Conventional models of continuum mechanics adequately describe only rather slow transfer processes. Experimental studies of high-rate and short-duration processes found out that the system’s response significantly lags behind the impact of force and is accompanied by self-organization of internal structure on the mesoscale. Far from local equilibrium the dynamic structure evolves towards a more stable state of the system, obeying certain internal laws. Unlike usual continuum models, the description of such processes cannot be localized both in space and time. The new interdisciplinary approach developed by the author, based on nonlocal transport equations with memory obtained in nonequilibrium statistical mechanics, describes structural evolution on the mesoscale using principles developed in the theory of control of adaptive systems. According to the Maximum Entropy principle, the goal of system evolution is to maximize the total entropy in the system. Speed Gradient principle determines the fastest way to the goal under constraints imposed and forms closed loops of internal control due to feedback between the structure evolution and macroscopic properties of the system. By transforming its dynamic structure, the system adapts to external influences and minimizes its irreversible losses. So, the inclusion of self-organization and closed loops of internal control is necessary for adequate modeling of processes that are far from local equilibrium. The proposed “flexible” modeling makes the description of nonequilibrium processes closed and provides an effective way to solve a wide range of modern practical problems related to the development of new technologies, intelligent systems and forecasting of catastrophic processes.