The paper introduces a model for the identification of milling parameters, leveraging known surface roughness parameters. The accurate surface geometry was built by using the Z-buffer algorithm, which enables modeling according to specified cutting modes and tool geometries while also allows accounting the additional tool deflections caused by vibrations. For quantitative characterization of the obtained surfaces, a computational module that calculates roughness parameters in accordance with GOST ISO 4287-2014. GOST ISO 4287-2014 was developed. To address the inverse problem of milling parameter identification from known roughness, a neural network model based on a multilayer perceptron was trained using generated surfaces that incorporate vibration data. The development of the algorithms and neural networks was made using the Julia programming language and the Flux library. The results demonstrate an error margin of no more than 12% in mill diameter identification and no more than 5% in the identification of cutting modes. The developed neural network model, in conjunction with the surface generator based on the Z-buffer algorithm, offers a powerful tool for generating representative surfaces with specified roughness, that could be used in numerical simulations of contact interactions. Moreover, this model holds significant potential as a supportive tool in manufacturing engineering, particularly in the preproduction phase.