Soft Sensor Modeling And Implementation Based On Dynamic Relevance Vector Machine For Animal Cells Suspension Culture

With the rapid development of animal husbandry,animal infectious diseases(such as avian influenza,foot and mouth disease,etc.)are also catching up.The market demand of veterinary vaccines has been greatly increased,and the production of vaccines by the traditional bottle culture method has been far from being able to meet the development of modern animal husbandry industry.The cell suspension culture technology gradually replaced the traditional bottle culture.It has gradually become the mainstream technology of industrial production of biological vaccines.Therefore,how to quickly and efficiently achieve vaccine production have become the focus of the current and future bio-pharmaceutical field and the core of competition.The suspension culture of animal cells is a highly nonlinear and time variant reaction process.In the process,due to the shortcomings(susceptible to bacterial infection,high cost,etc.)of the traditional physical instrument measurement methods,the real-time monitoring of some key state variables,such as glucose concentration,lactate concentration,cell density,and so on,has seriously restricted the synthesis of the products,thus affecting the yield and quality of biological products.The off-line measurement has a certain lag,which seriously affects the optimization control of the whole process.The emergence of soft sensor technology provides an effective way to solve the above problems.In order to solve the problem that the key state variables in BHK-21(Hamster kidney cells)suspension culture is difficult to be measured in real time,a soft sensor modeling method of dynamic relevance vector machine based on the angle of right triangle is proposed.Firstly,the main variables of the model are determined according to the process mechanism of the culture process.Then the auxiliary variables are determined by the correlation coefficient method based on the data collected in the field.Finally,the soft sensor model is established.On the basis of a thorough understanding of support vector machine,least squares support vector machine and relevance vector machine,which are compared with the soft sensor modeling method of dynamic relevance vector machine based on the angle of right triangle.The results show that,compared with support vector machines and least squares support vector machine,the solution of relevance vector machine is sparse better and more robust,and the model has high prediction accuracy.The dynamic relevance vector machinenot only has the advantages of the relevance vector machine,but also can introduce the dynamic characteristics of the industrial process into the model,which makes the forecasting effect more accurate and more able to reflect the essence of the industrial process.In order to realize real-time monitoring and optimal control of BHK-21 cell suspension culture,this paper makes full use of the powerful computing ability of MATLAB and the powerful script programming ability of WinCC,an intelligent monitoring system based on WinCC for BHK-21 cell suspension culture was established.The experimental results show that the monitoring system can display the key state variables in real time and accurately,which provides sufficient operating conditions for the further optimization control,reducing the occurrence of the fault,so as to achieve the purpose of economic and efficient production.