SYNTHESIS OF A MULTI-MODE INTELLECTUAL SYSTEM OF MANAGEMENT OF A WEAKLY FORMED PROCESS
The characteristic features of many industrial processes of biotechnological, biotechnological, food and other industries are noted. The main ones are the difficulty of obtaining a mathematical model due to the incompleteness of knowledge of kinetic laws, the presence of various types of noniinearit...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | Russian |
| Published: |
North-Caucasus Federal University
2022-08-01
|
| Series: | Современная наука и инновации |
| Subjects: | |
| Online Access: | https://msi.elpub.ru/jour/article/view/179 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The characteristic features of many industrial processes of biotechnological, biotechnological, food and other industries are noted. The main ones are the difficulty of obtaining a mathematical model due to the incompleteness of knowledge of kinetic laws, the presence of various types of noniinearities in the mathematical model, non-stationarity, as well as significant structural and parametric uncertainty that manifests itself in the process of functioning. The synthesis of a neural network process control system was carried out taking into account the multi-mode of its functioning in conditions of uncertainty. It is proposed that the process flow modes be identified using characteristic changes in the rate of change in the concentration of automatically controlled carbon dioxide dissolved in the culture medium in the air leaving the apparatus. Based on the analysis of the results of training the neural network by various methods, the feasibility of applying the Levenberg-Marquardt algorithm, which provides greater accuracy and high convergence rate near the minimum, and, therefore, can significantly accelerate the training procedure, has been established. Using the Levenberg-Marquardt algorithm, we trained neural network models of the control object, each of which, at certain modes of the process, is closest to the actual state of the object. This allowed us to implement the basic principle of multimode control systems, which consists in switching control algorithms with neural network models when changing process modes, which allows us to provide specified system quality indicators in each of the modes under changing operating conditions. |
|---|---|
| ISSN: | 2307-910X |