A Research Of Blast Furnace Fault Identification Based On Sparse Matrix

Iron and steel industry is a significant industry, which is related to national securityand lifelines of the national economy. The blast furnace process is a vital iron and steelsmelting process. It takes place in a closed environment and there occurs complexphysical and chemical reactions, which will cause large fluctuations. In case the blastfurnace fault happens, it will cause drastic life and economic losses. Thus ensuring thestability state of the blast furnace will be very meaningful for fault diagnosis of blastfurnace. In this paper, propose a new fault feature recognition method, sparse matrixfactorization algorithm is applied to the fault identification. According to the sparsematrix’s high computing speed and small storage capacity, boost the fault recognition rateto some extent. Main work in this paper are as follows：Firstly, analyzes the process of blast furnace production process and the faultsymptom of blast furnace. The fault diagnose model of blast furnace mainly includes thehybrid dynamic mechanism model, the iron-making process model, the furnace conditiondiagnosis reasoning model, the system optimization model, a model of holistic troublediagnoses, based on these models, this paper proposes a new fault identification method ofblast furnace——blast furnace fault identification method based on sparse matrix, thismethod includes: preprocessing of blast furnace fault identification system’s input, fuzzyprocessing of input, fault feature recognition of blast furnace, construction of fuzzy matrix,mathematical model of the nonnegative matrix factorization, non-negative matrixfactorization with sparseness constraints(NMFSC) algorithm study, and the blast furnacefault identification system experiment study. Finally, the system implementation of blastfurnace fault recognition includes: recognition system design, system test and analysis.In terms of this paper, make a combination of fuzzy matrix and sparse matrixfactorization algorithm to identify blast furnace fault. In the special field of blast furnacefault identification, non-negative matrix factorization with sparseness constraintsalgorithm has larger advantages. This algorithm can extract fault feature efficiently, reducethe false alarm rate and missing report rate, thus improve rate of fault identification.Practice shows that the method proposed by this paper can achieve good result andaccurately reflect the blast furnace internal fault condition in no time, provide firmguarantee to normal production of blast furnace. The blast furnace can carry on theproduction safely, stably and effectively. Improve the blast furnace productivity, producegreater efficiency as well.