A Comprehensive Operating Performance Assessment Framework Based on Distributed Siamese Gated Recurrent Unit for Hot Strip Mill Process

For modern industrial processes, the operating performance assessment has a significant role in ensuring high quality and efficiency in production under normal condition. However, with the switching of working conditions, aging of the equipments, and uncertain disturbance, the inherent nonlinear and dynamic characteristics make it difficult to make an accurate assessment. Meanwhile, industrial processes usually include multiple subsystems, and traditional distributed assessment methods seldom consider practical communications between different subsystems. In addition, normal operating performance assessment and fault level assessment are rarely investigated simultaneously in the existing methods. In this paper, a comprehensive operating performance assessment framework is proposed to deal with the above issues. First, a Siamese gated recurrent unit network is developed by Siamese neural network and gated recurrent unit. Then, based on the process knowledge, a distributed Siamese gated recurrent unit (DSGRU) framework with partial communication is proposed. After that, the assessment strategies are designed for normal and abnormal conditions, respectively. Finally, the effectiveness of the proposed framework is demonstrated on a real hot strip mill process (HSMP).