Research On Control And Performance Assessment For Non-gaussian Batch Processes

With the development of economy,the increase of product types and the acceleration of market demand,more and more industrial process need to turn raw materials in processing order for the product and to operate repeatedly for same products.Such industrial processes require both the ability to obtain more products by repeated operation and the ability to switch operating conditions by time period to adapt to different indexes.These complex characteristics make the process flexible and efficient.In order to meet these requirements,a small-scale methods of production with multistep,batch process,got the attention of people.However,noise is widespread in actual batch process,and its influence is inevitable.When these noises cannot be ignored,the performance of the control system designed with the deterministic theory will not meet the desired requirements.Therefore,anti-interference controller design strategy for the stable operation of the batch process has important theoretical significance and practical value.In addition,system output deviates from the desired trajectory may bring huge losses in the complex batch industrial control system,so the performance evaluation of control system is very important.In actual industrial process,the noise is mostly have a Gaussian distribution,so this subject flexibly improves the relatively mature control method of non-Gaussian continuous system and applied to the batch process with complex dynamic characteristics.The control and performance evaluation of batch processes with non-Gaussian noise are studied under the framework of statistical information theory.The methods adopted in this study are as follows:Firstly,the actual batch process can be divided into fast batch process and slow batch process according to the operation time of each batch.In order to break through the limitation that existing controllers can only deal with one type of process,a data-driven controller that can deal with both processes is proposed.The performance optimization index of non-Gaussian batch process is constructed by correntropy.Considering the practical operability,the sliding window and oversampling sampling methods are adopted to calculate the performance index,and the gradient descent method is used to optimize the performance index to obtain the optimal control.In addition,this article quotes the concept of "gold batch",a controlled-well batch selected from the historical batch is regarded as "gold batch",and then the performance of "gold batch" is regarded as a benchmark to evaluate the new batch,which data-driven method breaks through limitations of the model in the traditional studies.To sum up,the study on control and performance evaluation of batch process control with non-Gaussian noise is not only of great significance to the theoretical study of batch process,but also of certain guiding role to the actual industrial process,which has extensive application value.