A new control algorithm for nonlinear, non-stationary multichannel plant, convenient for practical use, is obtained. The ideological basis of the algorithm construction is the compensation of external additive influences on the state variables and output variables of the plant with the accuracy of etalon filters by means of the inverse model of this plant. Unobservable perturbations are evaluated by the mismatch between the corresponding plant variables and the inverse model. Controlled plant is represented as a system of ordinary differential equations in normal form with the same number of output variables and control actions. Functional algebraic equations are presented, solution of which yields the inverse model and the etalon filters. The solution leaves a certain freedom of choice of the etalon filters, which determine the dynamic properties of the control system. The structure of the control system is composed, according to the compensation principle in which the output and state variables of the plant are used for the perturbation estimation. As a result, there is a feedback in the system, which was not postulated initially, but was the result of evaluation and compensation of perturbations with the accuracy determined by the etalon filters. This approach made it possible to determine the structure and parameters of the controller by analytical method using physically obvious initial data. А consequence of the compensation and filtering of perturbations affecting the state variables is the correction of the plant’s own dynamics. The effectiveness of the proposed algorithm is shown by examples.с

The article deals with the problem of identification and tuning of mechatronic systems with state controllers using artificial neural networks (ANN) in conditions of wideband interference in measurement channels by optimizing the structure and parameters of a neural network identifier. The method based on application of radial ANN is proposed to reduce the tuning time of systems with state controllers. Discrete values of dynamics characteristics of system state coordinates are connected to the identifier inputs, according to which the ANN estimates of the variable parameters of the controlled object. Based on the estimates obtained, the parameters of the state controller are automatically calculated using the modal control method. The optimal composition of measurement channels is selected based on the ratio of information signal and interference powers, where signal is considered as the difference of dynamics characteristics with nominal and changed values of the object parameters. For each variable parameter, a state coordinate is found that gives a maximum value of the criterion, after which the optimal ANN structure is formed, providing a minimum identification error under interference conditions. In case of small variations in the parameters of object, it is recommended to determine the power of information signal using sensitivity functions of state coordinates. With large variations in parameters and different powers of noise in the measurement channels, direct calculation using matrix expressions provides more accurate estimates. The developed training algorithm allows us to determine optimal value of radial basis functions spread, which ensures a given accuracy in identifying the object parameters with minimum number of neurons in the network first layer. The proposed method of systems tuning makes it possible to obtain a given quality of control in conditions of parametric uncertainty of mechatronic object. At the same time, the optimal combination of measurement channels at ANN input according to the proposed criterion provides a minimum value of the identification error in conditions of wideband interference.

Nowadays the number of unexpected accidents depends to a great extent on the qualification of the personnel, who on the basis of the accumulated experience and the information received from the control systems intuitively determine the current technical condition of the object and the beginning of possible accidents. Sometimes their decision is too late and accidents occur. Usually, accidents occur as a result of occurrence and development of various malfunctions and at this time objects go into the latent period of emergency state, there noise emerges, correlated with the signal g(t) under control. Here, depending on the nature of the malfunction the dynamics of technological processes changes, under the influence of which, the spectra of useful signals X(t) and noise ε(t), g(t) = X(t) + ε(t) also change. When using the traditional technology, the noise is filtered. This introduces an additional error in the result of signal analysis, leading to a loss of important diagnostic information. To solve the problems of control of the beginning of the latent period of accidents it is proposed to use the estimate of the cross-correlation function between the useful signal and the noise as an informative attribute, using the technology of adaptive sampling of the analyzed signals g(t). To ensure the adequacy of the results of control by combining the proposed and traditional technologies information is formed about the beginning of the latent period of accidents, which allows the master to determine the time of the beginning of the latent period of accidents with sufficient reliability.

The paper proposes a system for model predictive temperature control of steel strip in the galvanized sheet metal production. Based on a review of proposals for predictive strip temperature control, we show that one of the main problems is the lack of reliable information about the current temperature state of the heat treatment section, since the temperature of the workspace is controlled locally at individual points. Taking this into account, the proposed system is based on the use of generalized estimates of the temperature of the workspace. This allows for precise predictive control in the absence of complete information about the current temperature state of the section. A generalized temperature estimate is given using a simplified interpretable model based on strip temperature values at the inlet and outlet of the section. To determine the feed forward impact on the power of heating and cooling systems of sections when compensating for disturbances in the product range and line speed, we propose a hybrid model. The model consists of an interpretable and an empirical component based on an artificial neural network. Using the example of a closed cooling section of a continuous hot-dip galvanizing unit, we studied the performance of a generalized temperature estimation stabilization system with a standard PID controller. We identified difficulties due to the complexity of simultaneous effective processing of disturbances according to the task, as well as disturbances caused by errors in the feed forward compensation of disturbances according to the model. To quickly stabilize the generalized temperature estimate in the event of modeling errors, the structure of the control system with two degrees of freedom was chosen. The closed-loop PID controller is configured to handle disturbances caused by modeling errors during predictive control. The direct open control loop is configured to handle disturbances based on a given value of a generalized estimate of the workspace temperature. The performance of the proposed system is demonstrated using the example of a closed strip cooling section. It is shown that the system is able to guarantee control quality even with the maximum possible errors of the hybrid model. The developed hybrid model can also be used to plan the dynamics of temperature changes in the workspace in the long term. The proposed model structures and control systems can also be used for heating sections.

The paper solves the problem of identifying the features of the interaction of underwater manipulators (UM) links, which autonomous underwater vehicles (AUV) are equipped with, with a viscous medium. It is shown that for precise control of the UM, viscous friction, as well as the added masses and moments of inertia of the UM links during its movement must be taken into account as correctly as possible. The procedure for identifying the parameters of interaction between a moving UM and a viscous medium is carried out before starting technological operations near work sites. These parameters are further used in the calculation of force and torque effects from the moving UM on the AUV body for its high-precision stabilization in the hover mode over the objects of work, as well as for the implementation of feedbacks of synthesized positional-force UM control systems that ensure automatic execution of underwater contact technological operations. In the first part of the article a dynamic model of UM, based on recurrent equations for soving the inverse problem of dynamics, is presented in the form of linear regression. In this model unknown parameters enter linearly into the equations of generalized forces (moments) acting in the joints of UM. This type of UM model allows the identification of all unknown parameters using a linear Kalman filter. The results of the study of the complete mathematical model of AUV with UM, performed in the second part of the article, confirmed the operability and high efficiency of the proposed method for identifying the parameters of the interaction of UM units with a viscous medium