Konokh I. Models and methods of effective management of continuous processes of raw material processing based on multi-agent systems

The work is devoted to solving the current scientific and applied problem of creating the theoretical foundations of effective continuous technological processes automated intelligent control. The solution is based on efficiency indicators for ensure compliance of production results with market conditions. The object of research - the processes of automated control of technological installations of continuous operation with maximum efficiency. The work is based on the proven statement that the efficiency of the operation, as the degree of compliance of its results with the goals of the enterprise owners, is determined through the cost estimates of energy, resource, time and the cost estimate of the finished product, taking into account technological constraints. The transition to cybernetic models of operations for continuous technological processes with channel transformation of raw materials has been carried out. The criteria for optimization of modes based on efficiency indicators have been synthesized and verified. General models of product systems have been developed, which include modules for calculating the duration of the operation, resource consumption, energy, associated costs for the transport and processing parts. The cost of the finished product is also calculated on the basis of quantitative and qualitative indicators, as well as the degree of their compliance with technological and organizational restrictions. Basing on the synthesized models, it is possible to compare the available technological modes from the efficiency view point, to search for the global optimum of their functioning, and to form controlling influences. It is proved that the use of the synthesized efficiency criterion requires the development of new methods for searching for an extremum both for a one-stage process and for multi-stage ones. The paper presents specialized search methods based on computational models of processes and allows us to perform optimization using an efficiency indicator. In order to implement the methods of optimal control in terms of efficiency indicators, specialized models of multi-agent systems have been developed. This allows to automatically identify the state of a dynamic process and perform predictive optimal control of essentially nonlinear, non-stationary objects. Using the efficiency criterion, it has been proved that the synthesis of a multi-agent control system is more expedient than using artificial neural networks of known architectures for similar purposes. It has also been proven that the obtained multi-agent structures provide automatic control of the validity of the classical control systems functioning for discrete and continuous processes, increase the reliability of their operation and reduce the cost of implementing such control.