| Bisphenol-A(BPA)is an important organic chemical raw material and it is also an important derivative of phenol and acetone.Generally,BPA is used to manufacture various polymer materials such as polycarbonate and epoxy resin.In industry,BPA is mainly produced by condensation reaction of acetone and phenol with acid medium as catalyst and followed by dehydration,crystallization,phenol removal and granulation.The production process of BPA is complex,and patent barriers and technical confi-dentiality are high.Although China has introduced advanced production technology from Japan and other countries,through digestion,absorption and technological transforma-tion,it has quite independent intellectual property rights,but the energy consumption and quality of products are not completely satisfactory.Based on mechanism analysis and industrial process characteristics analysis,the soft sensor modelling,optimization of operation conditions and fault detection of BPA synthesis reactor and BPA distillation tower are studied.The main research work of this paper is summarized as follows:1.According to the analysis of synthesis reaction mechanism and the operation characteristics and application conditions of a practical industrial BPA synthesis reactor,a steady-state mechanism model of industrial BPA synthesis reaction process is proposed for the first time based on the material balance and energy balance equations,which can be used to estimate the concentrations of key components at the outlet of reactor and to optimize the operation conditions of the reaction process.2.Aiming at the problem that it is difficult to determine the initial inlet concentra-tions and reaction rate constants in the mechanism model,an improved stochastic gradient Boosting-Gaussian process regression algorithm is proposed to estimate the initial con-centrations and reaction rate constants.The estimated results of initial concentrations and reaction rate constants are substituted into the mechanism model of the process then the mechanism model equation is solved.The simulation results show the effectiveness of the mechanism model.3.Based on the mechanism model of BPA synthesis reactor,the effects of different acetone tank discharge flow rate,first mother liquor tank return flow rate and feed tem-perature on acetone conversion and BPA selectivity are studied by sensitivity analysis.On this basis,a multi-objective optimization framework for BPA synthesis reaction process is established and the objective functions are optimized by constrained multi-objective optimization algorithm,which improved the conversion of acetone and the selectivity of BPA.4.Aiming at the characteristics of BPA dehydration process such as strong nonlin-earity and strong time-varying,an adaptive soft sensor model for BPA distillation process is established by combining local modeling with ensemble learning strategy.The just-in-time learning framework is used to divide the process into several categories and establish corresponding local models.According to the probability that the query samples belong to outlier samples in each category,the local models with excellent performance are select-ed to integrate and estimate the output.In order to make the model adaptive to process state,an adaptive strategy based on updating local model parameters and extracting local model online is proposed.The simulation results of BPA distillation process proved the validity of the model.5.Aiming at the unavoidable failure in BPA distillation process,two fault detection methods are proposed.The former measured the enrichment degree of process information in each principal component direction and reconstructed the principal component space according to the change rate of T~2statistics;the latter divided process variables into blocks by constructing nonlinear correlation coefficient between variables,and constructs KPCA model for each sub block.The simulation of the distillation process of BPA shows that the two methods improve the effect of fault detection significantly. |
PDF Full Text Request