surface with a maximum error about 10% of the total depth. The error curves float at the zero line, and the error curves were controlled at 20% of the total depth. By this method, the processing allowance of large diameter RB-SiC mirror blanks can be effectively reduced.