Development of the concept of control system work for plaster works based on neural network

Abstract

Preparation of mortar, its submission to the place of work and application to the surface is efficiently carried out by means of automated mechanical systems in tandem with concrete pumps combined with plastering machine units. The technical consistency and competitiveness of robotic plastering stations largely depends on the accepted kinematic models of the plastering robot being developed and the development of advanced control systems based on them. Simple kinematic circuits generate simple design solutions that are quickly prototyped and of low cost, but because of their design imperfection, these solutions are commonly used for limited production processes, have low precision and primitive control systems, which do not allow for deep automation production. Complex kinematic circuits of robots allow them to be used as a versatile tool in many stages of construction, but such design solutions require the creation of sophisticated control systems.The use of a neural network allows the architecture of the robot control system to be compromised, making it a versatile robot for performing various technological procedures, but requires a clear algorithm for describing the mathematical model of work and its training.This study proposes a mathematical model of a neural network based on a sigmoidal activation function for a two-link manipulator.