Apparatus for diagnosing rotationary machines based on classification of transient and steady states, and method of the same

The rotary machine diagnostic device and method classify transition and normal states using signal and valve information to reduce misdiagnosis in variable speed motors, enhancing fault detection accuracy by differentiating between transitional and normal states.

KR103046755B1Active Publication Date: 2026-11-11REPUBLIC OF KOREADEFENSE ACQUISITION PROGRAM ADMINISTATION
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
KR1020240076930
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2024-06-13
Publication Date
2026-11-11
Estimated Expiration
2044-06-13

AI Technical Summary

Technical Problem

Existing rotating machine diagnostic methods struggle to accurately distinguish between transitional and normal states, particularly in variable speed motors like pumps and compressors, leading to misclassification and missed fault diagnoses due to changes in current and vibration signals.

Method used

A rotary machine diagnostic device and method that utilize a preprocessing algorithm to classify transition and normal states based on signal information and valve opening/closing information, employing a signal collection unit, state classification unit, and rotary machine diagnostic control unit, including a learning unit trained with Deep Support Vector Data Description or One-Class Support Vector Machine to differentiate between transitional and normal states.

Benefits of technology

Effectively reduces errors in diagnosing abnormalities by distinguishing between transitional states and fault signals, ensuring accurate fault diagnosis in rotating machinery by considering the degree of valve opening and closing, thereby improving diagnostic precision.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112024063860610-PAT00004_ABST
    Figure 112024063860610-PAT00004_ABST
Patent Text Reader

Abstract

A rotary machine diagnostic device and method are disclosed, wherein a rotary machine diagnostic device determines whether to diagnose an abnormality in the rotary machine by classifying a transition state and a normal state based on a signal related to the rotary machine using a preprocessing algorithm. A rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention includes: a signal collection unit that collects a signal related to the rotary machine; a state classification unit that classifies the rotational driving state of the rotary machine into a transition state and a normal state based on signal information related to the signal and valve opening / closing information related to the opening / closing state of a valve related to the rotary machine; and a rotary machine diagnostic control unit that determines whether to perform an abnormality diagnosis of the rotary machine according to the rotational driving state classified for the rotary machine by the state classification unit. The rotary machine diagnostic control unit diagnoses an abnormality in the rotary machine when the rotational driving state of the rotary machine is in the normal state, and withholds the abnormality diagnosis of the rotary machine when the rotational driving state of the rotary machine is in the transition state.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to an apparatus and method for diagnosing abnormalities in a rotating machine, and more specifically, to an apparatus and method for diagnosing an abnormality in a rotating machine by classifying a transition state and a normal state based on signal information and valve opening / closing information related to the rotating machine using a preprocessing algorithm. Background Technology

[0002] Rotating machines, such as electric motors and generators, are utilized in various industries for human convenience. Rotating machines can be used for various purposes, for example, by converting electrical energy into kinetic energy or kinetic energy into electrical energy. If defects occur in the stator, rotor, bearings, or other components constituting the rotating machine, abnormalities occur in the electrical signals applied to or generated by the machine. Furthermore, mechanical abnormalities also occur within the rotating machine, causing mechanical vibrations to increase significantly compared to the normal state. For this reason, when diagnosing abnormalities in rotating machines, electrical signals applied to or generated by the machine are acquired, or vibrations of the machine body itself are received as vibration signals via vibration sensors and utilized for diagnosis.

[0003] Variable speed rotary machines are generally found in induction motors, and these induction motors can be utilized in pumps, compressors, etc. The change in rotational speed, or RPM (Rotation per minute), of a variable speed rotary machine is proportional to the AC frequency of the current signal. In the case of constant speed rotary machines, 50 / 60 Hz current is generated or utilized in accordance with national standards to rotate at a constant speed. However, since the rotational speed of a variable speed rotary machine varies depending on its application, the main frequency of the AC current generated or utilized also changes accordingly.

[0004] In variable speed rotary motors, the motor's RPM changes according to the flow rate (gas volume) of the inhaled fluid, and consequently, the load on the rotary machine also changes. The characteristic of variable speed motors, such as pumps and compressors, where rotational speeds vary, makes motor diagnosis difficult. This is because changes in current and vibration signals occur as the motor's rotational speed changes, and these variations can easily be mistaken for abnormal phenomena.

[0005] Algorithms are being researched to diagnose pump / compressor faults based on current or vibration signals in transitional states where the motor speed changes. However, there are limitations to their application in actual field settings due to reasons such as low reliability and the requirement for large amounts of training data. When the motor enters a steady state where it rotates stably at a constant speed, current-signal-based pump / compressor fault diagnosis is possible. Therefore, to diagnose pumps / compressors, a preprocessing step is required to detect transitional and steady states, and the algorithm must not confuse the transitional state with the fault state.

[0006] The rotational speed of pumps and compressors is generated by the amplitude, frequency, and phase of an alternating current signal. The frequency amplitude of the alternating current signal varies depending on the load of the pump or compressor, and the amplitude value of the main frequency can also vary depending on the degree of opening or closing of valves used to control the flow rate of fluid in the piping connected to the rotating machine. Conventional rotating machine diagnostic methods classify transitional and normal states by utilizing the amplitude of the main frequency; however, there is a risk that transitional and normal states may be misclassified based on the degree of valve opening or closing, and that a fault signal may be misidentified as a transitional signal, thereby skipping the fault process. The background technology described above merely explains the background in which the present invention was derived and does not imply that it is technology known prior to the filing of the present invention. The problem to be solved

[0007] The present invention aims to provide a rotary machine diagnostic device and method that determine whether an abnormality is diagnosed in a rotary machine by classifying transition states and normal states based on signal information related to the rotary machine and valve opening / closing information using a preprocessing algorithm.

[0008] Furthermore, the present invention aims to reduce errors in diagnosing abnormalities in rotating machinery caused by transitional states by effectively comparing transitional states and normal states, which cause significant difficulties in diagnosing rotating machinery such as variable speed pumps / compressors.

[0009] Furthermore, the present invention aims to reduce fault diagnosis errors in a rotating machine by acquiring and utilizing the degree of opening and closing of a valve related to the rotating machine, thereby distinguishing the amplitude change between the signal in the transition state of the rotating machine and the fault signal.

[0010] The technical problems to be solved by the embodiments of the present invention are not limited to those described above, and other technical problems can be inferred from the following embodiments. means of solving the problem

[0011] A rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention may include: a signal collection unit that collects signals related to the rotary machine; a state classification unit that classifies the rotational driving state of the rotary machine into a transition state and a normal state based on signal information related to the signals and valve opening / closing information related to the opening / closing state of a valve related to the rotary machine; and a rotary machine diagnostic control unit that determines whether to perform an abnormality diagnosis of the rotary machine according to the rotational driving state classified for the rotary machine by the state classification unit.

[0012] The above-described rotary machine diagnostic control unit diagnoses an abnormality of the rotary machine when the rotational driving state of the rotary machine is in the normal state; and may withhold the diagnosis of an abnormality of the rotary machine when the rotational driving state of the rotary machine is in the transition state.

[0013] Signal information related to the above signal may include the main frequency of the above signal.

[0014] The above state classification unit can be learned to classify the transition state and the steady state based on a difference vector including a frequency difference element representing a time-series change in the main frequency of the signal and a valve opening / closing difference element representing a time-series change in the opening / closing state of the valve.

[0016] The state classification unit may include: a main frequency calculation unit that calculates a first time point main frequency and a second time point main frequency corresponding to a peak from a frequency spectrum obtained by Fourier transforming the signal of the rotating machine at each of the first time point and the second time point prior to the first time point; a frequency difference element calculation unit that calculates a frequency difference element through a difference operation between the first time point main frequency and the second time point main frequency calculated from the signal of the rotating machine; a valve opening / closing information calculation unit that calculates first time point valve opening / closing information and second time point valve opening / closing information by extracting valve opening / closing values ​​corresponding to the first time point and the second time point in which the signal of the rotating machine is obtained, respectively, among valve opening / closing values ​​collected over time; a valve difference element calculation unit that calculates a valve opening / closing difference element by calculating a difference value between the first time point valve opening / closing information and the second time point valve opening / closing information calculated by the valve opening / closing information calculation unit; and a rotational driving state determination unit that determines the rotational driving state of the rotating machine at the first time point based on a difference vector including the frequency difference element and the valve opening / closing difference element.

[0017] The above signal may include at least one of an electrical signal applied to the rotating machine, an electrical signal generated from the rotating machine, and a vibration signal of the rotating machine detected by a vibration sensor.

[0018] The above-mentioned rotor is a variable speed rotor, the above-mentioned transition state is a state in which the rotational speed of the variable speed rotor changes, and the above-mentioned steady state may be a state in which the rotational speed of the variable speed rotor stably sucks in fluid and the rotational speed of the variable speed rotor is maintained at a constant level.

[0019] A rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention may further include a valve opening / closing measurement sensor for detecting the opening / closing state of the valve.

[0020] A rotary machine diagnosis method based on transition state and normal state classification according to an embodiment of the present invention comprises: a step of collecting a signal related to the rotary machine by a signal collection unit; a step of classifying the rotational driving state of the rotary machine into a transition state and a normal state by a state classification unit based on signal information related to the signal and valve opening / closing information related to the opening / closing state of a valve related to the rotary machine; and a step of determining whether to perform an abnormality diagnosis of the rotary machine by a rotary machine diagnosis control unit according to the rotational driving state classified for the rotary machine by the state classification unit.

[0021] The above-mentioned determining step may include a step of diagnosing an abnormality of the rotating machine when the rotational driving state of the rotating machine is in the normal state; and a step of withholding the diagnosis of an abnormality of the rotating machine when the rotational driving state of the rotating machine is in the transition state.

[0022] A rotary machine diagnostic method based on transition state and steady state classification according to an embodiment of the present invention may further include the step of learning a state classification model of the state classification unit to classify the transition state and the steady state based on a difference vector comprising a frequency difference element representing a time-series change in the main frequency of the signal and a valve opening / closing difference element representing a time-series change in the opening / closing state of the valve, by a learning unit.

[0023] The above state classification model can be trained by a semi-supervised learning model based on at least one of Deep Support Vector Data Description and One-Class Support Vector Machine.

[0024] The above classification step comprises: a step of calculating a first time point main frequency and a second time point main frequency corresponding to a peak from a frequency spectrum obtained by performing a Fourier transform on the signal of the rotating machine at each of the first time point and the second time point prior to the first time point; a step of calculating a frequency difference component representing the frequency change of the rotating machine by performing a difference operation between the first time point main frequency and the second time point main frequency calculated from the signal of the rotating machine; a step of calculating first time point valve opening / closing information and second time point valve opening / closing information by extracting valve opening / closing values ​​corresponding to the first time point and the second time point in which the signal of the rotating machine is obtained, respectively, among the valve opening / closing values ​​collected over time; and a step of calculating a valve opening / closing difference component by calculating a difference value between the first time point valve opening / closing information corresponding to the valve opening / closing value at the first time point and the second time point valve opening / closing information corresponding to the valve opening / closing value at the second time point. and may include a step of determining the rotational driving state of the rotary machine at the first time point based on a difference vector including the frequency difference element and the valve opening / closing difference element representing the main frequency and valve opening / closing change between the first time point and the second time point.

[0025] In the above learning step, the state classification model can be trained to distinguish between the amplitude change between the transition state of the rotating machine and the fault signal.

[0026] A rotary machine diagnostic method based on classification of transitional and normal states according to an embodiment of the present invention may further include the step of detecting the open / closed state of the valve by a valve open / closed measurement sensor for classifying the normal state and the transitional state.

[0027] According to an embodiment of the present invention, a computer-readable recording medium is provided on which a computer program for executing a rotating machine diagnosis method based on the transition state and steady state classification is recorded. Effects of the invention

[0028] According to an embodiment of the present invention, a rotary machine diagnostic device and method are provided, which classify a transition state and a normal state based on signal information related to the rotary machine and valve opening / closing information using a preprocessing algorithm to determine whether an abnormality has been diagnosed in the rotary machine.

[0029] In addition, according to an embodiment of the present invention, the transition state and normal state, which cause significant difficulties in diagnosing rotating machinery such as variable speed pumps / compressors, can be effectively compared, thereby reducing errors in diagnosing abnormalities in rotating machinery caused by the transition state.

[0030] In addition, according to an embodiment of the present invention, by acquiring and utilizing the degree of opening and closing of a valve related to a rotating machine, the amplitude change between the signal in the transition state of the rotating machine and the fault signal can be distinguished, thereby reducing the fault diagnosis error of the rotating machine. Brief explanation of the drawing

[0031] FIG. 1 is a configuration diagram of a rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention. FIG. 2 is a flowchart of a rotating machine diagnosis method based on transition state and steady state classification according to an embodiment of the present invention. FIG. 3 is a configuration diagram of a state classification unit constituting a rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention. Figure 4 is a flowchart illustrating step S20 of Figure 2 in more detail. Figure 5 is a flowchart illustrating the process of acquiring current data of a ship's motor and diagnosing abnormalities in the ship's motor. Figure 6 is an example diagram showing the change in current frequency according to the transition state and steady state. Figure 7 is an example diagram showing the change in current RMS according to the transition state and steady state. Specific details for implementing the invention

[0032] Hereinafter, embodiments of the present invention are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the invention. The present invention may be embodied in various different forms and is not limited to the embodiments described herein. It should be noted that the drawings are schematic and not drawn to scale. Relative dimensions and proportions of parts in the drawings are exaggerated or reduced in size for clarity and convenience in the drawings, and any dimensions are merely illustrative and not limiting. Also, the same reference numerals are used to denote similar features for identical structures, elements, or parts appearing in two or more drawings.

[0033] The embodiments of the present invention specifically illustrate ideal embodiments of the present invention. As a result, various variations of the illustrations are expected. Accordingly, the embodiments are not limited to specific forms of the illustrated areas and include, for example, variations in form resulting from manufacturing. All technical and scientific terms used herein, unless otherwise defined, have the meaning generally understood by those skilled in the art to which the present invention pertains. All terms used herein are selected for the purpose of further clarifying the present invention and are not selected to limit the scope of rights according to the present invention.

[0034] Expressions used in this specification, such as "comprising," "comprising," and "having," should be understood as open-ended terms implying the possibility of including other embodiments, unless otherwise stated in the phrase or sentence containing such expressions. Singular expressions described in this specification may include a plural meaning unless otherwise stated, and this applies likewise to singular expressions described in the claims. Expressions used in this specification, such as "first," "second," etc., are used to distinguish multiple components from one another and do not limit the order or importance of said components.

[0035] As used in this specification, 'module' and 'part' refer to a unit that processes at least one function or operation, and may refer to hardware components such as software, an FPGA, or one or more processors. In describing embodiments of the present invention, if it is determined that a detailed description of related known functions or known configurations may unnecessarily obscure the essence of the present invention, such detailed description may be omitted.

[0036] FIG. 1 is a configuration diagram of a rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention. FIG. 2 is a flowchart of a rotary machine diagnostic method based on transition state and normal state classification according to an embodiment of the present invention. Referring to FIG. 1 and FIG. 2, the rotary machine diagnostic device (100) based on transition state and normal state classification according to an embodiment of the present invention may include a signal collection unit (110), a valve opening / closing measurement sensor (120), a state classification unit (130), a rotary machine diagnostic control unit (140), and a learning unit (150).

[0037] The signal collection unit (110) can collect signals related to the rotating machine (10) (step S10 of FIG. 2). In an embodiment of the present invention, the rotating machine (10) may be a rotating machine such as a motor, pump, or compressor, but is not necessarily limited thereto, and may be provided as various devices that provide rotational power such as a generator, turbine, agitator, machine tool, motor, fan and blower, centrifugal separator, etc.

[0038] In one embodiment of the present invention, the rotor (10) may be a variable speed rotor, but is not necessarily limited thereto. For example, when the rotor (10) is provided as a variable speed rotor, a variable frequency drive (VFD) performs the role of converting the main frequency of an alternating current signal input to the driving circuit of the variable speed rotor, thereby controlling the rotational driving speed of the variable speed rotor.

[0039] In an embodiment of the present invention, the signal related to the rotating machine (10) collected by the signal collection unit (110) may include at least one of an electrical signal applied to the rotating machine (10) (e.g., an alternating current signal input to the rotating machine), an electrical signal generated from the rotating machine (10) (e.g., an output current signal of the rotating machine), and a vibration signal of the rotating machine (10) detected by a vibration sensor. The signal collection unit (110) may, for example, acquire the motor applied current in real time by a current sensor installed on a three-phase current cable connected to the motor of the rotating machine (10), or acquire the current generated from the rotating machine (10).

[0040] The signal collection unit (110) may include a current sensor or voltage sensor for measuring current and / or voltage signals, a pressure sensor for measuring fluid pressure in a pipe with a valve installed, a flux sensor for monitoring electrical characteristics by detecting changes in magnetic flux of the rotating machine, a power meter for measuring power consumption, an accelerometer such as a MEMS (Micro-Electro-Mechanical Systems) accelerometer for detecting vibration signals and converting mechanical vibrations into electrical signals, a velometer for measuring vibration speed of the rotating machine, a displacement sensor for measuring vibration displacement (e.g., a non-contact displacement sensor such as a laser displacement sensor, a contact displacement sensor such as an LVDT), an ultrasonic sensor for detecting noise generated in the ultrasonic frequency band to detect abnormal conditions of the rotating machine, a permanent magnet-based sensor (magnetic flux sensor) for evaluating the vibration state of the rotating machine by detecting changes in the magnetic field generated in the rotating machine, but is not limited to such examples.

[0041] The signal acquisition unit (110) can collect signals such as electrical signals and / or vibration signals related to the rotating machine through a data acquisition system (DAQ) that digitizes and stores electrical and vibration signals collected from various sensors for classifying the rotational driving state of the rotating machine, a wireless sensor network (WSN) that collects data from multiple sensor nodes using wireless communication, etc.

[0042] The signal information related to the signal of the rotating machine (10) collected by the signal collection unit (110) may include the main frequency of the signal regarding the rotating machine (10). That is, the main frequency of the signal regarding the rotating machine (10) can be utilized to classify the rotational driving state of the rotating machine (10).

[0043] The valve opening / closing measurement sensor (120) can detect the opening / closing state of the valve (20) associated with the rotor (10) so that the rotational driving state of the rotor (10) can be accurately classified into a normal state or a transitional state by the state classification unit (130) (step S10 of FIG. 2). The valve opening / closing information detected by the valve opening / closing measurement sensor (120) can be transmitted to the state classification unit (130).

[0044] A valve opening / closing measurement sensor (120) for measuring the degree of valve opening / closing can be implemented, for example, as a positioner installed on a valve actuator to detect the level of valve opening / closing, a limit switch, a potentiometer, a displacement sensor such as a linear variable differential pressure transducer (LVDT) or an ultrasonic displacement sensor to measure the displacement of a valve stem, a strain gauge attached to a valve stem or actuator to measure the deformation of the valve, a digital valve controller, a camera vision system that detects the position of the valve using a camera, but is not limited thereto.

[0045] The state classification unit (130) can classify the rotational driving state of the rotating machine (10) into a transitional state and a normal state based on signal information related to the signal of the rotating machine (10) collected by the signal collection unit (110) and valve opening / closing information related to the opening / closing state of the valve (20) collected by the valve opening / closing measurement sensor (120), according to a preprocessing algorithm during the preprocessing process before performing an abnormal diagnosis of the rotating machine (10) (step S20 of FIG. 2).

[0046] In an embodiment of the present invention, when the rotor (10) is configured as a variable speed rotor, the transition state is a state in which the rotational speed of the variable speed rotor changes (a state in which the driving state of the rotor is changing), and the normal state may be a state in which the rotational speed of the variable speed rotor stably sucks in fluid after passing through the transition state and the rotational speed of the variable speed rotor is maintained at a constant level. Since the transition state is a stage in which various states of the rotor change, if an abnormality diagnosis of the rotor is performed in the transition state, the possibility of an incorrect diagnosis may increase.

[0047] Accordingly, the rotary machine diagnostic control unit (140) determines whether to perform an abnormality diagnosis of the rotary machine (10) according to the rotational driving state classified for the rotary machine (10) by the state classification unit (130), and may diagnose an abnormality or postpone the abnormality diagnosis of the rotary machine (10) according to the rotational driving state of the rotary machine (10) (step S30 of FIG. 2). At this time, the rotary machine diagnostic control unit (140) may diagnose an abnormality of the rotary machine (10) when the rotational driving state of the rotary machine (10) is in a normal state, and may postpone the abnormality diagnosis of the rotary machine (10) when the rotational driving state of the rotary machine (10) is in a transitional state.

[0048] The learning unit (150) can learn the state classification model of the state classification unit (130) in advance to distinguish between the amplitude change between the transition state and the fault signal of the rotating machine (10). The learning unit (150) can learn the state classification model of the state classification unit (130) to classify the transition state and the normal state based on a difference vector including a frequency difference element representing the time-series change of the main frequency of the signal regarding the rotating machine (10) and a valve opening / closing difference element representing the time-series change of the opening / closing state of the valve (20).

[0049] Hereinafter, a rotary machine diagnostic device and method based on transition state and normal state classification according to an embodiment of the present invention as described above will be explained in more detail. The rotational speed of the rotary machine (10) changes according to the main frequency value of the alternating current signal related to the rotary machine (10). For example, the rotational speed of the rotary machine (10) may be generated in proportion to the main frequency value of the alternating current signal related to the rotary machine (10).

[0050] When the rotational driving state of the rotating machine (10) is in a normal state, the rotational speed of the rotating machine (10) is constant, so there should be no change in the main frequency value of the alternating current signal. In addition, as the load of the rotating machine (10) increases, the amplitude of the main frequency of the alternating current signal increases, and as the rotational speed of the rotating machine (10) increases, the load tends to increase. That is, the rotational speed and load of the rotating machine (10) are generally in a proportional relationship, and there may be a certain amplitude error value at the same main frequency.

[0051] The amplitude value of the main frequency may change depending on the degree of opening and closing of the valve for controlling the flow rate of the fluid flowing in the piping connected to the rotating machine (10). Therefore, it is necessary to infer the main frequency amplitude value according to the degree of opening and closing of the valve by adding valve opening and closing information (valve opening and closing sensing value) that can determine the degree of opening and closing of the valve (20). The data used to distinguish the rotational driving state of the rotating machine (10) as a transitional state or a normal state may include the main frequency amplitude of the n-second (n is the measurement time interval) signal, the main frequency of the n-second signal, and valve opening and closing information that can determine the valve opening and closing.

[0052] Since a problem arises where the frequency resolution is lower when performing a Fourier transform as the measurement time interval n becomes smaller, it is desirable to set n to a value greater than or equal to 1. If the measurement time interval n is too large, the probability of a transition state occurring within the signal related to the rotating machine (10) increases, so it is desirable to set the value of n according to the characteristics of the rotating machine (10).

[0053] To classify the transition state and the normal state, previously acquired current data of a defect-free rotating machine (10) may be utilized. The main frequency, main frequency amplitude, and valve opening / closing degree values ​​of the current signal of the rotating machine (10) generated at time t are respectively f t , a t , v t It is defined as such. In an embodiment of the present invention, a transition state and a steady state can be classified by utilizing a vector representing the difference between three datasets generated at the current time t and three datasets generated at the past time t-1. Accordingly, it is necessary to distinguish between a transition state signal and a fault signal.

[0054] FIG. 3 is a configuration diagram of a state classification unit constituting a rotary machine diagnostic device based on transition state and normal state classification according to an embodiment of the present invention. FIG. 4 is a flowchart showing step S20 of FIG. 2 in more detail. Referring to FIG. 1 to FIG. 4, the state classification unit (130) may include a main frequency calculation unit (131), a frequency difference element calculation unit (132), a valve opening / closing information calculation unit (133), a valve difference element calculation unit (134), and a rotational drive state determination unit (135).

[0055] The main frequency calculation unit (131) calculates the first time point main frequency (f) from the signal of the rotating machine (10) at each of the first time point (t) and the second time point (t-1) prior to the first time point (t). t ) and the second time point main frequency (f t-1 ) can be calculated (step S21 of FIG. 4). The first time point (t) and the second time point (t-1) may be time intervals that can be set by user input, pre-set in the system, or dynamically set according to the driving conditions of the rotating machine to classify the rotational driving state of the rotating machine.

[0056] The main frequency calculation unit (131) can determine one or more main frequencies from the signal of the rotating machine (10) through the frequency spectrum obtained through the Fourier transform. At this time, the main frequency may be the frequency in which the largest peak occurs in the frequency spectrum. Alternatively, the main frequency may be two or more peak frequencies determined in order of peaks in the frequency spectrum.

[0057] Fourier transform methods may include, for example, the Fast Fourier Transform (FFT), Continuous Fourier Transform (CFT), Discrete Fourier Transform (DFT), and Wavelet Transform, but are not necessarily limited to the exemplified Fourier transform methods.

[0058] The frequency difference element calculation unit (132) calculates the frequency difference element (f) based on the main frequency calculated by the main frequency calculation unit (131). t - f t-1 ) can be calculated (step S22 of FIG. 4). The frequency difference component calculation unit (132) calculates the first time point main frequency (f) determined from the signal of the rotating machine (10) at the first time point (t). t ) and the second time point main frequency (f) determined from the signal of the rotating machine (10) at the second time point (t-1) prior to the first time point (t). t-1 Calculate the difference value between ) and the frequency difference component (f t - f t-1 ) can be produced.

[0059] The valve opening / closing information calculation unit (133) receives a valve opening / closing value from the valve opening / closing measurement sensor (120), calculates valve opening / closing information related to the signal of the rotating machine (10), and provides it to the valve difference element calculation unit (134). The valve opening / closing information calculation unit (133) extracts a valve opening / closing value corresponding to the time points (t, t-1) when the signal of the rotating machine (10) is acquired, and from this, valve opening / closing information (v t , v t-1 ) can be calculated (step S23 of FIG. 4).

[0060] The valve opening / closing information calculation unit (133) receives valve opening / closing values ​​from the valve opening / closing measurement sensor (120), and among them, the first time point main frequency (f tDetermining the valve opening / closing value at the first time point (t) corresponding to ) and the first time point valve opening / closing information (v t ) can be calculated. In addition, the valve opening / closing information calculation unit (133) can calculate the second time point main frequency (f) among the valve opening / closing values ​​received from the valve opening / closing measurement sensor (120). t-1 Determine the valve opening / closing value at the second time point (t-1) corresponding to ) and the second time point valve opening / closing information (v t-1 ) can be produced.

[0061] The valve difference element calculation unit (134) can calculate the valve opening / closing difference element based on the valve opening / closing information calculated by the valve opening / closing information calculation unit (133) (step S24 of FIG. 4). The valve difference element calculation unit (134) calculates the first time point valve opening / closing information (v) determined from the valve opening / closing value at the first time point (t). t ) and second time point valve opening / closing information (v) determined from the valve opening / closing value at the second time point (t-1) prior to the first time point (t). t-1 Calculate the difference value between ) and the valve opening / closing difference element (v t - v t-1 ) can be produced.

[0062] The rotational drive state determination unit (135) is the frequency difference element (f) calculated by the frequency difference element calculation unit (132). t - f t-1 ) and the valve opening / closing difference element (v) calculated by the valve difference element calculation unit (134) t - v t-1 The difference vector x at the first time point (t) including ) t = [(f t - f t-1 ), (v t - v t-1 Based on )], the rotational driving state of the rotor (10) at the first time point (t) can be determined (step S25 of FIG. 4).

[0063] The difference vector used to determine the rotational driving state of the rotating machine (10) in the rotational driving state determination unit (135) is a frequency difference element (f) representing the change in the main frequency over time of the signal related to the rotating machine (10). t - f t-1 ) and a valve opening / closing difference element (v) representing the change in the opening / closing state of the valve over time related to the rotating machine (10) t - v t-1 It may include ).

[0064] The rotational drive state determination unit (135) can classify the rotational drive state of the rotating machine (10) into a normal state or a transitional state by a state classification model learned by the learning unit (150) based on Deep Support Vector Data Description (Deep SVDD), One-Class Support Vector Machine (One-Class SVM), etc.

[0065] At this time, for state classification models such as Deep Support Vector Data Description (Deep SVDD) or One-Class Support Vector Machine (One-Class SVM), the previously explained frequency difference component (f t - f t-1 ) and valve opening / closing difference element (v t - v t-1 A difference vector including ) is input as an input variable, and the rotational driving state of the rotor (10) can be output through a state classification model.

[0066] When a failure occurs in the rotating machine (10), a significant change occurs in the main frequency amplitude value of the current signal. If the main frequency amplitude among the three difference vectors is utilized, there is a risk that even if a failure occurs in the rotating machine (10), it may be mistakenly perceived as a transition signal and the failure-related process may be bypassed. Therefore, the difference vector used to distinguish between the transition state and the normal state consists of two types: the main frequency of the signal and a sensing value capable of determining the degree of valve opening and closing, which can be expressed as shown in Equation 1 below.

[0067] [Formula 1]

[0068]

[0069] The difference vector x shown in Equation 1 t In, f t represents the frequency of the current signal applied to the rotating machine at time t (the first time point), and v t represents the degree of valve opening or closing at time t. Also, in Equation 1, f t-1 represents the frequency of the current signal applied to the rotating machine at time t-1 (second time point) prior to time t (first time point), and v t-1 represents the degree of valve opening / closing at time t-1 (second time point) prior to time t (first time point).

[0070] Difference vector x t If both components of the difference vector are 0, that is, if the difference vector is [0, 0], the rotating machine is determined to be in a steady state (when the frequency difference component and the valve opening / closing difference component, which are components of the difference vector, are both 0 or within the allowable error range from 0), otherwise, the rotating machine can be determined to be in a transitional state (when at least one of the frequency difference component and the valve opening / closing difference component, which are components of the difference vector, is not 0 or is outside the allowable error range from 0).

[0071] By operating a rotor (10) that is free of defects while keeping the VFD fixed at a constant speed, the data set Data set X is acquired. When data is acquired in a steady state, the main frequency of the current signal and the difference factor of the valve opening / closing degree both have values ​​close to zero. In contrast, in the case of a transitional state, the main frequency of the current signal or the difference factor of the valve opening / closing degree changes over time, and therefore, when data is acquired in a transitional state, both factors of data set X have values ​​that move away from zero.

[0072] The learning unit (150) can train the state classification model of the state classification unit (130) by a semi-supervised learning model such as Deep Support Vector Data Description (Deep SVDD) and / or One-Class Support Vector Machine (One-Class SVM) on a normal data set created by sufficiently operating the normal rotation machine (10).

[0073] The Deep Support Vector Data Explanation (Deep SVDD) learning model is a method that solves the problem of anomaly detection in high-dimensional data by combining deep neural networks and data explanation methods. It can be used to identify abnormal data through a model trained on normal data, and can effectively detect outliers by learning the characteristics of high-dimensional data and mapping them to a low-dimensional space.

[0074] Deep Support Vector Data Description (Deep SVDD) learning models use artificial neural network structures, such as Convolutional Neural Networks (CNNs) or Deep Neural Networks (DNNs), to map input data into a feature space. They train the neural network models to ensure that the feature vectors of the training data converge to a center and to minimize the loss function.

[0075] One Class Support Vector Machine (One Class SVM) is a machine learning algorithm that can be used to solve problems such as anomaly detection. When the training data consists entirely of a single class (normal data), it learns the data distribution of that class and, based on this, can determine whether a new data point is normal or abnormal.

[0076] A One-Class Support Vector Machine (One-Class SVM) learns the distribution of training data when all training data is of a single class (normal data) to determine whether a new data point belongs to the same distribution. It then trains the model by performing a non-linear transformation using a kernel function to map the data to a high-dimensional feature space, and subsequently finding a hyperplane or hypersphere that encloses the normal data in that high-dimensional space.

[0077] One-Class Support Vector Machines (One-Class SVMs) have the advantage of being able to be trained using only normal data and not requiring explicit labeling for abnormal data. Additionally, One-Class Support Vector Machines can effectively model non-linear data distributions using various kernel functions and have the advantage of operating effectively even with high-dimensional data.

[0078] When data other than the training data is applied to the training model (state classification model) trained by the learning unit (150), it becomes possible to check whether the data is in a normal state. If, when applying the data, the rotating machine (10) is classified as being in a transition state, the transition state signal is not suitable for performing an abnormality diagnosis, so it can return to the data acquisition process without proceeding to the abnormality diagnosis process.

[0079] Previously, abnormal diagnosis was performed by analyzing the rotational state of a rotating machine through changes in the three elements of the current signal related to the rotating machine: amplitude, frequency, and phase. However, when a fault signal occurs in the rotating machine, changes in amplitude and phase may occur, and there was a risk that the fault signal related to the rotating machine could be mistakenly identified as a transition state signal, thereby bypassing the abnormal diagnosis process.

[0080] However, according to an embodiment of the present invention, by acquiring a valve opening / closing value regarding the degree of valve opening / closing related to a rotating machine using a valve opening / closing sensor or a pressure sensor, and classifying the rotational driving state of the rotating machine using the acquired valve opening / closing value, the change in amplitude between the transition state signal and the fault signal of the rotating machine can be distinguished, thereby reducing the fault diagnosis error of the rotating machine.

[0081] In the case of a transitional state signal, all three elements of the AC signal (frequency, amplitude, and phase) change, and in the case of a fault signal, the amplitude and phase components also change, making it difficult to classify the transitional state and the normal state based on only these three elements. Accordingly, in the embodiment of the present invention, in addition to the main frequency information of the current signal, the valve opening / closing sensing value can be used to clearly distinguish whether the amplitude change is due to the transitional state or the amplitude change is due to the fault state.

[0082] On the other hand, if the main frequency of the current signal and the valve opening / closing sensing values ​​related to the rotating machine are used as is to classify transitional and steady states, the number of possible scenarios increases excessively, which may make data acquisition difficult. Therefore, transitional and steady states can be classified by limiting the range of possible values ​​using the difference value (difference vector) between the current signal and the previous signal of the rotating machine current signal.

[0083] In addition, in an embodiment of the present invention, a semi-supervised learning model based on normal data is utilized, and by applying selected characteristic factors regarding the signal of the rotating machine and the degree of valve opening / closing, along with the difference values ​​of the characteristic factors, to the semi-supervised learning model, objective diagnosis of the rotating machine is possible through the classification of the rotating machine's transition state and steady state.

[0084] Although a vast amount of data must be acquired to diagnose a rotating machine without distinguishing between a transition state and a steady state, there are limitations to securing a large amount of data during the early stages of rotating machine development. Therefore, according to the embodiment of the present invention, the technology for diagnosing a rotating machine by distinguishing between a transition state and a steady state enables efficient diagnosis of the rotating machine by distinguishing between the transition state and the steady state even with the acquisition of a small amount of short-term data.

[0085] Hereinafter, an embodiment of the present invention is described using the process of acquiring current data of a ship motor to diagnose abnormalities in the ship motor as an example. FIG. 5 is a flowchart illustrating the process of acquiring current data of a ship motor to diagnose abnormalities in the ship motor. Once the current data of the ship motor is acquired (step S11 of FIG. 5), the current data can be segmented (S21), and after applying a Hanning window, a Fast Fourier Transform (FFT) can be performed for each segment (steps S22, S23 of FIG. 5).

[0086] In step S22, the Hanning window is a window function used in signal processing that can be applied to reduce spectral leakage caused by signal discontinuities during Fast Fourier Transform (FFT) analysis. By applying the Hanning window, frequency resolution is improved by smoothly attenuating the ends of the signal and mitigating abrupt changes, and frequency domain analysis can be performed more accurately by reducing spectral leakage. Meanwhile, in addition to the Hanning window, other windows such as the Hamming window, which can reduce spectral leakage, may also be applied.

[0087] Next, changes in the main frequency or amplitude are analyzed based on the Fourier transform data (step S24 in FIG. 5), and if there are changes in the main frequency or amplitude, it is classified as a transition state (step S25 in FIG. 5). In this case, the abnormal diagnosis of the ship motor can be performed without performing the abnormal diagnosis, and the process described above can be repeated by returning to the step of acquiring current data of the ship motor. If there are no changes in the main frequency or amplitude in the Fourier transform data (step S24 in FIG. 5), it is determined that the ship motor is in a normal state (step S26 in FIG. 5), and in this case, the abnormal diagnosis of the ship motor can be performed (step S31 in FIG. 5).

[0088] Even if a rotating machine (ship motor) rotates at a constant speed, the load may vary depending on the incoming flow rate, so a perfectly proportional relationship between the current frequency and amplitude may not be established. However, generally, as the rotational speed of the rotating machine increases, the incoming flow rate increases, and consequently, the load also increases. Therefore, the applied current frequency and amplitude of pumps used in ships can generally be said to be in a proportional relationship.

[0089] Figure 6 is an example diagram showing the change in current frequency according to the transition state and steady state. Figure 7 is an example diagram showing the change in current RMS according to the transition state and steady state. In the graph comparing the current frequency and amplitude of an actual ship LNG supply pump, the change in current signal amplitude can be influenced by various factors other than the motor load, making it difficult to determine whether the motor load is in a transition state or a steady state based on the amplitude change. Upon checking the actual current signal, it can be confirmed that there is a slight change in amplitude even when there is no change in the rotational RPM of the rotating machine. This makes it difficult to determine whether such a change in the amplitude of the applied current signal is caused by the motor transition state or by an actual equipment malfunction.

[0090] In an embodiment of the present invention, attention is paid to the relationship between the motor RPM and amplitude of a rotating machine. If the degree of valve opening / closing is known, it is substituted with the motor load to enable more accurate determination of the transition state and steady state. When the degree of valve opening / closing is constant, the flow rate sucked in according to the motor RPM is also constant, allowing for more accurate verification of the motor load. If the motor load is defined as the degree of valve opening / closing, it becomes possible to distinguish whether a change in the amplitude of the current signal is caused by a change in the motor load or by a malfunction of the rotating machine equipment.

[0091] As described above, the rotary machine diagnostic device and method based on transition state and normal state classification according to an embodiment of the present invention classify the transition state and the normal state by a preprocessing algorithm for diagnosing abnormalities in the rotary machine, and at this time, for classifying the transition state and the normal state, the main frequency of the current signal applied to the rotary machine and the sensing value through a sensor capable of confirming the opening and closing of the valve are utilized as characteristic factors.

[0092] In addition, the rotating machine diagnostic device and method based on transition state and steady state classification according to an embodiment of the present invention can accurately classify transition states and steady states by applying characteristic factors corresponding to the frequency difference value between the main frequency of the current signal and the main frequency of the previous signal, and the valve opening / closing difference value between the degree of valve opening / closing at the current time and the valve opening / closing information at the previous time, to a semi-supervised learning model.

[0093] The rotary machine diagnostic device and method based on transition state and normal state classification according to the embodiment of the present invention as described above can effectively perform transition state / normal state comparison, which causes significant difficulties in diagnosing rotary machines, particularly variable speed rotary machines such as variable speed pumps / compressors, and has the effect of reducing the risk of skipping the abnormal diagnosis process by incorrectly judging a fault signal as a transition state signal due to a change in amplitude / phase when a fault signal occurs.

[0094] At least some of the configurations of the embodiments described above may be implemented as hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an Arithmetic Logic Unit (ALU), a Digital Signal Processor, a microcomputer, a Field Programmable Gate Array (FPGA), a Programmable Logic Unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions.

[0095] The processing unit may execute an operating system and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of understanding, the processing unit may be described as being used as a single unit, but a person of ordinary skill in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements.

[0096] For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as a parallel processor, are also possible. Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure the processing unit to operate as desired or instruct the processing unit independently or collectively.

[0097] Software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by a processing device or to provide instructions or data to a processing device. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0098] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software.

[0099] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiments, and vice versa.

[0100] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without altering its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive, and the scope of the present invention is defined by the claims set forth below. Furthermore, all modifications or variations derived from the meaning and scope of the claims and their equivalents should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0101] 100: Rotating machine diagnostic device based on transitional and steady-state classification 110: Signal acquisition unit 120: Valve opening / closing measurement sensor 130 : Status Classification Section 131 : Main frequency calculation unit 132 : Frequency difference element calculation unit 133 : Valve Open / Closed Information Calculation Unit 134 : Valve difference element calculation section 135: Rotational drive state determination unit 140: Rotating machine diagnostic control unit 150 : Learning Department

Claims

Claim 1 A signal collection unit for collecting signals related to a rotating machine; a state classification unit for classifying the rotational driving state of the rotating machine into a transitional state and a normal state based on signal information related to the signal and valve opening / closing information related to the opening / closing state of the valve related to the rotating machine; and a valve opening / closing measurement sensor for detecting the opening / closing state of the valve. The system includes a rotary machine diagnostic control unit that determines whether to perform an abnormality diagnosis of the rotary machine according to the rotational driving state classified for the rotary machine by the state classification unit, wherein the transition state is a state in which the rotational speed of the rotary machine changes, and the normal state is a state in which the rotational speed of the rotary machine is maintained at a constant, and the signal information related to the signal includes the main frequency of the signal, and the state classification unit comprises: a main frequency calculation unit that calculates a main frequency at a first time point and a main frequency at a second time point corresponding to a peak from a frequency spectrum obtained by Fourier transforming the signal of the rotary machine at each of a first time point and a second time point prior to the first time point; a frequency difference component calculation unit that calculates a frequency difference component through a difference operation between the main frequency at the first time point and the main frequency at the second time point calculated from the signal of the rotary machine; a valve opening / closing information calculation unit that calculates valve opening / closing information at the first time point and valve opening / closing information at the second time point by extracting valve opening / closing values ​​corresponding to the first time point and the second time point in which the signal of the rotary machine is obtained, respectively, among valve opening / closing values ​​collected over time; and the first time point calculated by the valve opening / closing information calculation unit. A rotary machine diagnostic device based on transition state and steady state classification, comprising: a valve difference element calculation unit that calculates a valve opening / closing difference element by calculating a difference value between valve opening / closing information and the valve opening / closing information at the second time point; and a rotary driving state determination unit that determines the rotary driving state of the rotary machine at the first time point based on a difference vector including the frequency difference element and the valve opening / closing difference element. Claim 2 delete Claim 3 delete Claim 4 A rotary machine diagnostic device based on transition state and steady state classification according to claim 1, wherein the state classification unit is learned to classify the transition state and the steady state based on a difference vector comprising a frequency difference element representing a time-series change of the main frequency of the signal and a valve opening / closing difference element representing a time-series change of the opening / closing state of the valve. Claim 5 delete Claim 6 A rotary machine diagnostic device based on transition state and steady state classification according to claim 1, wherein the signal comprises at least one of an electrical signal applied to the rotary machine, an electrical signal generated from the rotary machine, and a vibration signal of the rotary machine detected by a vibration sensor. Claim 7 A rotary machine diagnostic device based on transition state and steady state classification according to claim 1, wherein the rotary machine is a variable speed rotary machine, the transition state is a state in which the rotational speed of the variable speed rotary machine changes, and the steady state is a state in which the rotational speed of the variable speed rotary machine stably sucks in fluid and the rotational speed of the variable speed rotary machine is maintained at a constant. Claim 8 delete Claim 9 A step of collecting a signal related to a rotating machine by a signal collection unit; a step of classifying the rotational driving state of the rotating machine into a transitional state and a normal state by a state classification unit based on signal information related to the signal and valve opening / closing information related to the opening / closing state of the valve related to the rotating machine; a step of detecting the opening / closing state of the valve by a valve opening / closing measurement sensor to classify the normal state and the transitional state; The method includes a step of determining whether to perform an abnormality diagnosis of the rotating machine according to the rotational driving state classified for the rotating machine by the state classification unit by the rotating machine diagnostic control unit, wherein the transition state is a state in which the rotational speed of the rotating machine changes, and the normal state is a state in which the rotational speed of the rotating machine is maintained at a constant, and the signal information related to the signal includes the main frequency of the signal, and the classification step comprises: a step of calculating a first time point main frequency and a second time point main frequency corresponding to a peak from a frequency spectrum obtained by Fourier transforming the signal of the rotating machine at each of the first time point and the second time point prior to the first time point; a step of calculating a frequency difference component representing the frequency change of the rotating machine by performing a difference operation between the first time point main frequency and the second time point main frequency calculated from the signal of the rotating machine; a step of calculating first time point valve opening / closing information and second time point valve opening / closing information by extracting valve opening / closing values ​​corresponding to the first time point and the second time point in which the signal of the rotating machine is obtained, respectively, among the valve opening / closing values ​​collected over time; and the first time point valve corresponding to the valve opening / closing value at the first time point. A step of calculating a valve opening / closing difference element by calculating a difference value between the opening / closing information and the valve opening / closing information at the second time point corresponding to the valve opening / closing value at the second time point;A rotary machine diagnosis method based on transition state and steady state classification, comprising the step of determining the rotational driving state of the rotary machine at the first time point based on a difference vector including the frequency difference element and the valve opening / closing difference element representing the main frequency and valve opening / closing change between the first time point and the second time point. Claim 10 delete Claim 11 delete Claim 12 A rotating machine diagnosis method based on transition state and steady state classification according to claim 9, further comprising the step of learning a state classification model of the state classification unit to classify the transition state and the steady state based on a difference vector including a frequency difference element representing a time-series change of the main frequency of the signal and a valve opening / closing difference element representing a time-series change of the opening / closing state of the valve by the learning unit. Claim 13 A rotating machine diagnosis method based on transition state and steady state classification according to claim 12, wherein the state classification model is learned by a semi-supervised learning model based on at least one of Deep Support Vector Data Description and One-Class Support Vector Machine. Claim 14 delete Claim 15 A rotating machine diagnosis method based on transition state and steady state classification according to claim 9, wherein the signal comprises at least one of an electrical signal applied to the rotating machine, an electrical signal generated from the rotating machine, and a vibration signal of the rotating machine detected by a vibration sensor. Claim 16 A rotary machine diagnosis method based on transition state and steady state classification according to claim 9, wherein the rotary machine is a variable speed rotary machine, the transition state is a state in which the rotational speed of the variable speed rotary machine changes, and the steady state is a state in which the rotational speed of the variable speed rotary machine stably sucks in fluid and the rotational speed of the variable speed rotary machine is maintained at a constant. Claim 17 delete Claim 18 A computer-readable recording medium having a computer program recorded thereon for executing a rotating machine diagnostic method based on transition state and steady state classification according to any one of claims 9, 12, 13, and 15 to 16.

Citation Information

Patent Citations

  • Charging equipment monitoring method and system

    CN115792607A

  • Method and apparatus for limiting torque in an electric drive motor

    KR1020110066228A

  • Machine Diagnosis and Prediction System using Machine Learning

    KR1020200101507A