RU2783860C2 - Device and method for vibration-acoustic analysis of industrial equipment containing rotating parts - Google Patents

Abstract

FIELD: analysis.

SUBSTANCE: invention is used for vibration-acoustic analysis of threshing and crushing equipment. A device for vibration-acoustic analysis of threshing and crushing equipment includes at least one sensor for obtainment of vibration data, data processing means with the possibility of obtainment and extraction of data on equipment operational modes from data obtained from sensors, while the device is equipped with at least one sensor for obtainment of data on acoustic oscillations and machine learning means made in the form of a neuromorphic chip and connected to data processing means for recognition and/or storage of equipment operational modes.

EFFECT: provision of a possibility of reliable recognition of abnormal and potentially emergency operational modes of industrial equipment with rotating parts.

17 cl, 4 dwg

Description

The proposed technical solution relates to systems for vibration and / or acoustic control of industrial equipment operating modes, in particular, rotation mechanisms using machine learning technology. A significant part of industrial equipment contains rotating parts such as rotors, turbine shafts, bearings and similar mechanisms and parts as main or auxiliary elements. All of them, to one degree or another, are subject to wear, which during the operation of the equipment can cause unwanted vibration or acoustic vibrations that propagate throughout the equipment and into the environment. Some types of fluctuations indicate serious malfunctions in the operation of the equipment, for example, the imminent failure of some parts, especially bearings of various types. This can lead to premature failure of the equipment itself, thereby damaging the production process. Vibration and acoustic analysis allows you to detect a malfunction at an early stage of its development. In some cases, it helps to determine the very cause of unwanted vibrations. However, vibration and acoustic patterns are a normal part of equipment operation due to the presence of many moving parts and mechanisms. A change in the vibroacoustic signature of the equipment may be associated with the transition from one normal operation mode of the equipment to another, caused, for example, by a change in the shaft rotation speed. Therefore, systems and methods are needed to analyze the operating modes of the equipment in order to identify modes that are potentially dangerous for the normal operation of the equipment.

Traditional methods for monitoring the condition of equipment are based on detecting a situation when the values of vibration velocity or vibration acceleration exceed threshold values. This type of analysis requires data that represents both normal and abnormal operation of the equipment, that is, specific values are needed for certain components or assemblies. Such data are obtained by calculation during the design of the installation or empirically during testing.

During the search, among the array of technical solutions, inventions were found that solve similar technical problems. So, a technical solution was found under the patent of the Russian Federation for the invention No. 2691225 "Device for measuring and evaluating the technical condition of equipment of the machine-building complex" (MPK G01M 7/00, DSTU, RF, s. No. 2018123580, 06/28/2018, publ. 06/11/2019) . According to the description, the invention relates to measuring devices, namely to devices for monitoring the technical condition of an object. The device consists of a housing including a monitoring module on a board with a microcontroller, an information storage module, an information analysis module, an information input/output module and a transceiver connected to external auxiliary equipment, a current sensor, a voltage sensor and a measuring head housing with sensors installed on its housing temperature, noise and vibration.

The disadvantage of this technical solution is mobility, which does not provide for long-term monitoring of the state of the equipment, and requires the presence of an operator who performs the measurement, processing and interpretation of the measurement results. It is known that defects in equipment with rotating parts do not appear sporadically, but over a certain, sometimes long, time interval and, thus, when using this technical solution, it is impossible to assess the development of a defect over time.

An evolutionary stage in the development of equipment diagnostic systems is the use of modern neural network models capable of conducting a deep analysis of its operating modes. Known technical solution according to the EU patent for the invention No. 3186598 "Monitoring of a device having a movable part (monitoring of a device having a moving part)" (IPK G01M 13/04; G05B 19/4065; Siemens AG, Germany, S. No. EP 14873124 A, 02.12.2014, file 02.12.2014, published 05.07.2017). According to the description of the technical solution, the invention relates to the monitoring of a device having a moving part, in particular a rotating part, the device comprising: a control module configured to receive a vibration signal of the device, provided with a sensor connected to the device providing the vibration signal spectrum, spectrum, to determine the base frequencies and spur frequencies, while the base frequencies are frequencies whose peak power corresponds to the natural frequencies of the device or fault frequencies, and the spur frequencies correspond to other frequencies. Then, base frequencies and side frequencies are processed, separately applying one-class classification to the base frequencies and side frequencies, combine the results of one-class classifications, thereby obtaining a classification signal that represents a confidence level of 1, and output a decision support signal based on the classification signal, this support signal decision indicates the error status of the monitored device.

However, the invention is also not without drawbacks, including the inability to receive information about the sound accompaniment of equipment operation, since it is known that the sound corresponding to any defect appears in the early stages of equipment damage, until the moment when deviations in the level of vibrations that occur can be fixed. inherent in this type of defect. The disadvantages also include the requirement to know the frequency of failure, which is a disadvantage when trying to detect previously unknown defects.

Closest to the claimed device is the system according to US patent for invention No. 7363111 "Methods and systems for analyzing engine unbalance conditions (Methods and systems for analyzing engine unbalance conditions)" (IPC G05B 23/02; G01M 1/38; Boeing Co, USA , file No. US 70843707 A, February 20, 2007, file December 30, 2003, published April 22, 2008). The invention discloses methods and systems for analyzing an engine imbalance condition. In one embodiment, the method includes receiving data from vibration sensors from a plurality of locations distributed throughout the engine and surrounding engine support structure, inputting said data into an inverse neural network model. The model establishes a relationship between the vibration data and the motor imbalance condition and outputs diagnostic information indicating the motor imbalance condition. In a further embodiment, the method includes processing the vibration data with a Fast Fourier Transform to extract the desired vibration data once per revolution before being fed into the inverse neural network model. At the same time, the disadvantages of this technical solution include the lack of mention of the possibility of retraining the neural network in real time, as well as the need to install multiple sensors and the lack of analysis of acoustic signals.

Thus, the current level of technology does not contain information about devices and methods that solve the problem of creating a manufacturable and reliable device and its method of operation, which make it possible to improve the ability to recognize abnormal and potentially emergency modes of operation of industrial equipment with rotating parts, which will reduce the failure rate due to early detection faults, reduce downtime and avoid costly repairs to key pieces of equipment. An additional technical result is the portability of the device, which allows expanding the scope of its application.

In the general case of implementation, a device for vibroacoustic analysis of equipment containing rotating parts is characterized by the presence of at least one sensor for obtaining vibration data, data processing means with the possibility of obtaining and extracting data on the operating modes of the equipment received from sensors, at least one sensor for obtaining data on acoustic vibrations and machine learning tools associated with data processing tools for recognizing and / or remembering the operating modes of the equipment.

Features common with analog are the presence of at least one sensor for obtaining vibration data, data processing means with the possibility of obtaining and extracting data on equipment operation modes received from sensors.

In general, the device is characterized in that it is provided with at least one sensor for obtaining data on acoustic vibrations and machine learning tools associated with data processing means for recognizing and/or storing equipment operation modes.

In the first particular case, the device is designed to be mounted on the equipment in such a way that the sensors capture the vibration and acoustic data of the rotating parts.

In another particular case of execution, machine learning tools are configured to cluster data.

In the following particular case, the device is provided with a power source.

In the following particular case, the vibration sensor is configured to receive vibration data along three axes.

In the development of this particular case, the vibration sensor is made in the form of an accelerometer.

In another particular case of execution, the sensor for obtaining data on acoustic vibrations is made in the form of a microphone.

In the next particular case, data processing facilities are implemented as a system on a chip.

It is expedient to implement machine learning tools in the form of a neuromorphic chip.

In another particular case of execution, the extracted data are mathematically converted signals from the sensors, formed into a feature vector.

It is advisable to supply the device with means of outputting information about the operating modes of the equipment.

The device in the general case of implementation contains sensors that dynamically receive data on the vibrational and acoustic components (signature) during operation of the equipment. These devices may be, for example, as indicated in particular cases, an accelerometer and a microphone, respectively. Then the data in the form of signals in digitized form is sent to the processing means in the form of a computer system with a processor capable of analyzing, mathematical transformations of the received data and transmitting them in processed form to machine learning tools. Under the processed type of data in the context of the present invention is meant a feature vector X, consisting of components extracted by mathematical, in particular, fast Fourier transforms and wavelet transforms, from signals received from vibration and acoustic vibration sensors. By means of machine learning, the feature vector incoming for analysis is recognized and checked for the condition of being known for the specified machine learning tools. In case of uncertainty, the feature vector is recorded and then clustered. The cluster is subsequently assigned a category. The use of machine learning tools for the analysis of the feature vector, built on the basis of a neuromorphic processor, allows you to quickly and dynamically analyze the feature vectors coming from the data processing tools and obtain real-time information about the equipment operation modes. This information is needed to assess the condition of the equipment and make maintenance decisions.

In some cases, the device can be equipped with its own power supply, which allows you to create a compact solution that is energy independent from external sources, which is advisable to use as a portable diagnostic equipment to check the condition of moving parts and mechanisms. However, the device can also be used as an integral part of the equipment control system. In this case, there are no requirements for its energy autonomy and transportability. Thus, it can be integrated into the electrical circuit of said control system.

The use of compact computer systems as data processing tools, in particular SoC (System-on-crystal), makes it possible to increase the manufacturability of the device, its computational efficiency and reduce the cost of manufacturing.

Providing the device with means of outputting information about the operation of the equipment refers to the use of various interfaces for connecting to peripheral equipment. It is proposed to use Ethernet as a standard interface. Additionally, a Wi-Fi interface can be implemented for connecting to mobile devices based on Android® and iOS® systems.

The device may contain a case in which all of its components are placed - this solves the problem of portability. The power supply may be integral with the case, or the device may be provided with mains power. The sensors, in the preferred embodiment, are located inside the case to ensure portability of the device. When the sensors are made inside the housing, the latter can be equipped with a pin installed in the appropriate mounting hole on the equipment to provide better contact for the transmission of vibroacoustic vibrations.

In general, the absence of a large number of nodes and elements of the device ensures a high level of reliability.

The group of inventions also includes a method partially disclosed in the description of the device above. The method of vibroacoustic analysis of equipment containing rotating parts is characterized by the following features: obtaining information about vibration and/or acoustic vibrations of rotating parts of equipment by means of data processing, mathematical processing of the obtained data with feature extraction, generation of a feature vector and transmission to machine learning tools for recognition and/or or memorization.

In the first particular case, the method is characterized by the fact that machine learning tools cluster the recognized feature vectors.

In clarifying this particular case, the method is characterized by the fact that by means of machine learning clusters are assigned categories corresponding to the operating modes of the equipment

In another particular case, mathematical processing includes the extraction of features from the received data on vibration and/or acoustic vibrations by means of fast Fourier transforms, wavelet transforms, empirical mode decomposition, empirical cumulative distribution function.

In the next particular case, the data received by the sensors is digitized before analysis by means of data processing.

In the fourth particular case, the method provides the output of information about the operating modes of the equipment to the means of displaying data.

In another particular case, if it is established by machine learning tools when recognizing the transmitted feature vector that if the feature vector is not learned by machine learning tools, then the specified feature vector is memorized by machine learning tools and subsequently clustered.

The basis of the method is a block of mathematical transformations provided by extracting features from the data coming to the processing means from sensors, further recognition of the feature vector generated from the results of these transformations in the form of a feature description of data on vibration and acoustic vibrations that changes dynamically, and clustering by means of machine learning of feature vectors into groups united by common signatures. Thus, by reasonable means, the problem of training and further recognition of the operating modes of the equipment is solved. Clusters can be assigned a category that allows you to navigate in the recognized modes, and it is possible to display information, for example, about abnormal modes. As noted in the description of the device, the extraction of features from the initial data is carried out by, in particular, fast Fourier transforms, wavelet transforms, which allow to extract the components that characterize it from the signal from the sensors. The categorization of clusters is carried out in order to identify abnormal and potentially emergency modes of equipment operation. The features extracted from the data coming from the sensors are obtained by mathematical transformations, including fast Fourier transforms, wavelet transforms, empirical mode decomposition, empirical cumulative distribution function. Transformation of data by these methods makes it possible to highlight the key features of the signals received from the sensors, which subsequently allows, based on the generated vectors, to determine the presence of an atypical operating mode in the vibroacoustic signature of equipment.

In order for the information to be perceived by the operator, the method can provide for the output of information, for example, about the abnormal operation of the equipment through an external interface to a peripheral device.

During the operation of the method, at the stage when the feature vectors are recognized by machine learning tools, the latter are programmed so that the (new) feature vectors unknown from previous iterations are recorded by the specified tools and automatically determined into the corresponding clusters. If the recorded vector cannot be assigned to an existing cluster, then a new cluster is created and the specified vector is assigned to the newly created cluster. This allows you to identify new modes of operation of the equipment, which may be abnormal and lead to malfunctions of its rotating parts.

The present invention is illustrated by the following figures:

Fig. 1 device with integrated sensors and external power supply;

Fig. 2 - device with remote sensors;

Fig. 3 - device with remote sensors and external power supply;

Fig. 4 block diagram of the vibroacoustic diagnostics algorithm.

In figures 1-3, the following positions are indicated:

1 vibration sensor;

2 - sensor of acoustic vibrations;

3 - data processing facilities (SoC);

4 RAM;

5 machine learning tools (neuromorphic processor);

6 - power supply;

7 interface;

8 - Wi-Fi and/or Bluetooth module;

9 - body.

The content of the blocks in figure 4 is given in the description below when disclosing private implementations of the present group of inventions in terms of the operation of the method.

To understand the principles of operation and features of various implementations of a group of inventions, the following is a description of the figures of the technical solution. Although the text of the description explains in detail the preferred embodiments of the technical solution, it should be understood that other embodiments of the group of inventions are also possible. Accordingly, there is no need to limit the scope of legal protection of a technical solution solely to the presented implementations and lists of components and components. However, when describing the preferred technical solutions, for a clear understanding of the basic principles of the invention by a specialist, it is necessary to clarify the terms used in the description.

It should be noted that the components and parts of the device used in the singular in the description and in the claims also represent plural forms, unless the opposite is expressly stated. For example, an indication of a composite element of the device also means an indication of the totality (set) of such elements.

Also, when describing the preferred embodiments, specific terms are used to ensure clarity of understanding. The term is intended to be used in the broadest sense in which it can be construed by those skilled in the art and to include all technical equivalents used in the same manner and for the same purpose. So, in particular, the phrase "equipment containing rotating parts" means devices in which, during their operation, some parts and / or mechanisms perform rotational movement. Such devices may include equipment containing shafts, turbines, drums, threshing and crushing devices. Vibration refers to mechanical vibrations of solids in the frequency range 1.6-1500 Hz, in relation to the present group of inventions, vibrations include vibrations that occur when equipment is operated with parts moving, mainly by rotation. Acoustic vibrations are understood as vibrations of an elastic medium: air, solids causing the propagation of elastic waves in them in the range from 6 to 20,000 Hz. Under the means of data processing is understood a set of software and hardware designed to receive, analyze, process, convert data, in particular, coming from vibration and acoustic vibration sensors. In general, these means can be made in the form of an electronic device equipped with a processor and memory that executes the corresponding program code. Under the extraction of information is meant the implementation of such mathematical transformations on the data received from sensors, the result of which leads to the possibility of obtaining information about the mode of operation of the equipment. Machine learning means a software and hardware system that uses neural network models to recognize, record and group data from sensors subjected to mathematical transformations. Clustering is understood as the distribution of a given sample of objects into non-overlapping subsets so that each cluster consists of similar objects, and the objects of different clusters differ. A feature vector is understood as a data input method for a machine learning tool, in which a vector is formed, composed of values (features) corresponding to a set of features that describe an object. The object, in the context of the present invention, are vibration and acoustic data obtained from sensors. The values from which the vector is formed are obtained by performing mathematical transformations of the specified data. The term memorization, referring to the operation of machine learning tools, characterizes the recording and subsequent clustering of the feature vector received for analysis.

The words "comprising", "comprising", "including" mean that at least the specified component, element, part or step of the method is present in the composition, object or method, but does not exclude the presence of other components, materials, parts, steps of the method , even if such a component, material, part, step of the method performs the same function as the specified one.

The materials from which the various elements of the present group of inventions are made, indicated below in the description of examples of a specific implementation of the system, are typical, but not mandatory for use. The components of the system indicated in the present examples of execution can be replaced by numerous analogues that perform the same function as the examples of the systems and methods used in the description given in the description.

Hereinafter, an electronic device is described that implements the vibroacoustic analysis method for detecting faults in bearings and gears of industrial equipment. The device uses an unsupervised self-learning algorithm for early detection of faults in industrial equipment based on a combined algorithm for vibroacoustic analysis of signal clustering and further recognition using the NM500 neuromorphic chip. At the stage of adjustment and training, the device studies data without having information about the types of failures, and when the studied modes of operation of nodes or assemblies converge to a certain number, it is assumed that the device has studied all the normal modes of operation of the equipment. Then, during operation, the distances between clusters of input processed data are compared with previously learned data (cluster centers), using the NM500 chip. If the numerical vector, composed of the features of the current cluster, characterizing the mode of operation of the equipment, does not coincide with the previously learned vectors, compiled according to the same principle, a conclusion is made about the presence of a new, previously unknown mode of operation of the equipment. Thus, the device is capable of detecting anomalous new data that was previously unknown. At the same time, it is assumed that the categories (clusters) discovered and learned at the training stage are categories that can be attributed to the normal operating modes of the equipment. If during operation of the device after the completion of the learning stage, the newly formed cluster that appeared during feature recognition belongs to the category of new and unknown, this is reported to the maintenance personnel.

The proposed device relates to electronic devices with internal software designed to process vibration acceleration along three axes and a sound signature coming from a three-axis accelerometer and microphone located inside the case.

In one of the special cases of implementation, in accordance with the schematic diagram of the components presented in Fig. 1, the device is structurally made in a housing with dimensions of 50 × 50 × 40 mm, at the base of which a metal threaded stud is fixed, which serves for fastening on industrial equipment (screwing into equipment) through which vibration and acoustic vibrations of the controlled equipment are transmitted. The housing can be made of any material that does not have significant vibration-absorbing properties. In the particular case under consideration, the body base is made of metal. An accelerometer (1) is installed inside the case, which records vibration vibration readings, a broad-spectrum microphone (2) for capturing acoustic vibrations of equipment during operation. These sensors are connected by electrically conductive circuits to a ZYNQ 7020 single-chip system (3) made on a printed circuit board containing appropriate connectors for connecting peripheral equipment, in this case, sensors, and a RAM module soldered on the board (4). RAM (4) can also be soldered on the board without the possibility of extraction. The system (3) is connected via a cross-connector on the printed circuit board to the NM500 neuromorphic processor (5). The system is powered by a built-in power supply (6) connected to the printed circuit board through the appropriate power circuit. The printed circuit board also has at least one interface connector (7) for transferring processed data to other devices, such as personal computers, terminals, mobile devices. In a particular case, an Ethernet connector is installed as an interface.

In another particular case, illustrated schematically in FIG. 2, the accelerometer (1) and the microphone (2) are placed outside the case. Accordingly, with such an implementation, either connectors for connecting cables leading from sensors must be made in the case, or the device must be equipped with components that provide wireless data transmission from sensors (1, 2) to a single-chip system (3). These can be transceivers operating according to the Wi-Fi standard, Bluetooth or other similar standards. At the same time, it is necessary to provide for high noise immunity of data transmission protocols and the availability of own power supplies for sensors (1, 2). Otherwise, structurally, the device repeats that described in Fig. one.

Further, in another particular case, schematically shown in FIG. 3, the device, in the rest of the design, can be made as in FIG. 1 as well as in Fig. 2, however, the difference lies in the use of the Wi-Fi/Bluetooth module (8), which can be made in the case. Using this module, it is possible to transmit diagnostic information about the state of the equipment in the form of, for example, error codes or insecure modes to mobile devices or terminals.

The device works as follows. In a particular case, the pin of the device body is installed in the corresponding slot of the equipment being diagnosed. During operation of the equipment, vibration and acoustic vibrations propagating through the equipment and into the environment are captured by a three-axis accelerometer (1) and a microphone (2), digitized by the sensors themselves, or transmitted for digitization to a ZYNQ 7020 single-chip system (3) for analysis. System (3) performs mathematical transformations with the received data, extracting features from them and forming a vector from the extracted features, thus compiling a feature description of the object. RAM serves to execute the program, store the program being executed, as well as to store the input, output and intermediate data processed by the system (3). The resulting feature vector is passed to the NM500 (5) neuromorphic processor for analysis by machine learning tools, which recognizes the feature vector and clusters it. Next, the distances between the input processed data clusters are compared with the previously learned data (cluster centers) using the NM500. If the feature vector, composed of the features of the current cluster, characterizing the mode of operation of the equipment, does not match the previously learned vectors, compiled according to the same principle, a conclusion is made about the presence of a new, previously unknown mode of operation of the equipment.

The method according to Fig. 4 consists of the following operations. Step 100 power supply and run the algorithm (fixing the device on the equipment being diagnosed), provided that the device is mounted on the equipment. At step 101, initial data is entered for specific equipment; in blocks 110, 120, data is received in dynamic mode from vibration sensors, along three axes, and acoustic vibrations, respectively. In block 111, the initialization check is carried out - the procedure for assigning the parameters of mathematical algorithms necessary for subsequent data processing. If initialization has not been performed, then in blocks 200 ... 202, the parameters of mathematical algorithms are initialized in accordance with the initial parameters of the equipment known in step 101. By converting the data at the output of block 203, a description of the signal characteristics is obtained, which allows further analysis. It should be noted that diagnostic information about the current mode of operation of the monitored equipment is not available until the equipment is monitored for a certain time. This is necessary to dynamically determine the upper and lower limits of the incoming signal and adjust to the characteristics of the input signal and must be performed continuously during operation of the device.

In blocks 300, ... 302, various mathematical transformations are performed by data processing tools, in particular, fast Fourier transforms, wavelet transforms, empirical mode decomposition, empirical cumulative distribution function to extract features from the data received for processing and generate a feature vector. The generated feature vector 303 is input to block 304, where further the machine learning tools recognize and check the knownness of the specified vector 305. If the vector is known, then step 310 discards it and returns to step 303, where a new feature vector is generated. If the feature vector is unknown, then it is written by machine learning tools and at step 321 it is determined that it belongs to one of the existing clusters. If the feature vector can be written to an existing cluster, then at step 330 a write is made, then a return is made to the step of obtaining a new feature vector 303. If the vector cannot be placed in an existing cluster, then at step 340 a new cluster is created, the vector is written to created cluster 341 and return to step 303.

Embodiments of the present invention are not limited to the above specific examples. Other forms of implementation of the technical solution may be proposed, without deviating from the meaning of the invention. Embodiments of the invention may provide for the implementation of the device with remote sensors, as schematically shown in FIG. 2, 3, the implementation of a device with a Wi-Fi or Bluetooth wireless interface, the ability to install a piezoceramic ultrasound sensor to obtain an acoustic signature in the ultrasonic range.

The exemplary embodiments disclosed above are given in order to show the industrial applicability of the device and to give a general impression of the capabilities of the proposed apparatus. The scope of legal protection of a technical solution is determined by the claims, and not by the description presented, and all changes made using equivalent features fall under the legal protection of the present invention.

Claims (17)

1. A device for vibroacoustic analysis of threshing and crushing equipment, including at least one sensor for obtaining vibration data, data processing means with the possibility of obtaining and extracting data on the operating modes of the equipment received from sensors, characterized in that it is equipped with at least one a sensor for obtaining data on acoustic vibrations and machine learning tools made in the form of a neuromorphic chip and associated with data processing tools for recognizing and/or storing equipment operation modes. 2. The device according to claim. 1, characterized in that it is made with the possibility of mounting on the equipment in such a way that the sensors capture the vibration and acoustic data of the rotating parts. 3. The device according to claim. 1, characterized in that the machine learning tools are made with the possibility of data clustering. 4. The device according to claim 1, characterized in that it is equipped with a power source. 5. The device according to claim. 1, characterized in that the vibration sensor is configured to receive vibration data in three axes. 6. The device according to claim 5, characterized in that the vibration sensor is made in the form of an accelerometer. 7. The device according to claim. 1, characterized in that the sensor for obtaining data on acoustic vibrations is made in the form of a microphone. 8. The device according to claim. 1, characterized in that the means of data processing are made in the form of a system on a chip. 9. The device according to claim 1, characterized in that the retrieved data are mathematically converted signals from the sensors, formed into a feature vector. 10. The device according to claim. 1, characterized in that it is equipped with means of outputting information about the modes of operation of the equipment. 11. A method for vibroacoustic analysis of threshing and crushing equipment, including obtaining information about vibration and/or acoustic vibrations of rotating parts of equipment by means of data processing, mathematical processing of the obtained data with feature extraction, generation of a feature vector and transmission to machine learning tools made in the form of a neuromorphic chip , for recognition and/or storage. 12. The method according to claim 11, characterized in that the machine learning tools cluster the recognized feature vectors. 13. The method according to claim 12, characterized in that, by means of machine learning, clusters are assigned categories corresponding to the operating modes of the equipment. 14. The method according to claim 11, characterized in that mathematical processing includes extracting features from the received data on vibration and/or acoustic vibrations by means of Fourier transforms, wavelet transforms, empirical mode decomposition, empirical cumulative distribution function. 15. The method according to claim 11, characterized in that the data received by the sensors is digitized before analysis by means of data processing. 16. The method according to p. 11, characterized in that it provides the output of information about the operating modes of the equipment to the means of displaying data. 17. The method according to claim 11, characterized in that if, when recognizing the transmitted feature vector, it is established by machine learning that if the feature vector is not learned by machine learning, then the specified feature vector is memorized by machine learning and subsequently clustered.

Images