Input-output networks considering graphlet-based analysis for production optimization: Application in ethylene plants

Complex systems and massive amounts of data present a huge challenge for modern chemical enterprises, but traditional data statistical analysis and data mining methods have certain limitations. Therefore, this study proposes a production optimization methodology about input-output (I–O) network construction considering the graphlet representation. Maximal structure network is built by mechanism modeling considering I–O relationships and constraints. Then solution structure network is obtained by the direct consumption coefficient matrix of specific production data. The proposed I–O graphlets introduce process information, and different nodes represent different I–O factors in an I–O graphlet. Graphlet-based characterization and analysis can be achieved by networks and graphlets characteristics statistics. The ethylene plant I–O network is provided based on the monthly production data of 26 ethylene plants in China to optimize the ethylene production. Compared with the correlation network, the I–O network has a clear structure and physical meaning of links. Meanwhile, the hierarchical clustering based on I–O graphlets has a higher precision clustering effect than the generic graphlets. Furthermore, the ineffective ethylene plants can improve the energy efficiency and economic benefits according to the optimal production benchmark of the effective ethylene plant. [Display omitted] •Novel input-output networks considering graphlet representation is proposed.•I–O network of ethylene plants is provided based on 26 ethylene plants.•Hierarchical clustering based on I–O graphlets has a higher precision clustering.•Production optimization of ineffective ethylene plants can be obtained.