Nonstationary Process Monitoring based on Cointegration Analysis with a Switching Scheme

摘要

Owing to the equipment aging, varying production plans and environment, many measurements of industry processes are nonstationary. In this paper, a nonstationary process monitoring for blast furnaces based on cointegration analysis with a switching scheme is proposed. A switching scheme is used to divide the process into to two working conditions according to whether the hot-blast stoves are switching. Cointegration analysis is conducted on each working condition to construct stationary variables from original nonstationary measurements respectively. Then, monitor the process in each working condition based on principal component analysis. A practical case study demonstrates the effectiveness of the proposed method.