SURFACE ROUGHNESS ESTIMATION BASED ON HYBRID LEADER TASMANIAN DEVIL OPTIMIZATION-ENABLED DEEP LEARNING

Surface roughness is a key indicator of surface characteristics and plays a crucial role in determining the texture of a surface, classifying it as either smooth or rough. It is essential for ensuring the proper functioning of the machine components. The traditional method for measuring surface roughness involves the use of a stylus device that relies on contact measurements. However, due to its limitations in real-time applications, researchers have developed image-based techniques for more efficient surface roughness assessment. This study presents an effective model for estimating surface roughness using ZF-Net, which is trained with the Hybrid Leader Tasmanian Optimization (HLTO) algorithm. The HLTO algorithm is the integration of the Hybrid Leader-Based Optimization (HLBO) algorithm and Tasmanian Devil Optimization (TDO) algorithm. The stages involved in this paper are pre-processing, data augmentation, feature extraction, feature fusion, and roughness estimation. Here, the pre-processing is done by utilizing a bilateral filter, and also, the feature fusion is performed by using the Deep Belief Network (DBN) model with an overlap coefficient. Finally, roughness is estimated by the ZF-Net, which is trained by HLTO. Moreover, the proposed model achieved the lowest Normalized Mean Absolute Error (MAE) of 0.109, Normalized Mean Squared Error (MSE) of 0.010, and Normalized Root Mean Square Error (RMSE) of 0.012, accordingly.