Identification Of The Change Of Dynamic Characteristic In The Multi-Phase Batch Process Monitoring

Owing to their operational flexibility, batch and semi-batch modes are preferred in a number of industries which usually process the low-volume, high-value added and variety products including pharmaceuticals, specialty chemicals, flavors and colorants products.The process monitoring and fault diagnosis are considered as the most important issues in the process industry. Benefited from the rapid development of process instrumentation and data acquiring, multivariate statistical process control (MSPC) methods, especially the principal component analysis (PCA), have been widely used for process monitoring and fault diagnosis. The PCA model is built to detect the running process, determine its operational status, guide the manufacture effectively, and improve product quality.In this thesis, the characteristics of batch process and the status of monitoring methods are analyzed and reviewed. Combined with the characteristics of batch processes and statistical method, a new batch process monitoring method is proposed. The main works and results include:(1) Describe the characteristics of batch process and data, and the concept, significance and methods of fault diagnosis. The development of process monitoring, its content and classification are introduced, especially the improved method of the MSPC. At the same time, Multi-way Principal Component Analysis (MPCA) which is widely used in batch process monitoring is discussed systematically.(2) Batch processes with multiple phases are commonly found in process industries. Process dynamics and correlations among variables also change with the phase transition. Conventional applied single MPCA model is obtained from the whole batch process, which is not sufficient to model the different process dynamic and variable correlation changes. A new phase identification method is proposed in this thesis based on the change of the first cumulative contribution between different PCAs. Fed-batch penicillin cultivation process is used as a case study; and monitoring result comparison is made between the proposed method and conventional single MPCA model. The result show that the method proposed in this thesis is able to monitor the fault in time.(3) The monitoring method is different in continuous process and batch process, but the dynamic characteristics and correlation of the data will also change in the continuous process, when fault occurs, which is similar to multi-phase in batch process. Phase identification method proposed in this thesis is applied to the monitoring on a municipal solid waste process (MSW). Results show that this method can make judgments on the failure and guide the production.