Detection of Gross Errors Using Mixed Integer Optimization Approach in Process Industry

Mei Cong-li,Su Hong-ye,Chu Jian
DOI: https://doi.org/10.1631/jzus.2007.a0904
2007-01-01
Abstract:A novel mixed integer linear programming (NMILP) model for detection of gross errors is presented in this paper. Yamamura et al. (1988) designed a model for detection of gross errors and data reconciliation based on Akaike information criterion (AIC). But much computational cost is needed due to its combinational nature. A mixed integer linear programming (MILP) approach was performed to reduce the computational cost and enhance the robustness. But it loses the super performance of maximum likelihood estimation. To reduce the computational cost and have the merit of maximum likelihood estimation, the simultaneous data reconciliation method in an MILP framework is decomposed and replaced by an NMILP subproblem and a quadratic programming (QP) or a least squares estimation (LSE) subproblem. Simulation result of an industrial case shows the high efficiency of the method.
What problem does this paper attempt to address?