For a gentle aircraft landing, it takes a skilled and experience pilot to harmonize the control activities with the state and position of the aircraft, while considering external disturbances (wind, rain, snow, signalization, etc.). An error at any time in the control process or significant disregard of the external factors can result in a rough landing or, in the worst case, a crash. A similar analogy can be applied to the process of monitoring the state of tools and machinery as the foundation of flexible autonomous production. The processing procedure is exceptionally dynamic and complex, and is reliant on a series of variables. These cannot be considered separately, but rather demands a holistic approach in which the characteristics of the machine, processes, tools, treatment and environment are intertwined. A partial examination of the system and the inability to transfer individual solutions makes the monitoring of processing procedure a limiting factor in the global applicability of autonomous systems. Adaptation to the environmental and processing conditions, and simple and fast learning are the key components for success in future supervision systems, and an important assumption in the transferability of supervision methods. As such, this paper is the result of an attempt to give a comprehensive overview of the tool monitoring process, and to propose a possible solution. The first part of the dissertation describes the structure of processing machines, followed by a description of the processing procedure, focusing on milling. The forms and mechanisms of wear and the methods of receiving the processing signal are analysed. A group of characteristics and the procedure for selecting the dominant characteristics in relation to the observed state are analysed. The influences of tool wear on individual characteristics are also analysed. An analysis of existing supervision systems is given, and a system based on a modularly oriented structure is proposed to allow for the development of software modules for various areas of work. A basic assessment is given of parameters of variable tangential axes as the means for separating out the mechanical part of the noise from digital signals of plant machine processing systems, in order to avoid the need for additional sensors and hardware. The procedure for determining the function of processability is described, as a possible means of supplementing measured data in detecting tool fraying. The structure of the modular neuron network, built by local experts specialized in individual areas of expertise, is described as the foundation of future systems. In the second part of the dissertation, the proposed methods are verified through experimental research. The selection of wear properties is analysed under different experimental conditions. The influence of a combination of various properties on monitoring quality is tested. The influence of changes in hardness and depth of the processed material, and the behaviour of the proposed monitoring system are analysed. The potential of the tool monitoring system, integrated into the machine management system, is investigated. The non-parametric method of assessing the strength spectrum in tool fraying for the milling process is tested.