Particle Size Distribution Modeling And Control For The Batch Crystallization Process

In recent years, the crystallization process is facing enormous challenges and opportunities with the rapid development of biological chemical and pharmaceutical industry. The requirements for the size and distribution of crystal are getting more and more demanding in order to meet the needs of industrial development. The particle size distribution (PSD) is a key factor for high quality production. And it determines the efficiency of downstream operations. The main difficulty in the control of PSD is the absence of effective measuring methods. Therefore, seeking an effective and high-performance control method for crystallization process becomes an urgent problem to be solved. This paper is devoted to study the on-line control of particle size distribution in the crystallization process and improved the product quality.At first, this paper introduced the reaction mechanism for crystallization process, which includes the nucleation rate model, the growth rate model, the solubility model, the relationship and model of the material balance model and particle balance model. On the basis of the model, this paper studied the method for the model numerical solution, established the mechanism model of the crystallization process, and analyzed the effect of different operating variables on the particle size distribution.Particle size distribution control is very complicated to realize due to the absence of effective measuring methods for PSD. Therefore, the tracking control of PSD which cannot be measured online is converted into the measurable moment control. However, another problem is that the crystallization distribution cannot be observed intuitively through the moment model. Aiming at this, this paper applied state estimation in the moment model to obtain the particle size distribution. The concentration change is estimated through detecting the changes of torque value, thus realizing the particle size distribution measurement, which is very important in the control of dynamic process.The model for crystallization process is usually established in the ideal case, which is difficult to meet the precision requirements. This paper applied the data modeling approach, acquired the relationship of distribution and moment for the crystallization process and established the neural network model between operating variables and low-order moment to reduce the difficulty for modeling. Then according to the given optimization index, we derived the adaptive iterative learning rate to control the low-order moment of particle size distribution. In this way, this paper realized the indirect control of the particle size distribution. Finally, the method is applied to the cobalt oxalate crystallization process and the simulation results proved the feasibility of the method.