The authors address the diagnosis problem in nonlinear systems by using the concept of the differential transcendence degree of a differential field extension. The authors also consider the algebraic observability concept of the variable which models the failure presence for the solvability of the diagnosis problem. The construction of a reduced-order uncertainty observer (unknown-input observer) to estimate the fault variable is the key step in the proposed approach. A simulation example that deals with a bioreactor process is presented to illustrate the effectiveness of the suggested approach.