摘要：The paper deals with a global supervision system for pipelines which involves the diagnostic of leaks and faulty sensors. By assuming a set of nonlinear partial differential equations for the fluid and measurements of flows and pressures at the ends of the pipeline, nonlinear analytical redundancy primary relations are generated which achieve a signatures’ matrix for five faults. Since the redundancy primary relations for the study case depend on time derivatives of known signals, additional manipulations are required for their implementation. Two algorithms are applied to manage this issue: one consists of transforming the primary relations by using proper filters such that a state space realization for the residual is achieved, and the second one consists of a substitution of the derivatives in the relations. The time derivatives are obtained by employing a uniform robust exact differentiation algorithm based on sliding modes. The comparison of the features of the residuals for the five faults with real data of a pilot plant of 200 [m] shows the advantage from the filtering of the redundancy relation with respect to the substitution of an exact time differentiation.
关键词：Multiple-faults isolation in a pipelinenonlinear redundancy relationssliding modes time differentiatorfiltering of nonlinear redundancy relations
