Development of Weighted Ensemble Deep Learning Network for Surface Roughness Prediction and Flank Wear Measurement

Recently in the modern manufacturing industries, surface roughness (SR) has become a difficult factor for characterizing the surface quality of the materials. Roughness is generally specified in any kind of operation or process that has been aided for providing quality information about products. During the workpiece of such materials, it creates heat at the cutting zone that might affect the surface of the machinery parts. Henceforth, predicting the SR is requisite to evade the making of ineffective products. Additionally, this process includes enormous tools that need measurement with certain values. The earlier models are found to be intricate while handling the prediction and measurement for the quality. Therefore, a novel model is implemented for cutting force, SR, and flank wear measurement. The initial stage is performed to collect the input data from standard data sources. Then, the ensemble deep learning technique is developed by integrating the one-dimensional convolution neural network, deep temporal convolution network, deep belief network, and long short-term memory. This ensemble technique is used for predicting the SR. Further, the weight-based fusion takes place between the predicted outputs of the developed ensemble techniques; thus, it is named as weighted average fusion-based ensemble deep learning. This network also employs the newly developed algorithm as opposition squid game optimizer to get the weighted fused predicted outcome. Finally, the performance of the model is examined and measured across divergent metrics, which are then compared with existing approaches. The findings of the developed model show 1.8% of MAE and 10.36% of RMSE, respectively. Thus, this resultant analysis of the developed model secures better performance than the existing approaches.