Využití neuronových sítí v oblasti spolehlivosti a údržby průmyslových zařízení

Abstract

The dissertation thesis "Use of neural networks in dependability and maintenance of industrial facilities" interferes with the concept of Industry 4.0, specifically focusing on the interconnection of artificial neural networks and operational dependability issues. The interconnection is applied on early control systems in area of printed circuit boards production. Early control systems represent one possibility to improve systems reliability, whereas they can involve implementation of progressive and sophisticated methods – in this concrete case machine vision with artificial neural networks. Also, thy can minimize economical expenses when defected or non standardized product occurred. In terms of this thesis were proposed methodology to improve optical control of produced printed circuit boards, where artificial neural networks are used to eliminate picture shift, before machine verification is applied. Proposed methodology is in dissertation thesis tested on specific printed circuit boards acquired from standard production line intended to automotive area. The testing was carried out in the laboratory conditions with usage of technical instrumentations, which is intended for industrial use. Obtained solution can be applied only on this early control system, nevertheless proposed methodology and procedures has general validity for any optical control with machine vision use. Obtained results confirm precarious presumption about possibility of use artificial neural networks as a modern and progressive toll which can be used in industrial systems reliability area. The interconnection with another technical instruments can be used for early control systems and also for early warning systems, reliability control systems, maintenance and quality.