Hybrid Optimization Assisted Deep Convolutional Neural Network for Hardening Prediction in Steel

Abstract Hardness is a property that prevents forced scraping or surface penetration of material surfaces against deformation. Indeed, some methods in the tradition of forecasting the mechanical properties of the steel used to recommend a new hardening forecast using a profound learning model. More particularly, an Optimized Deep Convolutional Neural Network (DCNN) framework is used that makes the prediction more accurate and precise. The input given to the model is the chemical composition of steel along with the distance from the quenched end, which directly predicts the hardening of steel as the model already knows of it. Moreover, to make the prediction more accurate, this paper aims to make a fine-tuning of Convolutional layers in DCNN. This paper suggests a new hybrid algorithm for optimal tuned, which is then hybridized Sea Lion Optimization (SLNO), Dragonfly Algorithm (DA), and Sea Lion insisted on Dragon Fly Modification (SL-DU). This is an optimal tuning. Finally, the performance of the proposed work is compared and validated over other state-of-the-art models for error measures. Finally, the performance of the adopted system was evaluated compared with other traditional systems and the results were achieved. According to the analysis, the MAE of the pattern used for distance 1.5 was 77.16%, 9.84%, 12.71%, and 23.36% better than regression, MVR, ANN, and CNN.