OBRABOTKAMETALLOV Vol. 23 No. 3 2021 MATERIAL SCIENCE EQUIPMENT. INSTRUMENTS 7 5 All formed information feature set (IFS) parameters Π = {Π1, Π2, … Π6} T∈ ℜ Π 6 have the following properties. First, IFS parameters are dimensionless, their values are usually positive, and they change monotonically as wear progresses. Second, wear w = 0 in all cases corresponds to the condition Πi(w) = 0, when w = 0. Third, from the variety of information features, those highly sensitive to changes in wear have been selected. Their area of application depends on the following. The information feature Π1 can be used if, as wear progresses, the equilibrium of the dynamic cutting system is asymptotically stable. It suffices to note that the internal amplification factor in the dynamic system depends not only on ρ, but also on the cutting depth. It changes to a lesser extent with variations in cutting speed and feed rate. It has been shown earlier that the evolution of the cutting system properties can be highly sensitive to small variations in technological parameters and disturbances. It can be concluded that the redistribution of amplitudes is highly sensitive to wear development. However, this holds true within the range of stable equilibrium, as well as in cases of low sensitivity of the system evolution to variations in initial system parameters and disturbances. To use the information feature Π1 for diagnosing wear on CNC machines, it is necessary to coordinate the diagnosis with the CNC program. Information assessments Π2 and Π3 are more universal but less sensitive to variations in the dynamic properties of the system and to changes in operating modes. When constructing diagnostic systems for turning structural steels at constant cutting modes and stable elastic deformation equilibrium, it is possible to divide wear information into 4–5 wear classes [27]. However, all three characteristics depend on the accuracy of the specific machine and its condition. Previous studies of the coherence function between force disturbances and deformations have shown that it increases with increasing frequency [3]. When selecting the space of information features, the features Π4, Π5, and Π6 are more resistant to interference. In this case, the main disturbances are associated with variations in the allowance, spindle group runout, and kinematic disturbances from the feed drives. All disturbances originating from the machine lie within the low-frequency range. At the same time, when installing the AE sensor, it is necessary to take into account the wave properties of the channel connecting the force emission generated in the cutting zone and the measured vibrations (oscillatory deformations, displacements, accelerations, etc.). There is a general rule here: the higher the frequency, the closer the measuring transducer should be to the cutting zone. In our opinion, specialized cutting tools with integrated measurement transducers hold the most promise for AE measurement. The presented method for determining diagnostic signs of tool wear and increased vibration activity of executive elements has practical significance for the creation and development of intelligent algorithms for monitoring systems. Changes in the assessments of diagnostic signs of wear serve as corrective parameters that can be used to build an adaptive control algorithm in the CNC unit, capable of extending the tool’s service life, but are not limited to this. Within the framework of the presented methodology, the tasks of determining diagnostic signs of degradation of the geometric topology of the part surface and the operational characteristics of the machine tool are of research interest, but these are subjects for future studies. The identified and collected information features in the VAE signals allow the creation of “Test Data” and “Training Data” databases for training the assessment of cutting process dynamics in machine learning models of diagnostic systems, which is another step towards the digital transformation of the machine tool industry. Conclusion The developed methodology, mathematical modeling, and digital and full-scale experiments have made it possible to form a rational information space for wear diagnosis, in which, based on known recognition methods, decisive rules can be constructed to classify information according to its correspondence to specific wear levels. For the practical application of the methodology in algorithms for monitoring, diagnosing, and controlling the cutting process, it is important to consider the following points. The basis for constructing tool wear diagnosis systems based on the observation of measurable vibration sequences can be both the dependence of vibrations on changes in the parameters of the dynamic connection
RkJQdWJsaXNoZXIy MTk0ODM1