In the diploma work, we analysed the data of the steel process of titanium steel production with the program Orange. We performed batch analyses with individual errors according to the assigned data on the manufacturing process. The first database was attributed production data in the EAF, with secondary metallurgy processes and continuous casting, and the second database was attributed production data with secondary metallurgy processes and continuous casting. The batches were compared by groups of errors and by individual errors, and the influential parameters for the occurrence of the error were also analysed. Data models were used in the analysis, namely neural networks, decision tree and AdaBoost model. With the obtained results, we produced graphs of violin arrangement, bar charts, tree representations and confusion matrices. We compared the accuracy of the forecast between individual data models.
|
