Modeling And Optimization Of Bed Fluid Flow Characteristics In Trickle Bed Reactor

The trickle bed reactor is a three-phase reactor in which gas and liquid flow down through the catalyst bed.It is widely used in petroleum refining and petrochemical fields.The performance of a trickle bed reactor largely depends on the fluid flow characteristics in the reactor bed.It is essential for fully exploiting the production potential of the reactor and optimizing the design of the reactor to know and understand fluid flow characteristics in the reactor.For the study of fluid flow characteristics in trickle beds,experimental analysis,empirical or semi-empirical theoretical models,and computational fluid dynamics simulation(CFD)are mainly used.Among them,experimental analysis is limited by operation,geometry and physical conditions,and is usually time-consuming and expensive;empirical or semi-empirical theoretical models are often established under certain assumptions or simplified conditions,and model prediction accuracy needs to be improved;although CFD can reduce the equipment cost of the experiment,it brings huge computational cost,and the simulation result is affected by the quality and quantity of the mesh.Therefore,it is still a huge challenge to develop a simple,economical and efficient method to study the fluid flow characteristics of the trickle bed.Due to the complexity of the multiphase flow inside the bed,the modeling optimization problem is usually unable to be described by the mechanism model.Artificial intelligence technology is an effective method to solve this problem.The bed flow regime transition and bed liquid distribution of trickle beds are important research directions for the fluid flow characteristics of trickle beds.In order to clearly describe the complex non-linear relationship between flow regime conversion,liquid distribution and the physical parameters of the trickle bed reactor,bed characteristic parameters and operating parameters,the random forest regression algorithm(Random Forest Regression,RFR)is used to model the flow transition boundary and liquid distribution of the trickle bed.First,use random sampling technology to preprocess the collected experimental data,and then use the gray wolf optimization(GWO)algorithm to optimize the number of decision trees ntree of RFR and the number of randomly selected features mtry to improve the reliability of prediction.Finally,80%of the original data set is used as the training set to train the RFR model with optimized hyperparameters,and the remaining 20%data is used to test the generalization ability of the model.GWO-RFR,a hybrid algorithm that combines GWO and RFR,predicts the transition boundary and liquid distribution of the trickle bed.The average absolute percentage error(MAPE)of the test set was 8.23%and7.10%,the mean square error(MSE)was 1.16e-06 and 4.8e-04,and the correlation coefficient(R~2)was 0.975 and 0.969,respectively.The prediction results of the established GWO-RFR transition boundary model and GWO-RFR trickle bed liquid distribution model are compared with empirical models,other machine learning models,etc.The results show that the GWO-RFR model has higher accuracy,stronger generalization ability and robustness compared with other models.In the process of modeling the transition boundary of the trickle bed,the random forest out-of-bag data(OOB)was used to sort the factors that affect the transition boundary of the trickle bed,and three characteristics were selected to analyze their influence on the transition boundary.In addition,in the process of modeling the liquid distribution of the trickle bed,the random forest out-of-bag data(OOB)was also used to order the factors that affect the liquid distribution of the trickle bed.Based on the established GWO-RFR model,the particle swarm optimization algorithm is used to optimize its operating parameters,which provides a new method for optimizing the distribution of liquid in the trickle bed.In order to solve the problem that the standard gray wolf optimization algorithm is easy to fall into local optimality,and the convergence speed and optimization accuracy need to be improved,an improved gray wolf algorithm-Chaos Gray Wolf Optimization Algorithm(CGWO)is proposed.Five improved gray wolf algorithms and 11 standard test functions are selected to compare and analyze the chaotic gray wolf algorithm.The research results show that the accuracy and convergence speed of CGWO are better than other algorithms.Finally,the algorithm is coupled with the random forest regression algorithm to establish CGWO-RFR and apply it to the prediction of the transition boundary of the trickle bed.The comparison shows that the model has higher prediction accuracy,generalization ability and faster calculation speed than the decision tree model,RFR model,and GWO-RFR model.In order to realize the visualization of the trickle bed flow pattern modeling process,the trickle bed flow pattern transition boundary prediction system was developed through the graphical user interface GUI.This system not only simplifies the operation steps,but also provides a large amount of data for the operation of the trickle bed as a reference,which has certain engineering practical value.