Application Of Multivariate Statistics In The Process Of Continuous Casting Mold Detection Research

The present thesis breaks up into two parts: the analysis of mold heat flux for round billet continuous casting and the application of multivariate statistical method for monitoring of mold process during continuous casting.The billet quality is determined greatly by the mold thermal behavior. In the first part, the average and the local mold heat fluxes are simultaneously analyzed for round billet continuous casting, based on the measurements of heat flux in plant trial, focusing on the influence of the main operational parameters on the heat flux, such as casting speed, EMS current, mold level, and so on, and the discussion of distributions of average mold heat flux and that in high heat flux region.The development trends of high-speed continuous casting and non-defects billet require the on-line monitoring of the key parameters of mold process. The second part of this thesis introduces the multivariate statistical method, analyzing and diagnosing the mold process in continuous casting. The abnormity in mold is monitored, using the key factor analysis method, and the results show that this method can precisely identify the abnormities in process and diagnose faults using monitoring map.The research of PCA, based on the neural network,is eliminating the system non-linearity using neural network, integrating the neural network and principal component analysis. The simulation results indicate that NNPCA has a distinct improvement in the monitoring and diagnosing of process abnormity.