Adaptive Coefficients Multi-view Projection and Its Application for Fault Detection

Fault detection technology is playing an increasingly important role in industrial processes. Therefore, a fault detection algorithm based on adaptive coefficient multi-view projection is proposed in this paper. This method uses the consistency and complementarity of hidden features in the multi-view data to enhance the accuracy of the model. Meanwhile, the manifold learning idea is used to maintain the local geometric structure of the data. The most important thing is to introduce the spectral radius to assign the weight coefficient of each view and different views, which reduces the influence of artificially set parameters and the time for artificial parameter adjustment when training the model. On this basis, considering the non-Gaussian of industrial data, monitoring statistics suitable for this method are proposed in the feature subspace and residual subspace separated from the original data. A real industrial process proves the validity of the detection method.