Online Steady State Detection Based on Rao-Blackwellized Sequential Monte Carlo

Online detection of whether a data stream has reached the steady state is known to be an important problem in many applications such as process control, data reconciliation and fault detection. This paper introduces a novel online steady state detection algorithm under the Bayesian framework based on a multiple change‐point state space formulation and the sequential Monte Carlo methods. A Rao‐Blackwellization technique is proposed to substantially reduce the variance of Monte Carlo estimation and greatly enhance the computational efficiency. In addition, a resampling scheme called the Optimal Resampling is used for eliminating duplicate samples and the robustness of steady state detection is significantly improved by using the information of the particles more efficiently. Numerical studies based on simulated signals and application to a real data set are used to evaluate the performance of the proposed method and compare with other existing methods from the literature. The proposed method is shown to establish a more robust performance than other methods. And it is much more computationally efficient than the standard sequential Monte Carlo method.