Novel Multidimensional Feature Pattern Classification Method and Its Application to Fault Diagnosis

Abstract

With the development of modern process industries, data-driven fault diagnosis methods have attracted more and more attention. In this paper, a novel nonlinear fault diagnosis method based on multidimensional feature pattern classification (MDFPC) is proposed. The proposed MDFPC method integrates multikernel independent component analysis (MKICA) with adaptive rank-order morphological filter (AROMF). First, some dominant independent components capturing nonlinearity are extracted from the historical process data using the MKICA algorithm, getting the template signal and the testing signal of each fault pattern. Then, a multidimensional signal classification method based on AROMF is developed to achieve the diagnosis of fault patterns. The effectiveness of the proposed fault diagnosis method is demonstrated by carrying out a case study using the Tennessee Eastman process.