A multi-source heterogeneous data monitoring method based on latent subspace
A high level of monitoring is necessary for the safety and product quality of the electrical fused magnesia furnace (EFMF). In this paper, a monitoring method based on latent subspace for EFMF is proposed to fully mine the effective information of multi-source heterogeneous data in the process. By minimizing the distance of different types of data in the subspace, the corresponding projection matrix is obtained. Then the data is projected into the obtained subspace to estimate whether fault occurs.In summary, the main contributions of this paper are threefold. (1) Heterogeneous data is applied to monitoring EFMF, and information collaborative processing of different source data is realized through common latent subspace. (2) The statistics of the system space and the residual space are weighted and fused to reduce the false alarm rate without affecting the missing alarm rate. (3) An EFMF fault-grading scheme is proposed to enhance the practicability of the algorithm and reduce the workload of operators. The proposed method is applied to the data of the EFMF and compared with the traditional algorithm. The experimental results show that the method is effective.
Train set contains 300*7 normal measurement data and 300 figures，which is used to training model.
Test set contains 200*7 measurement data and 200 figures，which is used to test model.