Neural network optimization in automated control systems for complex technological processes and production

Annotation: The complexity and responsibility of decisions made when managing potentially hazardous and risky operations, as well as operations with a high cost of errors, often preclude the com-plete automation of complex technological process control. The article presents the problems of control methodology, along with the main provisions and stages of the methodology for case-based adaptation of a multi-dimensional technological process control loop to changing operating conditions. A de-scription of the neural network adaptation procedure and princi-ples of choosing the type and topology of the neural network adjusting the control loop of technological installations is pro-vided. The conditions for optimal utilization of multilayer neu-ral network architectures and formulas for calculating the num-ber of their interneuronal connections and layers are given.

Keywords: pi-controller, neural network, precedent management, optimization, technological process