31 July 2019 in Multimedia Tools and Applications
18 August 2020 in Multimedia Tools and Applications
6 October 2026 in SSRN Electronic Journal
Adjoint design sensitivities depend on the constitutive model supplied to the adjoint. Agreement with a finite-difference gradient verifies that derivative. It does not show that every constitutive model still compatible with the measurements supports the same design direction.For steady conduction with a movable heat source, thermal resistance converts the state, the objective, the design gradient, and the curvature into exact moments. The reduction reveals exact decision-equivalence classes of physically distinct conductivity fields. A finite change of resistance splits into an exact linear term and an exact remainder.A discrete conjugate heat-transfer channel has no such reduction. The flux derivative inside the discrete adjoint is analytic. Central differences of the reduced objective match it to 3.1×10⁻¹² at step 10⁻⁵. Sensor and monitor temperatures are interpolated to fixed physical coordinates. The reported quantity is the critical measurement tolerance at which the feasible design gradient reaches zero. On a 96×64 mesh that tolerance is 0.01073 with three wall sensors and 0.01509 when the heated-face temperature of the ambiguous conductivity strip is added. The same ordering holds on 12×8, 24×16, 48×32, and 96×64, as a numerical search to a stated tolerance. On the coarse mesh, a Taylor enclosure with an explicit fourth-derivative remainder puts both signs in the three-sensor feasible set and only the negative sign in the four-sensor set. All channel observations are synthetic temperatures from the nominal discrete model.