Simulation And Optimization Of Complex Polymerization Processes With Transport Phenomenon

Polymer products occupy an important position in our daily life.Many polymers are produced through free radical polymerization.While experimenting directly on chemical objects can be costly,establishing a mathematical model of the research object to obtain the relationship between operating conditions,reactor data and free radical quality indicators,and then calculating based on the model to obtain better operating conditions and reactor design can greatly reduce production consumption,improve safety,and increase production efficiency.With the thoroughgoing research on the free radical polymerization process and the improvement of computing technology,the modeling of the free radical polymerization process not only needs to consider the conventional reaction kinetic mechanism,but also the influence of transport phenomenon.As a result,some polymerization problems with high computational complexity have arisen,which pose challenges for the research on simulation and optimization.In some radical polymerization reactions,due to the existence of molecular diffusion,there will be automatic acceleration during the reaction time.Describing this phenomenon via a conditional model results in non-smooth and strong nonlinear features to optimization,which is difficult to solve using existing methods.Besides,considering the transport phenomenon existing in the actual reactors,distributed features have to be computed.It is proposed to combine computational fluid dynamics(CFD)technology with polymerization to discard the ideal reactor hypothesis and reduce the gap with the actual situation.However,this greatly increases the computation complexity,and the derivative information is also difficult to obtain;thus,special optimization algorithms need to be designed.In this paper,an event triggered conditional model of the free radical polymerization considering the diffusion effects caused by molecular diffusion at the micro level is established;then a spatial distributed model of free radical polymerization considering the CFD is established considering transport phenomenon at the macro level.To deal with the optimization of the complex polymerization,model reconstruction techniques are proposed.The specific research work is summarized as follows:1.The radical polymerization reaction process model with diffusion effects is established and the dynamic operation optimization problem is formulated.Most of the models used in the operation optimization of the free radical polymerization reaction only consider the conventional reaction kinetic mechanism and ignore the molecular diffusion-controlled problems brought by the increasing polymer viscosity in the reaction system.According to reference,the reaction kinetic mechanism equations and the condition model of the diffusion-controlled effects are both established in the model of the free radical polymerization process.Taking the reaction process of polymethyl methacrylate(PMMA)as an example,a model simulation is carried out,and the influences of gel effect,glass effect and cage effect during the reaction process are analyzed.Then,based on this model,a dynamic operation optimization proposition is established.The calculation difficulties caused by the derivative discontinuity of the model are analyzed.2.Adaptive smoothing strategy is proposed.Process optimization problems are often solved by gradient-based algorithms for the high efficiency in convergence.However,direct application of gradient-based algorithms to the radical polymerization reaction process with conditional models results in high failure rate and long calculation time due to the existence of the model derivative discontinuity.Thus,the idea using smoothing equations to reconstruct the original conditional model is proposed.Taking into account the nature of the smoothing equations and the characteristics of the optimization proposition,an adaptive smoothing strategy is proposed.This method successfully solves the dynamic operation optimization,and the calculation principle of this strategy is analyzed.3.Modeling method of molecular weight distribution(MWD)in non-ideal reactors based on CFD is proposed.Perfect-mixing condition cannot be achieved in real situation.Thus,the state variables in the reactor are distributed,transport phenomenon of mass,momentum and energy are existed and have important influence on reaction process.In this study,by using computational fluid dynamics(CFD)methods,transport phenomenon and reaction are combined.Free radical polymerization process model in non-ideal reactors is established and the calculation method of molecular weight distribution in non-ideal reactor is further proposed.Then,two applications of the CFD models are conducted on the low-density polyethylene(LDPE)in the tubular reactor and the tank reactor,respectively.The calculation method of the MWD in non-ideal reactors is verified.4.Optimization on CFD model using a global surrogate model is proposed.Simulation of a CFD model takes a long time.Furthermore,existing CFD software cannot provide derivative information.It is difficult to directly use the gradient-based algorithm for the optimization with the CFD model.Therefore,Kriging method is used to reconstruct the CFD model.To obtain a reliable surrogate model,accuracy analyses of the surrogate model under different sampling points and different basis functions are carried out.With the obtained global surrogate model,the process optimization for non-ideal reactors can be implemented.5.Optimization algorithm based on iterative update of surrogate models is presented.Establishing high-fidelity global surrogate model to solve the optimization propositions with the CFD model requires a lot of CFD simulation calculations,which brings high computation burden.Thus,an optimization method based on iterative updated surrogate model is further proposed,which starts the calculation with an initial rough surrogate model and constantly updates the surrogate model during the calculation process,till it approaches the optimal result.This method can ensure that the obtained solution is the local optimal solution of the original optimization proposition,with high computational efficiency at the same time.Two optimization problems of the LDPE process in the tubular reactor and the tank reactor are presented by using this new method.The number of CFD simulations are greatly reduced,which proves the reliability and efficiency of this method.