Building a Simulation Model of a Neonatal Incubator with a Neural Controller

A mathematical model for simulation studies of neonatal incubators and their control systems was presented. The mathematical model is based on a numerical description of the physical and chemical processes in the incubator. The incubator control system is organized using a neuro-controller. The neuro-controller is built on the basis of a deep neural network such as a multilayer perceptron. The functioning of the neurocontroller organized through a gradient-based neural network control algorithm. The ultimate goal of modeling is to select the optimal incubator structure and parameters of the control system. Therefore, the mathematical model is focused on simulation studies of the functioning of a neonatal incubator under the disturbing influences as close as possible to real ones. Verification of the mathematical model and implementation of the control system with a neuro-controller was carried out using the neonatal incubator “ARDO Amelie’’, the “Raspberry Pi 4B 4GB RAM’’ mini-computer and the author’s made controlled hydrodynamic neonatal phantom. The software was implemented using the Python programming language.