This paper deals with an information-theoretic approach to process condition monitoring based on generalized models. The overall conclusion regarding the process status under this approach is made by means of a running comparison of provisional rated values that comply with the nominal model of a process with matching characteristics based on the current observations. The paper suggests using a consistent measure of stochastic dependence of random processes as a process behaviour indicator. Namely, quadratic Tsallis mutual information is proposed as such a consistent measure of dependence. Its features are considered in the light of problems the condition monitoring of technological processes.