Apparatus and method for monitoring electro- and/or mechanical systems

摘要

An apparatus and associated methodology that monitors a system. The method determines scheduling parameters of a system. The system includes a plurality of subsystems. The method determines diagnostic parameters of the system and generates a linear parameter varying (LPV) model. The LPV model relates an input, the diagnostic parameters, and the scheduling parameters to an output. In addition, the method monitors each of the plurality of subsystems as a function of the LPV model.

主权项

1. A method of fault diagnosis comprising:
determining, using processing circuitry, scheduling parameters of a system wherein the system includes a plurality of subsystems; determining, using the processing circuitry, diagnostic parameters of the system; generating, using the processing circuitry, a linear parameter varying (LPV) model wherein the model relates an input, the diagnostic parameters, and the scheduling parameters to an output; calculating a first vector estimate xQo associated with a first matrix by applying
{circumflex over (x)}Q0=(ρ(ξ){circle around (x)}ψ0T(k))†y0(k)where ρ(ξ) is a deviation vector, ψo is a data matrix, and yo is an output vector;
identifying the LPV model by estimating recursively using singular value decomposition matrices using data obtained by perturbing one or more diagnostic parameters and the first matrix which is determined with no perturbations, wherein the identifying includes applying the following formulation:
{circumflex over (x)}Qi=(ρ(ξ){circle around (x)}ψiT(k)Δγi)†(yi(k)−ψiT(k){circumflex over (Q)}0ρ(k));{circumflex over (x)}Qij=(ρ(ξ){circle around (x)}ψijT(k)ΔγiΔγj)†(yij(k)−ψijT(k)({circumflex over (Q)}0(ξ)+{circumflex over (Q)}i(ξ)Δγi+{circumflex over (Q)}j(ξ)Δγj)ρ(ξ));and{circumflex over (x)}Qijk=(ρ(ξ){circle around (x)}ψijkT(k)ΔγiΔγjΔγk)†(yijk(k)−ψijkT(k)({circumflex over (Q)}0+{circumflex over (Q)}i(ξ)Δγi+{circumflex over (Q)}j(ξ)Δγj+{circumflex over (Q)}k(ξ)ρ(ξ)Δγk+{circumflex over (Q)}ij(ξ)ρ(ξ)ΔγiΔγj)ρ(ξ)) where xQi, xQij, xQijk are vector estimates associated with matrices; monitoring each of the plurality of subsystems as a function of the identified LPV model; scheduling the LPV model to track variations in an operating regime; and diagnosing one more faults of one or more components of the system using the generated LPV model.