Abstract. The main problem of automation consists in the unity of the solution of two tasks: the definition of the control object model (identification) and the design of the controller (regulator) for the object on the basis of this model. To the great regret, in most cases the calculated regulators in the system together with real objects behave completely differently than they should, according to the calculation results. The problem is that when identifying is inevitable, the model of the object is determined with an error. First, as a rule, the mathematical expression of the model of the object is adopted in a more simplified form than is actually the case, since the model of any object, even the most simple one, in practice, if to take into account all its features, is extremely complicated. Second, identification methods rely on measurement results, and any measurement is inevitably carried out with inaccuracy, and therefore the coefficients of the mathematical model are not determined accurately. Third, the parameters of a model of even an almost stationary object can, in the course of its functioning, vary somewhat in time, either depending on the prescribed equilibrium point, or depending on external conditions. Most often, this problem of the discrepancy between the results of calculation of actual results is either ignored in publications or solved by improving the form of the transient process to the desired by iterative way, that is, experimentally, on the basis of primitive algorithms or intuitively. This situation should be considered unsatisfactory. To eliminate this negative phenomenon, it is proposed to forcefully limit the speed of the obtained model of the object by introducing an additional delay link, and then use the new obtained model to calculate the regulator. One of the most effective methods of calculation is the method of numerical optimization. This paper gives example of the successful use of the proposed method, which proves its effectiveness.