Raw Anode Volume Density Prediction Algorithm Based on the Genetic Algorithm

• Published in: SN computer science Vol. 3; no. 5; p. 354
• Main Authors: Cao, Danyang, Tian, Xuefa
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 30.06.2022; Springer Nature B.V
• Subjects: Adaptive algorithms; Aluminum industry; Artificial neural networks; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Datasets; Density; Electrolytes; Evolutionary algorithms; Genes; Genetic algorithms; Information Systems and Communication Service; Methods; Neural networks; Optimization; Pattern Recognition and Graphics; Production controls; Raw materials; Software Engineering/Programming and Operating Systems; Survey Article; Vision; Neural network; Competitive evolution; Genetic algorithm; Prediction; Adaptation
• ISSN: 2661-8907; 2662-995X; 2661-8907
• DOI: 10.1007/s42979-022-01248-0

In the process of aluminum electrolytic anode production, raw anode volume density is an important indicator of anode quality, which is of great significance to ensure the quality of raw anode volume density through raw anode production control parameters and the ratio of raw material. Considering the nonlinear characteristics of the raw anode production process and combining the advantages of the neural network in the nonlinear prediction problem, deep neural networks are used to model the raw anode volume density. For the uncertainty of neural network model structure, a competitive evolutionary adaptive genetic algorithm is proposed to determine the network model structure. The algorithm selects well-performing individuals through competitive fitness values to form progeny populations. During the process of genetic variation, the variation probability is adaptive calculated from the fitness values of the parent and the current number of population iterations to converge the optimal results. Experimental results show that in terms of production data, the optimization ability of the neural network model structure is significantly improved compared with other algorithms, with the root mean square error of the prediction value of the raw anode volume density is 0.005, which is smaller error than other methods.