Use este identificador para citar ou linkar para este item:
http://repositorio.ufla.br/jspui/handle/1/59816
Título: | Predição da análise de cal livre na produção de cimento por meio de aprendizado de máquina e uso de dados sintéticos |
Título(s) alternativo(s): | Prediction of free lime analysis in cement production through machine learning and use of synthetic data |
Autores: | Rodríguez, Demóstenes Zegarra Melgarejo, Dick Carrillo Dias, Vinicius Vitor dos Santos |
Palavras-chave: | Aprendizado de máquina Cal livre Cimento Clinquer Machine learning Free lime Cement |
Data do documento: | 10-Fev-2025 |
Editor: | Universidade Federal de Lavras |
Citação: | SOUSA, Alexandre Ladeira de. Predição da análise de cal livre na produção de cimento por meio de aprendizado de máquina e uso de dados sintéticos. 2025. 81 p. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Lavras, Lavras, 2024. |
Resumo: | Cement production is a complex process that involves mining, grinding of raw materials and heating of materials in clinker kilns. Throughout this production chain, homogenization and chemical reactions occur that alter the structure of the compounds to obtain a final product that meets the quality standards required by regulations. In the clinkering stage, one of the cru- cial parameters to be monitored is the free lime content, which directly impacts the quality of the cement and the efficiency of the process. However, this analysis is punctual, usually per- formed every two hours, and includes collection, sample preparation and calibration steps in X-ray equipment. Therefore, this work aims to develop a predictive model for the analysis of free lime in clinker kilns, using machine learning techniques and synthetic data generation. A consultation with process experts and related research was carried out to define the variables with the greatest impact on the free lime value, their respective limits that characterize stability in the operation, as well as cases in which the sampling of the variables could be impaired. Based on this, a historical database of the selected quantities was created, followed by data processing and increase of the database with the generation of synthetic data through techni- ques based on real data, being interpolation between samples and disturbance through Gaussian noise. Then, machine learning algorithms were applied to predict the free lime content, eva- luating the performance of each one and aiming to guide proactive adjustments in the process. In the performance analysis, the proposed predictive model obtained the indices of R2 = 0.966, MSE = 0.02 and RMSE = 0.141, compared to R2 = 0.73 and MSE = 0.1162 of the bibliographic reference most similar to the present application. The experimental results obtained indicate that it is possible to predict free lime, highlighting the relevance of the work for improving the energy efficiency of the plant, reducing waste, and providing greater stability in the quality of the clinker and, consequently, of the cement produced. |
URI: | http://repositorio.ufla.br/jspui/handle/1/59816 |
Aparece nas coleções: | Ciência da Computação - Mestrado (Dissertações) |
Arquivos associados a este item:
Arquivo | Descrição | Tamanho | Formato | |
---|---|---|---|---|
DISSERTAÇÃO_Predição da análise de cal livre na produção de cimento por meio de aprendizado de máquina e uso de dados sintéticos.pdf | 1,95 MB | Adobe PDF | Visualizar/Abrir | |
IMPACTOS da pesquisa.pdf | 186,26 kB | Adobe PDF | Visualizar/Abrir |
Este item está licenciada sob uma Licença Creative Commons