Analyze and application machine learning technique for diagnosis industrial production process

Abstract

This paper proposes an application protocol of diagnosis process in a reactor machine of a complex process. By using Bayes Learning Technique, this protocol is trained by learning historical production database in order to diagnosis the failure cause of this reactor in production process. Application in automation field, the model propose is structured automatically from collected data extract directly from sensors. Based on Expectation Maximization algorithm in machine learning, we show that the result of this model is to classify and also to identify the root causes of drift problem in a specific scenario simulation. Therefore, our key contribution to support maintenance tool machine for increasing life-cycle engineering.