Modern i nd ustr y has bee n e xperiencing rapid developme nt i n a uto matio n a nd i ntelligence to meet the conti nuo usly i ncreasing perfor ma nce req uireme nts on prod uct q uality and reliability, eco no mic benefi ts, and safety. T his ma kes moder n i nd ustrial systems, such as power systems, electronic machines, robotics, chemical systems, etc., become more and more large, comprehensive, and complex. Various advanced control and optimization methods have been developed for these comple x ind us trial syste ms to achieve satisfactory system performance, however, the design and implementation of these methods rely heavily on the availability of an accurate system model. Therefore, modeling is a prerequisite for system analysis and synthesis. The system model can be obtained by using the first principle method through logic derivations based on physical and chemical laws.

However, the complexity and the lack of prior knowledge of modern industrial systems greatly limit applications of the firstprinciple method. System identification provides an alternative for complex industrial system modeling and recovers the mathematical model of systems from informative process data. Modern industrial systems are typically nonlinear or time-varyingand ma y have time-delay property which makes the parameter identification and structure identification of these systems very challenging; The practical process data may irregular, dual-rate, or multirate sampled and are usually corrupted by the white/color noise, outliers, missing data, measurement delays, etc, which have imposed extra difficulties on complex industrial system identification. Therefore, modeling and identification for complex industrial systems taking the practical problems in process data set intoconsideration are challenging and meaningful tasks.