Control Of Particle Size Distribution In Crystallization Process Based On Real-time Image Analysis

The crystallization process is a key link in the separation and purification of crystal products.It is widely used in the pharmaceutical,food,chemical and other industries.Among them,the crystal size distribution is the main focus of the crystallization process control.The crystal size distribution determines the characteristics of the crystal product and also affects some of the characteristics of the crystal product.The downstream operating unit has an important influence.However,due to the lack of online measurement of crystal size distribution,most of the current researches use offline laboratory analysis to adjust the operating conditions of the crystallization process to control the crystal size distribution,which often results in large fluctuations in the crystal size distribution and low qualification rate of crystal products.problem.Although scholars have used real-time image analysis technology to complete the online measurement of the particle size distribution in the crystallization process,it has not been used in the closed-loop control of the particle size distribution.The online control of the crystal particle size distribution has become a problem to be solved in the study of the crystallization process.In response to the above problems,this thesis has carried out the research on the online measurement and online control of the particle size distribution during the crystallization process.The main work of this thesis is as follows:(1)Aiming at the problem that it is difficult to measure the particle size distribution online during the crystallization process.First of all,this thesis uses the image processing method based on deep neural network to accurately segment the crystal.Secondly,the actual one-dimensional size of the crystal is measured by using the average value of the length and width of the best circumscribed rectangular frame of the crystal target area and the pixel equivalent.After that,through the introduction of two evaluation indicators of similarity deviation and variation coefficient,the minimum number of crystal particles required to characterize the crystal size distribution in the reactor was analyzed.Finally,the density estimation of the Gaussian kernel function is used to smooth the crystal size distribution(Crystal Size Distribution,CSD)calculated by the numerical value to obtain a real-time crystal size distribution curve.(2)Aiming at the problem of on-line control of the particle size distribution of the crystallization process,this thesis selects the Gaussian function to approximate the particle size distribution curve of the crystallization process according to the typical non-Gaussian nature of the crystal particle size distribution,and uses the iterative learning algorithm to complete the parameter tuning of the Gaussian function.On the basis of the real-time particle size distribution curve measured in(1)is decoupled from the weight,and the tracking control of the particle size distribution is transformed into the tracking control of the weight vector characterizing the particle size distribution,and then based on the input variables in the crystallization experiment(Temperature)and the change rule between output variables(CSD),and on this basis,the particle size distribution controller is designed to control the output weight,and then realize the tracking control of the target particle size distribution curve.(3)Aiming at the verification and analysis of the control method of particle size distribution in the crystallization process,this thesis integrates a variety of process analysis techniques,uses software development tools,and based on the description of the particle size distribution control problem,starting from the analysis of system requirements,the crystallization process particle size distribution control is designed Based on this system,the particle size distribution control of the constant cooling rate alum crystallization process and the feedback-based particle size distribution control of the alum crystallization process are experimentally studied.The research results show that compared to the constant cooling rate crystallization process,the particle size distribution control method of the crystallization process based on real-time image analysis proposed in this thesis can well realize the tracking control of the crystal size distribution.