Product Prediction And Optimization Of Shift Device Based On Digital Modeling

In the context of industrial intelligence construction,strengthening petrochemical enterprise production management,operation optimization,and upgrading equipment intelligence and digitization are the focus of current construction.The article takes the shift device of a factory as the background,and aims to build the production optimization module in the intelligent management platform.According to the current status of data,use hybrid modeling and data-driven modeling techniques to conduct simulation optimization research on key reactors and the device,in order to achieve the goal of product quality control in the construction of enterprise intelligent platform and improve the intelligent level of the device.The specific research content and conclusions are as follows:(1)The Core reactor R2202?R2204 hybrid modeling and product prediction.Combined the transfer mechanism of the reactor,estimated the reaction kinetic parameters based on the operating data,and establishd a hybrid model through optimization algorithms and differential equation solving In order to make up for the lack of sample data caused by the incomplete collection of reactor operating data and the shortage of measuring instruments,based on the design data,Aspen Plus was used to simulate actual production fluctuations to generate sample data,which was used as the data source for reactor hybrid modeling.The results show that the kinetic equation calculated by the GA is more reasonable than PSO and SA.The average relative errors of the training samples of the reactors R2202?R2204 are 3.633%,3.237%,and2.048%,respectively,and the verification sample is 5.106%,3.817%,2.504%.The product prediction error is small,and the model can be used for subsequent real-time optimization and model integration.(2)BP neural network modeling and product prediction of shift device.Collected 1041 sets of data in the actual operation of the whole device,and screened the data variables through the screening rules established based on the combination of empirical knowledge and MIC.Finally,161 variables were reduced to 23 variables(85.71% reduction),which improved the calculation efficiency.Then the BP neural network is used to model,and the network structure parameter analysis shows that the structural model with the L-M algorithm,3 hidden layers,and the number of neurons of 10,9,and 9,respectively,has higher accuracy.The relative average deviation between the simulated value and the real value is 1.193%,which proves that the model is reliable and can be used for product quality prediction and real-time tuning(3)Reactor operation variables analysis,shift device operation optimization,and production optimization module architecture design in the intelligent management platform.Analyzed the influence of the flow rate,temperature and water-to-carbon ratio of the reactor inlet mixed gas on the reaction process,based on the above-built model.The results show that in the operating range,reducing the flow rate and lowering the inlet temperature will increase the reaction conversion rate;for reactor R2202,the larger the water-to-carbon ratio,the higher the conversion rate,while the water-to-carbon ratio of reactors R2203?R2204 is 11 and 54,the minimum CO concentration is 0.706% and 0.279%.The GA-BP method is used to optimize the8 operating variables that have a greater impact on the device from the perspective of global optimization.After optimization,the CO concentration is 0.342%,which is about 42.424%lower than the original,which is less than the technical index requirement(0.5%),strengthening the quality control of the product.Finally,the article preliminarily designs the production optimization module architecture in the intelligent management platform of the shift device,with a view to providing guidance for the subsequent platform construction and accelerating the process of the intelligent construction of the device.