Background: Patents suggest that fault diagnosis is an important technology for maintenance,
which is usually employed to avoid system catastrophic damage and ensure reliability.
Objective: As traditional diagnosis method was limited in linear systems, a novel diagnosis method was
proposed to address hidden failure in nonlinear system.
Methods: This method was based on the combination of strong tracking square root unscented Kalman
filter (STSRUKF) and sequential residual χ2 (chi-square) test. After discussing some related patents and
methods, the χ2 test was introduced to check the STSRUKF abnormal residual, and this test solved the
problem of inability to locate fault for SRUKF. Firstly a χ2 test was used to locate faulty parameters via
detecting STSRUKF residual outputs generated by all the potential faulty models. Secondly, the located
faulty parameters were estimated by STSRUKF to indicate the faulty degree.
Results: A simulation case was used to verify the reliability of the presented approach. The experiment
results indicated that the proposed algorithm could locate faulty parameter accurately, and the
STSRUKF estimation accuracy was higher than square root unscented Kalman filter (SRUKF) and
strong tracking unscented Kalman filter (STUKF).
Conclusion: The method provides references for hidden fault detection and parameter estimation.