Taylor-Gorilla troops optimized deep learning network for surface roughness estimation.

Abstract

In order to guarantee the desired quality of machined products, a reliable surface roughness assessment is essential. Using a surface profile metre with a contact stylus, which can produce accurate measurements of surface profiles, is the most popular technique for determining the surface roughness of machined items. One of the limitations of this technique is the work piece surface degradation brought on by mechanical contact between the stylus and the surface. Hence, in this paper, a roughness assessment technique based on the suggested Taylor-Gorilla troops optimizer-based Deep Neuro-Fuzzy Network (Taylor-GTO based DNFN) is proposed for estimating the surface roughness. Pre-processing, data augmentation, feature extraction, feature fusion, and roughness estimation are the procedures that the suggested technique uses to complete the roughness estimate procedure. Roughness estimation is performed using DNFN that has been trained using Taylor-GTO, which was created by combining the Taylor series with the Gorilla troop's optimizer. The created Taylor-GTO based DNFN model has minimum Mean Absolute Error, Mean Square Error, and RMSE of 0.403, 0.416, and 1.149, respectively.