Scientific and technological progress has significantly changed the very process of developing technical devices and significantly complicated their principles. Many modern systems are designed with using knowledge-cognitive systems, and they themselves use hard-formalized and fuzzy rules, often based on big data. Accordingly, there is an urgent task of adapting the modern education system to these changes. Soon, knowledge can't be transferred to trained engineers within the framework of existing teaching technologies; the skills of using, extracting and transforming knowledge become the basis for promising technologies of continuous self-education. This article suggests a model of the structure of cognitive management systems based on control levels. The levels of this model reflect the procedures for transforming of knowledge forms. As practical steps to introduce such an educational approach for engineers, it is proposed to increase the amount of research work of students of technical specialties. The article contains examples of the formulation of problems requiring research in the study of physics, electrical engineering, signal processing and other disciplines. The most simple to implement is the use of a "black box" to conceal the structure of the object for investigation or use non-standard modes and settings of well-known devices, for example, ADCs. Also, approaches are considered for assessing quality of knowledge that obtained under the cognitive control system.