Application Of Fault Detection Method Based On 2d-dkpca In Penicillin Fermentation

Pharmaceutical industry is very important to a country. It is not only a main part of a country's economic, but also a special welfare work which including prophylaxis, treatment and health care. So it has been paid high attention in all countries.Fermentation is indispensable in pharmaceutical industry. The process of microbial fermentation is very complicated which has high nonlinearity, dynamic and relativity. So an important premise of improving production efficiency is to develop an affective fault detection method.At present, there have been several fault detection methods based on multivariable statistical method. For example, the Principal Component Analysis (PCA) method and the Kernel Principal Component Analysis (KPCA) method. They have good performance when monitoring linearity system or the static system. But they can not capture the nonlinearity and dynamic between variables and batches, so they have poor effect when monitoring systems with nonlinearity and dynamic like fermentation process.Under the background of penicillin fermentation in industry, the paper introduces several fault detection methods of fermentation at present. On the basis of the thorough analysis of the feature of microbial fermentation and accumulating the experiences, the paper put forward a new fault detection method which is based on Two Dimensions-Dynamic Kernel Principal Analysis (2D-DKPCA).Using this method and combining the practical process of penicillin fermentation, the paper design a fault monitoring software to inspecting the process of penicillin fermentation. The main functions of this software including online monitoring, fault alarming, recording sampling data and inquiring fault record.