Aplicação de Machine Learning para o estudo de comportamento de uma planta de produção de estireno

Abstract

Artificial Intelligence (AI) and its Machine Learning (ML) algorithms has been calling attention in what it is concerned to several areas lately, and industries are not left out of this. In fact, the sector represented by them tends to gain a lot with those algorithms, since they are based on a massive amount of data, which exists in great amount in the industries, which can help with the training and learning processes. Thus, using those algorithms, companies and industries are able to make production and financial predictions, optimizations, among others. Because of that, there is the possibility of a significant gain by some of these companies, since those predictions directly assist in improvements in what it is concerned with making decisions based on the data generated by the process itself. Taking into consideration this context, the objective of this work is the application of two ML algorithms with supervised learning: Neural Network and Multiple Linear Regression. Such algorithms were used in a digital blueprint of Styrene, in order to understand how some parameters of its process influence the expected final product. From the data training, it is possible to perform the prediction of Styrene by predefined parameters. As a consequence, there is the possibility of optimizing the process, while knowing which parameters most influence the desired final product. Concluding, the evaluation of the best model is carried out using specific statistical measures.