Research On Fault Diagnosis Of Batch Process Based On Modified FDA Of Kernel Function Theory

Process monitoring is the important guarantee of safety production in industrial production, and it plays an important role. in improving product quality and economic benefit. Statistical process monitoring is a data driven method whose theoretical basis is multi-variate statistical theory. Process monitoring theory dectects and diagnoses the abnormal situation in the process through the analysis of monitoring data,ensuring product quality and production efficiency.According to the characteristic of batch process,process monitoring method based on the Fisher discriminant analysis is improved.The main research contents are as follows:1. The singular matrix may occur when. diagnosing fault in the complex batch process. A singular value decomposition algorithm based on kernel Fisher discriminant analysis is proposed to the above problems in this paper.Firstly,the original data is mapped from the original space to a high dimensional space by using the kernel function.Secondly, the processed data is mapped into the non singular orthogonal matrix by means of singular value decomposition. Finally, process monitoring and fault diagnosis theory is realized through the Fisher discriminant analysis algorithm.2. Industrial process data from multiple data sources or heterogeneous data sets,and the effect of processing data based on the single kernel method is not ideal.The combination kernel and kernel Fisher discriminant analysis algorithm is proposed in this paper. The effectiveness of the algorithm is verified by the experiment of beer fermentation.3. The kernel local Fisher discriminant analysis algorithm is proposed,which could analyze of global Euclidean distribution structure and the local epidemicdistribution structure of the sample data.The superiority of the algorithm is verified by the experiment of beer fermentation.