Method, system and computer program for detecting faults in an induction motor

Abstract

A method, system and computer program for detecting faults in an induction motor using electrical signal analysis (ESA) and fast Fourier transform (FFT) techniques are disclosed. The method comprises receiving an input signal (10) from an induction motor; transforming the input signal into the frequency domain using an FFT (11) to obtain an FFT spectrum; applying an envelope detection algorithm (12-15) to the FFT spectrum to obtain an envelope representing amplitude variations of frequency components; calculating (16) a root mean square (RMS) of the envelope; and analyzing the calculated RMS envelope (17) to detect at least one frequency signature indicative of a fault in the induction motor. The method enhances the fault detection capability by addressing the noise sensitivity and resolution limitations of conventional FFT-based ESA techniques.

Description

Technical Field

[0001] This disclosure relates to the field of predictive maintenance technology, and more specifically to fault detection in induction motors using electrical signal analysis (ESA) and fast Fourier transform (FFT) techniques. Background Technology

[0002] Predictive maintenance technologies involve using advanced monitoring and diagnostic tools to predict equipment failures before they occur. These technologies aim to reduce downtime, improve operational efficiency, and enhance workplace safety by providing real-time data on the condition of machinery. Predictive maintenance relies on a variety of technologies, including vibration analysis, thermal imaging, and electrical signal analysis, to monitor the health of equipment and identify potential problems early.

[0003] Predictive maintenance technology aims to ensure the continuous and efficient operation of industrial equipment. By monitoring the condition of machinery, facility managers can schedule maintenance activities based on the actual needs of the equipment rather than a fixed schedule. This approach helps prevent unexpected failures, reduces maintenance costs, and extends the lifespan of the equipment. In the context of induction motors, predictive maintenance focuses on detecting common faults such as misalignment and broken rotor bars, which, if not addressed promptly, can lead to serious operational disruptions.

[0004] An induction motor is an electric motor that operates based on the principle of electromagnetic induction, where the current required to generate torque is produced through the electromagnetic induction of a magnetic field from the stator windings. It typically consists of a stator and a rotor; the stator generates a rotating magnetic field, which induces motion in the rotor. Induction motors are widely used in industrial and commercial applications due to their robustness, reliability, and simple structure.

[0005] Induction motors are widely used in a variety of industrial, commercial, and residential applications due to their robustness, reliability, and simplicity of construction. They have industrial applications in areas such as pumps, conveyors, compressors, machine tools, and fans and blowers. They can also be used to drive elevators and escalators, or in electric vehicles, etc.

[0006] Electrical signal analysis (ESA) is a widely used technique in predictive maintenance of induction motors. ESA involves analyzing the motor's current and voltage signals in the frequency domain using a Fast Fourier Transform (FFT). By examining the frequency components of these signals, characteristic fault features indicating the presence of specific defects can be identified. Until now, this approach has always been based on the analysis of the amplitude of the fault's frequency components, providing valuable insights into the motor's condition and aiding in the early detection of faults.

[0007] While ESA and FFT are effective in fault detection, these methods face several challenges that may limit their accuracy and reliability. Notably, although several studies have been conducted using FFT, they all face difficulties in identifying the characteristic frequency components of faults.

[0008] In "Fault Detection of Induction Motors Using Feature Selection via Principal Component Analysis with Fast Fourier Transform," International Journal of Precision Engineering & Manufacturing, Vol. 20, No. 9, pp. 1543-1553, 2019, Young-Jun Yoo proposes a method that involves calculating the FFT signal of induction motor current data and using Principal Component Analysis (PCA) to find the probabilistic bounds of the FFT peak frequency and corresponding value variations through Hotelling's T2 value. This allows for the detection of faults in induction motors by reducing the dimensionality of the fault characteristic frequency component (FCFC) using PCA. Hotelling's T2 can be used as an indicator for detecting faults such as broken rods or broken end ring connector conditions.

[0009] While interesting, this method has the same drawbacks as other existing technical solutions based on ESA via FFT.

[0010] In practice, noise sensitivity is a significant issue because FFT can be highly sensitive to noise, making it difficult to distinguish fault signals from background noise. Furthermore, the resolution of FFT is constrained by the length of the analyzed signal, which can lead to insufficient frequency resolution for accurate fault identification. Shorter signals may not provide sufficient frequency resolution to accurately identify faults. These limitations necessitate the development of improved techniques to enhance fault detection in induction motors.

[0011] Therefore, a method is needed to enhance the fault detection capability of ESA by addressing the noise sensitivity and resolution limitations of FFT. Summary of the Invention

[0012] This disclosure improves upon this situation.

[0013] A method for detecting faults in induction motors is proposed, comprising:

[0014] Receive input signals from the induction motor;

[0015] The input signal is transformed to the frequency domain using the Fast Fourier Transform (FFT) to obtain the FFT spectrum.

[0016] The envelope detection algorithm is applied to the FFT spectrum to obtain the envelope representing the amplitude changes of the frequency components of the FFT spectrum;

[0017] Calculate the root mean square (RMS) of the envelope;

[0018] The calculated RMS envelope is analyzed to detect at least one frequency characteristic representing a fault in an induction motor, referred to as a fault characteristic.

[0019] On the other hand, a system for detecting faults in an induction motor is proposed, comprising:

[0020] The signal acquisition module is configured to receive input signals from the induction motor;

[0021] The processing unit is configured as follows:

[0022] The input signal is transformed to the frequency domain using the Fast Fourier Transform (FFT) to obtain the FFT spectrum.

[0023] The envelope detection algorithm is applied to the FFT spectrum to obtain the envelope representing the amplitude changes of the frequency components of the FFT spectrum;

[0024] Calculate the root mean square (RMS) of the envelope;

[0025] The calculated RMS envelope is analyzed to detect at least one frequency characteristic representing a fault in an induction motor, referred to as a fault characteristic.

[0026] In another aspect, computer software is proposed, including instructions that, when executed by a processor, implement at least a portion of the methods as defined herein. In yet another aspect, a computer-readable non-transitory recording medium is proposed on which software is registered, and when executed by a processor, to implement the methods as defined herein.

[0027] The following features can be implemented, either individually or in combination:

[0028] Applying envelope detection algorithms to FFT spectra includes:

[0029] - Rectify the FFT spectrum;

[0030] - Smooth the rectified FFT spectrum with a low-pass filter to obtain the envelope.

[0031] The detected faults belong to the following groups:

[0032] -Misalignment fault;

[0033] - Broken rotor bar failure.

[0034] The calculated RMS envelope includes:

[0035] - The total energy of at least one fault feature is calculated by summing the harmonic fault areas between the curve representing the RMS envelope and the x-axis.

[0036] - Output a fault indicator representing the calculated total energy;

[0037] - Compare the fault indicator with a threshold to detect a fault when the fault indicator exceeds the threshold.

[0038] The input signal belongs to the following group:

[0039] - Voltage signal;

[0040] - Current signal;

[0041] -Reactive power signal;

[0042] - Active power signal.

[0043] The FFT spectrum is normalized before the envelope detection algorithm is applied.

[0044] The processing unit is also configured to calculate the total energy of at least one fault feature by summing the harmonic fault area between the curve representing the RMS envelope and the x-axis, and the system also includes a fault indicator module configured to output a fault indicator representing the calculated total energy and compare the fault indicator with a threshold to detect a fault when the fault indicator exceeds the threshold.

[0045] The system for detecting faults also includes a user interface configured to display fault detection results to the operator. Attached Figure Description

[0046] Other features, details, and advantages will be shown in the following detailed description and accompanying drawings, wherein:

[0047] Figure 1 This is a block diagram illustrating the calculation of the RMS envelope of the FFT spectrum of the input signal according to an embodiment.

[0048] Figure 2 A comparison of the FFT and envelope of the current signal of a faulty motor with misalignment faults according to an embodiment is shown.

[0049] Figure 3 A comparison of the FFT envelopes of a healthy electric motor and an electric motor with misalignment faults according to an embodiment is shown.

[0050] Figure 4 A comparison of the FFT and envelope of the current signal of a faulty motor with a broken rotor bar fault, according to an embodiment, is shown.

[0051] Figure 5 A comparison of the FFT envelopes of a healthy electric motor and an electric motor with a broken rotor bar fault, according to an embodiment, is shown.

[0052] Figure 6A set of fault indicators is shown based on the envelope energy of the reactive power signal of a 4kW induction motor with three broken rotor bars, according to an embodiment.

[0053] Figure 7 A set of fault indicators based on the envelope energy of the current signal of a 4kW induction motor with three broken rotor bars, according to an embodiment, is shown.

[0054] Figure 8 A set of fault indicators based on the envelope energy of the voltage signal of a 4kW induction motor with three broken rotor bars, according to an embodiment, is shown.

[0055] Figure 9 This is a block diagram of a system for detecting faults according to an embodiment. Detailed Implementation

[0056] The general principle of this disclosure is to enhance fault detection capabilities in induction motors by utilizing a combination of Electrical Signal Analysis (ESA) and Fast Fourier Transform (FFT) techniques, supplemented by an envelope detection algorithm. Envelope detection is a signal processing technique used to extract the envelope of a signal. The envelope represents a smoothed curve that outlines the peaks of an oscillating signal. This method involves transforming the input signal from the induction motor to the frequency domain using FFT to obtain an FFT spectrum. The envelope detection algorithm is then applied to this spectrum, which may include rectifying and smoothing the FFT spectrum to obtain an envelope representing the amplitude variations of the frequency components. By calculating the root mean square (RMS) of this envelope, the method effectively reduces noise and highlights the dominant frequency components. This enhanced frequency analysis allows for accurate detection of fault characteristics, such as misalignment and rotor bar breakage, by emphasizing specific frequency components and providing a clearer and more interpretable spectrum.

[0057] In fact, each fault has a specific frequency characteristic, which is a specific pattern or set of frequencies that appear in the spectrum of electrical signals when a particular fault occurs in an induction motor. The fault frequency characteristic can be regarded as a "fingerprint" that helps identify the type of fault affecting the motor.

[0058] This disclosure allows for the detection of specific frequency characteristics in the calculated RMS envelope that indicate faults in an induction motor, and for monitoring changes in the amplitude of these detected fault characteristics that indicate how the fault develops over time.

[0059] Now for reference Figure 1 It shows a block diagram for calculating the RMS envelope of an input signal according to an embodiment of a method for detecting faults in an induction motor.

[0060] Input signal 10 represents an electrical signal acquired from the induction motor. This signal can be any electrical signal acquired from the motor, such as a voltage signal, a current signal, or an active or reactive power signal. The signal acquisition module (described below in the section on the system for fault detection) Figure 9 The input signal (marked 101) is captured, and this input signal contains information about the operating state of the motor. Input signal 10 serves as the initial data source for subsequent processing steps, where the signal undergoes transformation and analysis to detect potential faults in the motor. This input signal 10 can be sampled at any frequency sampling rate (e.g., 2kHz, 4kHz, 6kHz…) without significantly affecting the efficiency of fault detection, such as when considering information about… Figures 6 to 8 As shown.

[0061] Unit 11, used to calculate the Fast Fourier Transform (FFT), transforms the input signal 10 from the time domain to the frequency domain. This transformation produces an FFT spectrum that displays the frequency components of the input signal, which is then rectified by unit 12. Rectification involves converting all negative values ​​in the FFT spectrum to positive values, ensuring the entire spectrum is positive. This can be done using full-wave rectification or half-wave rectification, in which both the positive and negative halves of the signal are converted to positive, and in half-wave rectification, only the negative half is converted. This step prepares the signal for envelope detection because the rectified FFT spectrum simplifies subsequent smoothing. The rectified FFT spectrum highlights the amplitude variations of the frequency components, which is necessary for fault detection.

[0062] Gain 13 amplifies the rectified FFT spectrum by a factor of two. This amplification enhances the signal amplitude, making the dominant frequency component easier to detect and analyze. Gain adjustment ensures that the signal's salient features are more pronounced, thus facilitating the identification of fault characteristics in induction motors.

[0063] Downsampling component 14 reduces the sampling rate of the amplified and rectified FFT spectrum. Downsampling involves selecting a subset of data points, thereby effectively reducing the number of samples while preserving the characteristics of the signal. This step helps manage data size and computational load, making subsequent processing steps more efficient. The downsampling factor can vary depending on the type of fault being analyzed, with different factors applied to misalignment and broken rotor bar faults (e.g., downsampling factor 15 provides good results for misalignment faults, while downsampling factor 5 can be used for broken rotor bar faults, characterized by a greater number of sidebands in the fault features).

[0064] Then, low-pass filter 15 smooths the downsampled FFT spectrum by removing high-frequency noise components that may be introduced by downsampling. This filtering process produces a smooth curve representing the envelope of the FFT spectrum. Low-pass filter 15 ensures that only relevant low-frequency components corresponding to fault characteristics are retained. The cutoff frequency of the low-pass filter is selected based on the specific requirements of the fault detection process (e.g., 10 rad.s⁻¹).

[0065] Component 16 then calculates the root mean square (RMS) of the smoothed envelope by defining a window size. Recall that the RMS is a statistical measure of the amplitude of a changing signal, providing an average that accounts for both positive and negative variations. Therefore, the RMS calculation provides a measurement of the amplitude of the changing envelope, effectively reducing noise and highlighting the dominant frequency components. This step enhances the clarity of the spectrum and makes fault characteristics easier to identify. The RMS value represents the total energy of the envelope, which is used in subsequent analysis to detect faults.

[0066] The final output of the envelope detection algorithm is Envelope17, which summarizes the amplitude variations of the frequency components in the FFT spectrum, providing a clear and interpretable representation of the signal's frequency content. It can also be done using Matlab. ® The envelope is derived using the RMS envelope function in [the original text].

[0067] The envelope 17 is then analyzed to detect specific frequency characteristics indicating faults in the induction motor. By examining the envelope, the system can identify and monitor the evolution of faults such as misalignment and broken rotor bars, thereby ensuring timely maintenance and preventing operational interruptions.

[0068] Figure 1 The method illustrated for fault detection is particularly advantageous in applications such as mechanical diagnostics, where it enhances fault detection by emphasizing specific frequency components while providing a clearer and more interpretable spectrum. It offers a robust solution for fault detection in induction motors by overcoming the noise sensitivity and resolution limitations of traditional FFT-based ESA techniques. The combination of FFT, envelope detection, and RMS calculation provides a clearer and more detailed representation of the signal's frequency content, thereby improving fault detection and diagnosis.

[0069] In practice, fault detection in induction motors is crucial for maintaining operational efficiency and preventing unexpected failures. Two common faults are misalignment and rotor bar breakage, each producing different characteristics in the motor's current spectrum.

[0070] Now about Figure 2 and Figure 3 Describe the application of detecting misalignment faults in induction motors.

[0071] Misalignment faults occur in induction motors when the rotor and stator are not properly aligned. This misalignment results in an uneven air gap between the rotor and stator, which affects the motor's magnetic field and reduces its efficiency, increases vibration and noise levels, and adds additional mechanical stress to motor components. Misalignment faults generate specific harmonic components in the motor's current spectrum, which can be detected and analyzed to identify the fault.

[0072] The frequency of misaligned fault feature markers can be expressed by the following equation:

[0073] [Mathematical Expression 1]

[0074]

[0075] Where f misalignment It is the frequency of the misalignment fault characteristic, f s It is the power supply frequency, f m It is the mechanical rotor frequency, and k is an integer representing the harmonic order. Recall that harmonic components are frequencies that are integer multiples of the dominant frequency, and they typically appear in the signal spectrum due to periodic interference.

[0076] Figure 2 The figure shows a comparison between the normalized Fast Fourier Transform (FFT) spectrum of the current signal of a faulty motor with misalignment and its corresponding normalized FFT envelope. This figure illustrates how the envelope detection algorithm enhances the visual representation of the fault detection process by highlighting the dominant frequency components.

[0077] In fact, according to the embodiments, it can be applied Figure 1 The envelope detection algorithm normalizes the FFT spectrum before applying it. Normalizing the FFT spectrum before applying the envelope detection algorithm ensures that the amplitude values ​​of the frequency components are scaled to a common reference. This normalization process enhances signal clarity by reducing the impact of signal amplitude variations, making fault characteristics easier to identify.

[0078] Normalization also helps mitigate the impact of noise in the FFT spectrum. By scaling the amplitude values, this method reduces the influence of random noise fluctuations, making the dominant frequency components more apparent and easier to analyze. The normalization process also makes the method more robust to variations in the input signal. Variations in signal amplitude due to load changes, power supply voltage fluctuations, or other factors are considered during normalization, ensuring that the fault detection process remains reliable under changing conditions.

[0079] exist Figure 2In the diagram, the horizontal axis or X-axis represents the frequency components of the signal, ranging from 0 to 100 Hz. The vertical axis or Y-axis represents the normalized amplitude values ​​of the frequency components. The solid line represents the normalized FFT spectrum of the current signal. This spectrum shows the frequency components of the signal, including fault characteristic frequencies. The dashed line represents the normalized FFT envelope17 obtained after applying the envelope detection algorithm. This envelope17 outlines the amplitude variations of the frequency components, providing a clearer representation of the dominant frequency.

[0080] The main curve shows a prominent peak at the power supply frequency (approximately 50Hz) and at f s ±f m The sidebands at the location. These sidebands are characteristic of misalignment faults.

[0081] The illustration on the left shows the frequency components in A magnified view, where f s It is the power supply frequency, and This is the mechanical rotor frequency. The inset shows the normalized FFT (solid line) and normalized FFT envelope 17 (dashed line) for this specific frequency component. The inset on the right provides the frequency component in... An enlarged view. Similar to the inset on the left, it shows the normalized FFT (solid line) and normalized FFT envelope 17 (dashed line) for this frequency component.

[0082] Envelope 17 more clearly highlights amplitude changes, making fault characteristics more identifiable. In fact, if the signal resolution is reduced, the amplitude of the frequency peaks associated with the fault also decreases, making them difficult to detect when working on the FFT (solid line), as is done in the prior art, rather than when working on the RMS envelope of the FFT (dashed line).

[0083] As in Figure 2 It is also evident that the envelope detection algorithm reduces the impact of noise, making the main frequency components more obvious and easier to analyze.

[0084] Figure 3 The figure shows a comparison of the FFT envelopes 17 between a healthy motor and a motor with misalignment faults. This figure provides a visual representation of how the envelope detection algorithm distinguishes between the two conditions by highlighting fault features in the spectrum.

[0085] The horizontal axis, or X-axis, represents the frequency components of the signal, ranging from 0 to 100 Hz. The vertical axis, or Y-axis, represents the normalized amplitude values ​​of the frequency components. The solid line represents the normalized FFT envelope of the current signal of a healthy motor. This line is used as the baseline for comparison. The dashed line represents the normalized FFT envelope of the current signal of a motor with a misalignment fault. This line indicates the presence of fault characteristics in the spectrum.

[0086] The illustration on the left provides the frequency components in A magnified view, in which It is the power supply frequency, and This is the mechanical rotor frequency. The illustration shows the normalized FFT envelopes of both a healthy motor (solid line) and a motor with a misalignment fault (dashed line). The fault characteristics are clearly visible in the dashed lines, indicating the presence of a misalignment fault.

[0087] The illustration on the right provides the frequency components in An enlarged view. Similar to the inset on the left, it shows the normalized FFT envelopes of both a healthy motor (solid line) and a motor with a misalignment fault (dashed line). The fault characteristics are again visible in the dashed lines, confirming the presence of the misalignment fault.

[0088] The main curve shows the prominent peak at the power supply frequency (approximately 50Hz) and at... The sidebands at the location. These sidebands are characteristic of misalignment faults.

[0089] The dashed line (misalignment fault) indicates the characteristic frequency of the fault. and A significant amplitude change at the solid line (in a healthy motor) indicates the presence of a misalignment fault and constitutes a fault characteristic.

[0090] In general, Figure 3 This study demonstrates the effectiveness of the envelope detection algorithm in distinguishing healthy motors from those with misalignment faults. Clear visualization of fault characteristics in the spectrum facilitates accurate detection and diagnosis of misalignment faults in induction motors.

[0091] Regarding misalignment faults, focusing on the first harmonic or even just the first two harmonics may be sufficient, as subsequent harmonics have less significant values. It is worth noting that the second harmonic may superimpose with electrical harmonics. And it might be better to focus only on the first harmonic. For example, regarding... Figures 6 to 8 As shown, the indicator can be calculated as the sum of the total energy for the fault region of the selected harmonic.

[0092] Now combine Figures 4 to 8 This paper describes an application for detecting broken rotor bars in an induction motor, based on an example of a 4kW induction motor with three broken rotor bars.

[0093] A broken rotor bar fault occurs in an induction motor when one or more bars in the rotor cage break or are damaged. This fault disrupts the uniform magnetic field required for efficient motor operation, leading to performance degradation, increased vibration, and potential overheating. Broken rotor bar faults produce fault characteristics characterized by sidebands near the power supply frequency. The frequencies of these sidebands can be expressed by the following equation:

[0094] [Mathematical Expression 2]

[0095]

[0096] in, It is the frequency characteristic of a broken rotor bar fault. It is the power supply frequency. It is the slip of the electric motor, and It is an integer representing the harmonic order.

[0097] The slip in an induction motor refers to the difference between the synchronous speed (the speed of the rotating magnetic field in the stator) and the actual speed of the rotor. Slip is a critical parameter in the operation of an induction motor because it determines the motor's torque and efficiency. It is usually expressed as a percentage of the synchronous speed. It can be represented by the following equation:

[0098] [Mathematical Expression 3]

[0099]

[0100] in It is synchronous speed, and This refers to the rotor speed. Due to slip, the rotor speed is always slightly less than the synchronous speed. Synchronous speed Power frequency The number of poles in the electric motor is determined.

[0101] Figure 4 The figure shows a comparison between the normalized Fast Fourier Transform (FFT) spectrum of the current signal of a faulty motor with a broken rotor bar fault and its corresponding normalized FFT envelope. This figure illustrates how the envelope detection algorithm enhances the visual representation of the fault detection process by highlighting the dominant frequency component associated with the broken rotor bar fault.

[0102] The horizontal axis or X-axis represents the frequency components of the signal, ranging from 30 to 75 Hz. The vertical axis or Y-axis represents the normalized amplitude values ​​of the frequency components. The solid line represents the normalized FFT spectrum of the current signal. This spectrum shows the frequency components of the signal, including fault characteristic frequencies. The dashed line indicates the application... Figure 1 The normalized FFT envelope is obtained after the envelope detection algorithm.

[0103] The main curve plot shows the power supply frequency ( The prominent peak at ) and at The edge bands at the location. These edge bands are characteristic of broken rotor bar faults and are marked on the diagram:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] Figure 5 This figure shows a comparison of the FFT envelopes between a healthy motor and a motor with a broken rotor bar (BRB) fault. The graph illustrates how the envelope detection algorithm distinguishes the two conditions visually by highlighting fault features in the spectrum.

[0111] The horizontal axis, or X-axis, represents the frequency components of the signal, ranging from 0 to 100 Hz. The vertical axis, or Y-axis, represents the normalized amplitude values ​​of the frequency components. The solid line represents the normalized FFT envelope of the current signal of a healthy motor. This line is used as the baseline for comparison. The dashed line represents the normalized FFT envelope of the current signal of a motor with a broken rotor bar fault. This line indicates the presence of fault characteristics in the spectrum.

[0112] The main graph shows prominent peaks at the power supply frequency (approximately 50 Hz) and sidebands at specific frequencies. These sidebands are characteristic of broken rotor bar faults.

[0113] In fact, the dashed line (BRB fault) shows a significant amplitude change at the fault characteristic frequency, which is not present in the solid line (healthy motor). This indicates the presence of a broken rotor bar fault.

[0114] The normalized FFT envelope provides a clearer representation of the frequency components, making fault features more discernible. The envelope detection algorithm effectively highlights the differences between healthy and faulty states.

[0115] Another aspect of this disclosure is to provide an indicator for detecting these faults (e.g., misalignment faults or broken rotor bar faults) and for monitoring their evolution over time. According to... Figure 1 After calculating the FFT envelope 17 of the electrical input signal 10, the total energy of the fault characteristics is calculated. This total energy represents the FFT envelope curve (…). Figure 3 and Figure 5 The sum of the harmonic fault areas between the dashed line on the X-axis and the X-axis. By calculating the total energy of the fault characteristics, this method captures the cumulative effect of all harmonic components associated with the fault. This method provides a more comprehensive measurement of the presence and severity of the fault, thereby enhancing the sensitivity of fault detection.

[0116] As in Figure 3 and Figure 5As can be observed above, misalignment faults exist ( Figure 3 ) or broken rotor bar failure ( Figure 5 In this case, the area of ​​fault harmonics increases significantly. Summing the harmonic fault region between the RMS envelope curve and the x-axis effectively reduces the impact of noise on the fault detection process. This method emphasizes the dominant frequency component, making fault characteristics more identifiable and improving the clarity of the spectrum.

[0117] Figure 6 , Figure 7 and Figure 8 The total energy of the envelope output for the broken rotor bar (BRB) fault interval is shown in histogram form by calculating it using different types of signals (reactive power, current, and voltage). Each graph presents results for various speed references and torque percentages, providing a comprehensive comparison of fault detection capabilities using different signal types. Furthermore, indicators are calculated for different resolution frequencies to assess the impact of resolution on the indicator's relevance. Each bar in the histogram represents the sum of the fault regions (all harmonics) at a given operating point (speed and load) of the induction motor.

[0118] The output, a fault indicator representing the calculated total energy, provides a quantifiable measure of the presence and severity of the fault. This fault indicator can be easily monitored and compared to predefined thresholds, enabling automated and objective fault detection.

[0119] Figure 6 The indicators associated with reactive power characteristics are shown. Figure 6 (a) to Figure 6 (e) Each represents the envelope output of the BRB fault interval at different speed references (80%, 90%, 95%, 100%, and 105%) and varying torque percentages (80%, 90%, 95%, and 100%). The horizontal axis represents the torque percentage. The vertical axis represents the envelope output value. The legend indicates different conditions (healthy and faulty) at various resolution frequencies (2kHz, 4kHz, and 6kHz).

[0120] Figure 7 The indicator associated with the current signature is shown. Figure 7 (a) through 7(e) represent the envelope output of the BRB fault interval at different speed references (80%, 90%, 95%, 100%, and 105%) and varying torque percentages (80%, 90%, 95%, and 100%). The horizontal axis represents the torque percentage. The vertical axis represents the envelope output value. The legend indicates different conditions (healthy and faulty) at various resolution frequencies (2kHz, 4kHz, and 6kHz).

[0121] Similar to Figure 6 and Figure 7, Figure 8 Indicators associated with voltage characteristics are shown. Figure 8 (a) through 8(e) represent the envelope output of the BRB fault interval at different speed references (80%, 90%, 95%, 100%, and 105%) and varying torque percentages (80%, 90%, 95%, and 100%). The horizontal axis represents the torque percentage. The vertical axis represents the envelope output value. The legend indicates different conditions (healthy and faulty) at various resolution frequencies (2kHz, 4kHz, and 6kHz).

[0122] As in Figure 6 As can be observed above, the envelope output value of the reactive power signal shows a clear difference between healthy and fault conditions, with higher values ​​in the fault condition, particularly at higher torque percentages. By comparison, Figure 7 Current signal and Figure 8 The envelope output value of the voltage signal also shows a clear difference between healthy and faulty conditions, but compared with... Figure 6 Compared to reactive power signals, the overall value is lower. Using reactive power signals, compared to current or voltage signals, allows for the derivation of the most relevant indicators for detecting broken rotor bars in induction motors. Similarly, it can be shown that when using active power signals, the most relevant indicators for detecting misalignment faults are obtained.

[0123] In all the graphs, the envelope output value increases with higher speed reference and torque percentage, indicating stronger fault characteristics under these conditions, which makes it easier to detect BRB faults.

[0124] Each Figure 6-8 The graphs in the figure show the envelope output values ​​at different resolution frequencies (2kHz, 4kHz, 6kHz). As can be observed, the envelope output values ​​vary with the resolution frequency, with some resolution frequencies showing a more significant difference between healthy and faulty conditions. However, Figure 6-8 This demonstrates that faults can be effectively detected regardless of the test resolution or frequency.

[0125] Figure 6 , Figure 7 and Figure 8 Together, we demonstrate the effectiveness of using different signal types (reactive power, current, and voltage) to detect broken rotor bar faults in induction motors, and the robustness of the proposed method to frequency resolution. Each signal type provides valuable insights, with the reactive power signal showing the highest envelope output value, followed by voltage and current signals. Comparisons between different speed references, torque percentages, and frequency resolutions highlight the robustness of the fault detection method, making it a reliable tool for predictive maintenance in industrial applications.

[0126] The proposed technical solutions are of interest because they can directly use electrical signals provided by electrical sensors already embedded in the motor driver, without requiring additional specific sensors.

[0127] The metrics can then be compared to defined thresholds to determine if they indicate a major malfunction of the induction motor. The values ​​of the thresholds can be derived using deep learning algorithms and can depend on the type of induction motor or its operating environment.

[0128] Figure 9 This is a block diagram of a system 100 for detecting faults in an induction motor. The system is designed to process, analyze, and detect faults such as misalignment and broken rotor bars from the input signal 10 from the induction motor.

[0129] System 100 includes a signal acquisition module 101, which is responsible for receiving input signals from the induction motor. The input signal can be a voltage signal, a current signal, or a power signal (active or reactive). The signal acquisition module captures the signal, which contains information about the operating state of the motor.

[0130] The memory unit 102 stores the acquired signal data and any intermediate processing results. It provides the system with the necessary storage to perform complex calculations and analyses.

[0131] Processing unit 103 is the core component responsible for transforming the input signal 10 to the frequency domain using Fast Fourier Transform (FFT). It applies an envelope detection algorithm to the FFT spectrum, calculates the root mean square (RMS) of the envelope, and analyzes the calculated RMS envelope to detect fault characteristics. Processing unit 103 performs the necessary calculations to identify the presence of faults in the induction motor.

[0132] According to an embodiment, system 100 may further include a fault indicator module 104, which calculates the total energy of a fault characteristic by summing the harmonic fault area between the FFT envelope curve and the x-axis. The fault indicator module 104 may output a fault indicator representing the calculated total energy and compare the fault indicator with a threshold to detect a fault when the fault indicator exceeds the threshold. The fault indicator module 104 provides a clear indication of the presence and severity of a fault in the induction motor.

[0133] According to an embodiment, system 100 may further include a user interface 105 that displays fault detection results to an operator. It provides a visual representation of the detected faults and their severity, allowing the operator to take appropriate maintenance measures. The user interface ensures that the system's output is easily accessible and interpretable by the user.

[0134] System 100 is capable of real-time fault detection, thereby providing timely and accurate information about the condition of the induction motor. Specifically, the input signal 10 from the induction motor is first captured by the signal acquisition module 101. The data is then stored in memory 102 and processed by the processing unit 103. The fault indicator module 104 analyzes the processed data and generates a fault indicator, which can be displayed to the operator via the user interface 105.

[0135] Industrial application

[0136] This disclosure is not limited to the detection of misalignment faults or broken rotor bar faults described herein, which are merely examples. This disclosure includes every alternative that would occur to those skilled in the art upon reading this document. It is worth noting that the technical solutions presented herein can be used to detect any faults that affect the frequency characteristics of an induction motor.

[0137] Citation List

[0138] Non-patent literature

[0139] For any purpose, the following non-patented elements are referenced:

[0140] -Young-Jun Yoo, “Fault Detection of Induction Motors Using Feature Selection by Principal Component Analysis via Fast Fourier Transform”, International Journal of Precision Engineering and Manufacturing, Vol. 20, No. 9, pp. 1543-1553, 2019.

Claims

1. A method for detecting faults in an induction motor, comprising: Receive input signal (10) from the induction motor; The input signal (10) is transformed to the frequency domain using Fast Fourier Transform (FFT) (11) to obtain the FFT spectrum; The envelope detection algorithm (12-15) is applied to the FFT spectrum to obtain the envelope representing the amplitude change of the frequency components of the FFT spectrum; Calculate the root mean square (RMS) of the envelope described in (16); The calculated RMS envelope (17) is analyzed to detect at least one frequency characteristic representing a fault in the induction motor, referred to as a fault characteristic.

2. The method for detecting faults according to the preceding claim, wherein, Applying the envelope detection algorithm to the FFT spectrum includes: - Rectify the FFT spectrum (12); - Smooth the rectified FFT spectrum (15) with a low-pass filter to obtain the envelope.

3. The method for detecting faults according to claim 1 or 2, wherein, The detected fault belongs to the following group: -Misalignment fault; - Broken rotor bar failure.

4. The method for detecting a fault according to any one of the preceding claims, wherein, The calculated RMS envelope includes: - The total energy of the at least one fault feature is calculated by summing the harmonic fault areas between the curve representing the RMS envelope and the x-axis; - Output a fault indicator representing the calculated total energy; - Compare the fault indicator with a threshold to detect a fault when the fault indicator exceeds the threshold.

5. The method for detecting a fault according to any one of the preceding claims, wherein, The input signal (10) belongs to the group including the following: - Voltage signal; - Current signal; -Reactive power signal; - Active power signal.

6. The method for detecting a fault according to any one of the preceding claims, wherein, The FFT spectrum is normalized before the envelope detection algorithm is applied.

7. A system (100) for detecting faults in an induction motor, comprising: The signal acquisition module (101) is configured to receive an input signal (10) from the induction motor. The processing unit (103) is configured as follows: The input signal is transformed to the frequency domain using a Fast Fourier Transform (FFT) to obtain the FFT spectrum; An envelope detection algorithm is applied to the FFT spectrum to obtain an envelope representing the amplitude variation of the frequency components of the FFT spectrum. Calculate the root mean square (RMS) of the envelope; The calculated RMS envelope is analyzed to detect at least one frequency characteristic representing a fault in the induction motor, referred to as a fault characteristic.

8. The system (100) for detecting faults according to claim 7, wherein, The processing unit (103) is further configured to calculate the total energy of the at least one fault feature by summing the harmonic fault area between the curve representing the RMS envelope and the x-axis, and wherein the system (100) further includes a fault indicator module (104) configured to output a fault indicator representing the calculated total energy and compare the fault indicator with a threshold to detect a fault when the fault indicator exceeds the threshold.

9. The system (100) for detecting faults according to claim 8 further includes a user interface (105) configured to display fault detection results to an operator.

10. The system (100) for detecting faults according to any one of claims 7 to 9 is further configured to implement the method for detecting faults according to any one of claims 1 to 6.

11. A computer software comprising instructions that, when executed by a processor, implement at least a portion of the method according to any one of claims 1 to 5.

12. A computer-readable non-transitory recording medium on which software is registered, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images