Improvement of osprey optimization algorithm and weighted RBM features for predicting the surface roughness and flank wear measurement using 1DCNN with ridge regression

• Published in: Machining science and technology Vol. 29; no. 1; pp. 43 - 92
• Main Authors: V, Dilli Ganesh, S, John Justin Thangaraj
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 02.01.2025
• Subjects: Adaptive osprey optimization algorithm; deep belief network; flank wear measurement; multiscale atrous spatial pyramid pooling based one dimensional convolution neural network with ridge regression; surface roughness and cutting force prediction
• ISSN: 1091-0344; 1532-2483
• DOI: 10.1080/10910344.2025.2458050

Surface finish is considered the significant factor in the evaluation of product quality. Surface roughness (SR) is highly utilized as an index for determining the surface finish at the time of the machining process. The flank wear (FW) is computed with the varying geometrical connections between a worn and a new cutter, which assures a full FW mapping. In addition, it is essential to forecast the cutting forces for attaining the stability of the machining system and its dimensional accuracy. To meet this requirement, an adaptive technique is proposed for SR, cutting force and FW quantity. The deep features are extracted by using the deep belief network (DBN) from the data values; here attributes are optimized using the adaptive osprey optimization algorithm (AOOA). Finally, the resultant features are given as input to the multiscale atrous spatial pyramid pooling-based one-dimensional convolution neural network with ridge regression (MASPP-1DCNN-RR) for the prediction task. Diverse performance metrics are used to show the efficiency of the recommended scheme. From those experiments, the developed AOOA-MASPP-1DCNN-RR obtains 36.9%, 30.97%, 27.07%, 16.39% and 3.57% improved performance on FW prediction than the Adaboost, support vector machine (SVM), ASPP-1DCNN, ridge regression (RR) and ASPP-1DCNN-RR, respectively, based on analysis of RMSE.