In order to solve the problem of ensuring the required level of reliability of modern radio electronic systems for space, aviation and special purposes on the basis of ultra-high-speed heterostructure nanoelectronics devices, methodology was elaborated aimed at design, development and technological optimization of such devices. It is proposed to use a special efficiency preservation coefficient as the target function, which makes it possible to take into account parametric alterations within the assigned operational tolerance and is characterized by practice-oriented advantages in comparison with the so-called integral probability of performing the given functions. The problem of reduced accuracy in forecasting electrical properties of semiconductor structures in nanoelectronics devices with a low dimensional channel is solved. A gerontological physical and mathematical model for AlGaAs-heterostructure nanoelectronics devices with transverse transport was developed, which made it possible to solve the problem of limited applicability of the methods established by the classical statistical theory of reliability to ensure increased requirements to durability of such devices in electronic systems for space purposes operating in severe conditions and operating modes.

Devices based on nanoscale nonlinear elements with the transverse transport of charge carriers are widely used in various modern radio-electronic systems. However, at present, neither a time-optimal nor an accuracy-optimal approach has been developed for predicting the operating parameters with regard to the time factor and external influence of such devices. In this paper, we propose a variant of solving the problem of modeling the degradation of a nanoelectronic device with the transverse transport of carriers on the basis of a feed-forward artificial neural network. The use of neural-network approaches in the simulation of resonant-tunneling diodes can significantly (by several orders) increase the speed of such models. Training of the developed artificial neural network by data obtained from the reliability tests of low-dimensional semiconductor heterostructure diodes allows an increase in the model accuracy by several times by taking into account the influence of technological errors (occurring during manufacturing) and various degradation changes (occurring with time and under the influence of external factors during the operation of resonant-tunneling diodes).