System and method for detecting electrical imbalance and early winding fault of industrial induction machine

Abstract

Systems and methods for detecting phase voltages of a multi-phase electric machine are described. The method includes receiving, by a controller, a set of readings from an auxiliary coil mounted in the electric machine; converting, by the controller, the set of readings into a plurality of principal components, wherein the principal components are based on feature vectors and feature values generated on the basis of the reading spectrum; and inputting the plurality of principal components into a trained machine learning model to obtain a predicted phase voltage of the motor.

Description

Technical Field

[0001] The present disclosure relates to detecting winding voltage and winding damage in electric motors during operation. In particular, systems and methods are described herein for predicting early winding faults or voltage imbalances between different windings in electric motors and electric motor power supplies fed from an electrical grid. Background Art

[0002] Winding imbalance can occur in electric motors for a variety of reasons. One possible cause of unbalanced three-phase power in an industrial environment is a large number of single-phase industrial equipment being fed by the same phase. This can cause a reduction in voltage on that phase while leaving the other phases unaffected. Another possible cause of voltage phase imbalance at the motor is an electrical short between turns in the motor’s stator windings, which can result in degraded performance. This failure mode can occur when the dielectric coating (insulation) of the magnet wire of the stator field winding becomes damaged, allowing adjacent turns to become electrically connected, reducing the number of effective turns in that coil. Over time, the insulation of the magnet wire can become compromised due to insulation breakdown, contamination, or relative movement caused by vibration between the windings that wears away the insulation. Motors exposed to frequent start-stop cycles may be most susceptible to this failure.

[0003] Motor design makes the machine relatively insensitive to small phase imbalances. However, at certain voltage imbalance levels, the motor begins to be severely affected, which can lead to permanent damage, shortened life, or immediate, unexpected shutdown. If not addressed early, the worsening of winding insulation breakdown can lead to catastrophic motor failure if a short circuit connects two different windings or any winding to ground. Summary of the Invention

[0004] A first aspect of the present disclosure provides a method for detecting phase voltages of a multi-phase motor. The method includes: receiving, by a controller, a set of readings from an auxiliary coil installed in the motor; converting, by the controller, the set of readings into a plurality of principal components, wherein the principal components are based on eigenvectors and eigenvalues ​​generated based on a spectrum of the readings; and inputting the plurality of principal components into a trained machine learning model to obtain predicted phase voltages of the motor.

[0005] According to an implementation of the first aspect, the method further comprises providing an alert based on comparing the predicted voltage with the expected voltage value.

[0006] According to an implementation of the first aspect, determining the plurality of principal components includes scaling a set of readings received from the auxiliary coil.

[0007] According to an implementation of the first aspect, converting the set of readings into a plurality of principal components includes detecting, by the controller, a dominant frequency component of the set of readings.

[0008] According to an implementation of the first aspect, training the machine learning model includes: storing a plurality of principal components in a training database; and providing labels for the stored principal components and predicted phase voltages.

[0009] According to an implementation of the first aspect, the labels provided to the stored principal components and predicted phase voltages include at least one of an operational load label and a measured voltage label.

[0010] According to an implementation of the first aspect, the operational load tag indicates power consumed by a load connected to the multi-phase electric machine.

[0011] According to an implementation of the first aspect, the predicted voltage indicates a supply voltage imbalance and / or a value outside a nominal range measured during operation of the multi-phase electric machine.

[0012] According to an implementation of the first aspect, the method further includes: calculating a difference between the predicted voltage and the measured voltage; and comparing the calculated difference to a threshold value to determine whether an imbalance exists in the multi-phase electric machine.

[0013] According to an implementation of the first aspect, the method further includes: calculating a difference between the predicted voltage and the measured voltage; and comparing the calculated difference to a threshold value to determine the health of the rotor of the multi-phase electric machine.

[0014] According to an implementation of the first aspect, the method further comprises: calculating a difference between the predicted voltage and the measured voltage; and comparing the calculated difference with a threshold value to determine degradation of the stator winding insulation during operation of the multi-phase electric machine.

[0015] A second aspect of the present disclosure provides a method for detecting phase voltages of a multiphase motor. The method includes: receiving, by a controller, a set of readings from auxiliary coils installed in the motor; calculating, by the controller, a root mean square (RMS) value of the set of readings from the auxiliary coils installed in the motor; and inputting the RMS value into a trained machine learning model to obtain predicted phase voltages of the motor.

[0016] A third aspect of the present disclosure provides a system for detecting phase voltages of a multi-phase motor. The system includes a controller configured to: receive a set of readings from an auxiliary coil installed in the motor; convert the set of readings into a plurality of principal components, wherein the principal components are based on eigenvectors and eigenvalues ​​generated based on the set of readings; and input the plurality of principal components into a trained machine learning model to obtain predicted phase voltages of the motor.

[0017] According to an implementation of the third aspect, the controller is further configured to provide an alert based on comparing the predicted voltage with the expected voltage value.

[0018] According to an implementation of the third aspect, the controller configured to convert the set of readings into a plurality of principal components is further configured to detect a dominant frequency component of the set of readings.

[0019] According to an implementation of the third aspect, the controller configured to train the machine learning model is further configured to: store a plurality of principal components in a training database; and provide labels for the stored principal components and the predicted phase voltages.

[0020] According to an implementation of the third aspect, the labels provided to the stored principal components and predicted phase voltages include at least one of an operational load label and a measured voltage label.

[0021] According to an implementation of the third aspect, the controller is further configured to: predict a load connected to the multi-phase motor; and compare a rated load associated with the multi-phase motor with the predicted load; and provide an alarm based on determining whether the predicted load exceeds the rated load of the multi-phase motor.

[0022] According to an implementation of the third aspect, the predicted load is used to calculate the energy usage of the multi-phase electric machine.

[0023] According to an implementation of the third aspect, energy usage of the multi-phase electric machine is used to determine the efficiency of the electric machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The subject matter of the present disclosure will be described in more detail below based on the exemplary drawings. All features described and / or illustrated herein may be used individually or in combination in various combinations. The features and advantages of various embodiments will become clear by reading the following detailed description with reference to the accompanying drawings, which illustrate the following:

[0025] Figure 1 shows a simplified block diagram depicting a system for detecting winding voltage imbalance in an electric machine during operation according to one or more examples of the present disclosure;

[0026] Figure 2A-2C shows different configurations of auxiliary coils mounted on a stator for detecting winding voltage imbalance in an electric machine during operation according to one or more examples of the present disclosure;

[0027] Figure 3 is a block diagram of an exemplary sensor device associated with an auxiliary coil according to one or more examples of the present disclosure;

[0028] Figures 4A-4C depicts a graph showing data collected from a multi-coil auxiliary coil assembly installed in a stator of an electric machine according to one or more examples of the present disclosure;

[0029] Figures 5A-5Ca graph depicting performance of a multi-output regressor using only the RMS value of a single coil of an auxiliary coil assembly attached to a stator of an electric machine according to one or more examples of the present disclosure;

[0030] Figures 6A-6C depicts a graph generated by analyzing spectral information from an induced voltage signal according to one or more examples of the present disclosure;

[0031] Figure 7 An exemplary process for detecting voltage phase imbalance in a multi-phase electric machine according to one or more examples of the present disclosure is shown; and

[0032] Figure 8 An exemplary process for training a machine learning model to detect winding voltage imbalance in a multi-phase electric machine according to one or more examples of the present disclosure is shown. DETAILED DESCRIPTION

[0033] This disclosure describes detecting unbalanced voltages at the stator windings of AC motors that can be present in industrial environments. Unbalanced voltages at the windings can degrade the power factor, efficiency, and life of motors and generators. Industrial motors are typically designed to be powered by a specific single-phase or three-phase AC voltage, such as 110V, 220V, 230V, 400V, 460V, etc. While this voltage can be nearly perfect at the substation, voltage variations may be encountered at the final electrical load location (power outlet in industrial, residential, or other locations).

[0034] Detecting voltage imbalances in motor windings is highly desirable to trigger corrective action, either internally or externally. Motor winding voltage monitoring can be easily performed on motors powered by a variable frequency drive (VFD). While VFDs can monitor and regulate motor supply voltages across the various stator windings, they are expensive. Furthermore, VFDs may require specific environmental operating conditions. Furthermore, many generator applications don't even require a variable speed option, effectively rendering a VFD useless in these situations.

[0035] For direct-on-line (DOL) motors without a VFD, winding voltage monitoring can be accomplished with potentially expensive motor health monitoring equipment. Typical motor health monitoring equipment only measures the motor voltage applied at the motor terminals. In the early stages of a winding fault, the terminal voltages may still be largely balanced, so this traditional approach may not detect an impending motor failure. Many types of motor health monitoring equipment for DOL motors must be temporarily connected by trained technicians. In these cases, motor monitoring is not performed continuously, but only at certain maintenance intervals or when a problem is suspected. The added expense associated with this monitoring method is a major factor preventing many customers from implementing winding voltage monitoring.

[0036] This disclosure provides a low-cost, permanent monitoring solution for early-stage winding fault detection in electric motors. The present invention is implemented as an integral motor component and imposes no additional installation costs on the customer. The cost of the present invention is significantly lower than that of VFDs or typical motor health monitoring devices that are either permanently connected or partially connected by a technician. Customers can utilize this disclosure to reduce unplanned downtime.

[0037] Figure 1 A simplified block diagram depicting a system for detecting winding voltage imbalance in an electric machine during operation is shown, according to one or more examples of the present disclosure. Figure 1 A multiphase electric machine 102 is depicted, including a rotor 104 and a stator 106. The rotor 104 and stator 106 can have multiple different windings connected to different phases of a voltage source. The multiphase electric machine 102 can be connected to a load 120. An auxiliary coil assembly 108 is attached to the stator 106 of the multiphase electric machine 102. The auxiliary coil assembly 108 is positioned so that the rotating magnetic field generated by the currents in the windings of the rotor 104 and stator 106 induces a voltage in the auxiliary coil 122. The magnitude and phase angle of the induced voltage in the auxiliary coil spectrum correspond to the voltages in the different windings connected to the rotor 104 and stator 106 of the multiphase electric machine 102. The induced voltage in the auxiliary coil 122 is a function of the voltage in the windings of the stator 106. By analyzing the voltage signal captured by the auxiliary coil 122, the voltages in the windings of the rotor 104 and stator 106 can be approximated. The calculated voltage in the windings of the stator 106 can be used to identify problems within the insulation of the motor windings and provide information about the status of the applied external power supply. The calculated voltage in the windings of the stator 106 can also be used to estimate the operating torque and power of the motor. The voltage signal from the auxiliary coil assembly 108 can also be used to identify faults associated with the rotor 104.

[0038] In some embodiments, the current in the stator windings induces a voltage in the auxiliary coil assembly. The stator winding that is most aligned with a particular coil of the auxiliary coil assembly will induce the highest voltage in the auxiliary coil 122, while the stator winding that is least aligned with the auxiliary coil assembly will induce the lowest voltage in that auxiliary coil 122. These induced voltages add together to form a total induced voltage in the auxiliary coil 122.

[0039] If the auxiliary coil 122 includes two or more coils, and if these coils are placed at different angular positions around the stator, there will be magnitude and phase differences between the total induced voltages in these coils. Changes in the voltage of one of the motor stator windings or the rotor induced voltage (i.e., if the motor becomes unbalanced) can also cause phase differences between the coils of the auxiliary coil assembly. The magnitude of the induced voltage in the coils of the auxiliary coil assembly will also be affected.

[0040] An imbalance in the stator winding voltage or rotor induced voltage (e.g., from a winding fault or grid problem) can be uniquely detected by a phase change in the induced voltage in the coils of the auxiliary coil assembly 122 and can be distinguished from motor torque load changes. The auxiliary coil assembly 108 includes a sensor device 110 that measures the voltage induced in the coils of the auxiliary coil assembly 108 by the rotating magnetic field of the motor stator windings of the multi-phase motor 102. The auxiliary coil assembly 108 also includes an analog-to-digital (A / D) converter 112 for sampling the voltage signal measured by the sensor device 110 of the auxiliary coil assembly 108. Typical sampling rates can be 3, 6, or 10 kHz. In some embodiments, the A / D converter 112 can be a low-power microcontroller with an integrated A / D converter and a processor for signal conditioning and interpretation. The A / D converter 112 provides the sampled voltage from the auxiliary coil 108 to a controller 114. Figure 3 The sensor device 110 is described in more detail below.

[0041] The controller 114 transmits the sampled voltage from the auxiliary winding 108 to the memory 118 for storage. In some embodiments, the controller 114 may process the sampled voltage and then provide the sampled voltage to the machine learning model 116. The machine learning model 116 uses the sampled voltage to predict voltage phase imbalance of the multi-phase motor 102 during operation.

[0042] The machine learning model 116 is trained to predict the occurrence of winding voltage imbalance in the multi-phase electric machine 102 based on information received from the multi-phase electric machine 102 and the auxiliary coil assembly 108. In some embodiments, the machine learning model 116 may be a single-output regression model that uses information obtained from the sensor device 110 connected to the auxiliary coil assembly 108 to output a predicted voltage imbalance of the stator windings or a predicted load 120 of the multi-phase electric machine 102. In some other embodiments, the machine learning model 116 may be a multi-output regression model that outputs a predicted value of the load 120 connected to the multi-phase electric machine 102 in addition to the predicted winding voltage imbalance.

[0043] In such an embodiment, this information may be used to detect deviations from nominal winding voltages in the multi-phase motor 102 caused by faults in the power supply or by faults in the motor's stator or rotor.

[0044] The machine learning model 116 can be programmed with more than one algorithm to predict winding voltage imbalance of the multi-phase motor 102. For example, the machine learning model 116 can be configured to use the RMS value of the voltage induced in the auxiliary winding to predict the winding voltage imbalance of the multi-phase motor 102. In other examples, the machine learning model 116 can be configured to use the frequency spectrum of the induced voltage of the auxiliary winding and a data reduction technique called principal component analysis to predict the phase imbalance of the multi-phase motor 102. Both methods are explained in more detail below.

[0045] Figure 2A-2C Depicted are different configurations of stator-mounted auxiliary coils for detecting winding voltage imbalances in a multi-phase electric machine during operation, according to one or more examples of the present disclosure. Figure 2A A single coil auxiliary coil assembly 108 is shown associated with the stator 106 of the multi-phase electric machine 102. In the single coil auxiliary coil assembly 108, the coil span is such that the coil side begins at one particular winding slot and extends over the other winding slots at least once. In this configuration, the voltage and current in any winding of the stator 106 induce a related voltage in the auxiliary coil 122 of the auxiliary coil assembly 108. This allows the auxiliary coil 122 to be used to detect problems in any winding of the multi-phase electric machine 102 using only the single coil auxiliary coil assembly. Figure 2A In FIG, the stator 106 winding has four poles and three windings. Specifically, these three windings include a U-shaped winding 202, a V-shaped winding 204, and a W-shaped winding 206. The coils of the auxiliary coil assembly 108 at least partially cover each of the three windings and are therefore able to pick up voltage variations in all three windings. The auxiliary coil is connected to a sensor device 110, which measures the voltage induced in the auxiliary coil 108.

[0046] Figure 2B and Figure 2C Auxiliary coils are shown arranged in three or nine winding configurations, respectively, according to one or more examples of the present disclosure. In some embodiments, the auxiliary coil assembly 108 consists of an insulated conductor ring that is inserted ("embedded") into specific stator slots adjacent to the windings of the stator 106 of the multi-phase electric machine 102. Specifically, Figure 2B depicts a 3-coil auxiliary coil assembly 108, Figure 2C A nine-coil auxiliary coil assembly 108 is depicted. In some cases, the more coils in the configuration of the auxiliary coil assembly 108, the more sensitive the auxiliary coil assembly 108 is to detecting winding voltage imbalances, and the more reliable the machine learning algorithm is in detecting winding voltage imbalances in the multi-phase electric machine 102. The windings and configuration of the auxiliary coils are disclosed in greater detail in PCT Application No. PCT / IB2022 / 054404, the entire contents of which are incorporated herein by reference.

[0047] In some embodiments, the auxiliary coil assembly 108 is composed of a flexible printed circuit board (PCB) having a polyimide or similar high-temperature polymer substrate. The conductive coil is formed by other conductor loops that are wave-wound, concentrically nested, or attached to the flexible polymer substrate. Most flexible PCB coils include a single layer of conductor, although multi-layer flexible PCBs are also possible. The flexible PCB can be bent and bonded to the tips of the stator teeth using an adhesive. The flexible PCB can be rectangular or have any other shape to cover certain areas of the stator. Based on the induced voltage in the auxiliary coil at the air gap, it can detect winding problems in any phase as well as irregularities in the power supply.

[0048] Figure 3 is a block diagram of an exemplary sensor device associated with an auxiliary coil according to one or more examples of the present disclosure. The sensor device 110 includes a processor 304 (such as a central processing unit (CPU)) and / or logic for executing computer-executable instructions to perform the functions, processes, and / or methods described herein. The computer-executable instructions are stored locally and accessed from a non-transitory computer-readable medium (such as a data storage device 306), which may be a hard drive or flash drive. Random access memory (RAM) 308 is main memory for loading and processing instructions executed by the processor 304. A communication segment 312 can be connected to a wired or cellular network, as well as a local or wide area network. The sensor device 110 may also include a detector 302 for detecting a voltage induced in the auxiliary coil 108. The sensor device 110 may also include an indicator 310 that generates a message or alarm when a voltage detected or predicted voltage phase imbalance in the auxiliary coil exceeds a specified threshold. A bus may connect the processor 304, RAM 308, data storage device 306, and / or communication segment 312. The components within sensor device 110 can communicate with each other using a bus. The components within sensor device 110 are merely exemplary and may not include every component within sensor device 110. Additionally and / or alternatively, sensor device 110 may also include components that may not be included. For example, sensor device 110 may not include communication segment 312.

[0049] Figures 4A-4C A diagram illustrating data collected from a particular coil of a multi-coil auxiliary coil assembly installed in a stator of an electric motor according to one or more examples of the present disclosure is depicted. For example, the auxiliary coil assembly 108 may have three coils A, B, and C, each corresponding to a winding U, V, W of the multi-phase electric motor 102. Figure 4AData is shown collected from coil C of auxiliary coil assembly 108 while multi-phase motor 102 is operated under load 120. Winding W of multi-phase motor 102 is powered by variable voltages, including 270V, 256.5V, and 283.5V using a controllable three-phase power supply. Figure 4A The sampled auxiliary coil voltage induced in the coil C of the auxiliary coil assembly 108 in each case is shown in FIG. Figure 4A Graph 400 plots the voltage induced in the auxiliary coil assembly 108 on the y-axis versus time on the x-axis. Graph 400 includes three curves 402, 404, and 406. Curve 402 represents the auxiliary voltage induced in coil C of the auxiliary coil assembly 108 when the multi-phase motor 102 is powered by 283.5V. Curve 404 represents the auxiliary voltage induced in coil C of the auxiliary coil assembly 108 when the multi-phase motor 102 is powered by 270V. Curve 406 represents the auxiliary voltage induced in coil C of the auxiliary coil assembly 108 when the multi-phase motor 102 is powered by 256.5V. These curves are sampled at different times, and the phase relationship between the three curves 402, 404, and 406 is selected to better distinguish these curves.

[0050] The processor 304 of the sensor device 110 calculates Figure 4A The peak and RMS voltages of the sampled voltage signals of the voltage induced in the coils of the auxiliary coil assembly 108 of the motor are shown. Specifically, Figure 4B The RMS values ​​of curves 402, 404, and 406 are shown in FIG. Point 452 depicts Figure 4A Point 454 depicts the RMS voltage of signal 406 in graph 400. Figure 4A The RMS voltage of signal 404 in graph 400. Point 456 depicts Figure 4A The RMS voltage of signal 402 in graph 400 .

[0051] Figure 4CGraphs of data generated from three different coils of the auxiliary winding assembly 108 while the multi-phase motor is operating at different loads are shown. Specifically, graph 476 depicts the peak voltage induced in coil A of the auxiliary winding assembly 108. Graph 478 depicts the peak voltage induced in coil B of the auxiliary winding assembly 108, and graph 480 depicts the peak voltage generated in coil C of the auxiliary winding assembly 108. In each graph, different inductance values ​​are used in series with one winding of the motor stator 106 to introduce winding voltage imbalance. The inductance value is varied by connecting different impedance values ​​to the multi-phase motor 102. For example, inductance values ​​of 0 mH, 5 mH, and 10 mH are used in series with the multi-phase motor 102. The different introduced inductance values ​​are plotted on the x-axis, and the induced peak voltage is plotted on the y-axis. The multi-phase motor 102 is operated at different loads 120 of 0 horsepower, 4.5 horsepower, and 8.3 horsepower. As can be inferred from the different figures, the RMS voltage induced in the auxiliary winding 122 generally decreases as the motor load increases and the phase imbalance increases. In such an embodiment, we cannot use only the peak or RMS value to distinguish between imbalance and load changes (see paragraph 26). When the winding voltage is unbalanced, the RMS voltage decreases more, and when the motor load increases, the RMS voltage decreases less. Figure 4C As shown, in each of coils A, B, and C, the RMS voltage value is nearly linearly related to the load torque (i.e., power) or winding voltage imbalance. Furthermore, when motor load and winding voltage imbalance are present simultaneously, the combined drop in measured RMS voltage is even higher. Using the RMS value of the voltage induced in the auxiliary coil 108 to determine winding voltage imbalance is useful when the processing and memory capabilities of the processor 304 associated with the sensor device 110 are limited, although it should be noted that the usefulness of using only the RMS value is limited (i.e., one cannot distinguish between load changes and winding voltage imbalance).

[0052] The measured RMS values ​​of the coils of the auxiliary coil assembly 108 are stored in memory and paired with known torque values ​​or known winding voltage imbalances. The known torques and known winding voltage imbalances are referred to as 'labels'. These data pairs are used to form a training data set and are stored in memory 108. The training data set is used to train the machine learning model 116 so that the machine learning model 116 can predict the voltage phase imbalance of the multi-phase motor 102. In the case where the machine learning model 116 is a single-output model, only a single label is provided to the training data set. This label can be used to identify the measured voltage phase imbalance of the multi-phase motor 102. This label is used to train the machine learning model 116 to predict the voltage phase imbalance in the multi-phase motor 102.

[0053] In the case where the machine learning model 116 is a multi-output model, two or more labels can be provided to the training data set. These labels can be used to identify the measured load 120 and the measured voltage phase imbalance of the multi-phase motor 102. These labels are used to train the machine learning model 116 to predict the voltage phase imbalance in the multi-phase motor 102 and the load 120 connected to the multi-phase motor 102. The measured values ​​of the load and operating parameters of the multi-phase motor 102 during operation are compared with the predicted values ​​of the load and other operating parameters of the multi-phase motor 102 generated by the machine learning model 116. This comparison is used to train the machine learning model 116. The machine learning model 116 is then used to predict the winding voltage imbalance in the multi-phase motor 102. If the predicted winding voltage imbalance exceeds a certain threshold, an alert is generated to prompt the motor administrator to take remedial action.

[0054] To accurately predict winding voltage unbalance in a motor, data from at least two coils of an auxiliary coil assembly installed in the motor stator should be used. The angular spacing of the coils provides additional information needed to infer the load power and voltage unbalance in the windings. The angular spacing of the coils results in the following characteristics:

[0055] • A change in torque load, ie power, will scale the coil voltage equally.

[0056] • Changes in winding voltage imbalance in one or more windings affect some coils more than others.

[0057] Figure 5A The performance of a multi-output regressor using only the RMS value of a single coil of an auxiliary coil assembly connected to the stator of the electric machine is shown. Figure 5A Graph 502 plots the actual power of the motor on the x-axis and the predicted power on the y-axis. Similarly, Figure 5A Graph 504 plots predicted voltage phase imbalance on the y-axis and actual winding voltage imbalance on the x-axis. Figure 5A As can be seen from graphs 502 and 504 , when only the RMS value of a single coil of the auxiliary coil assembly 108 is used, the load power and winding voltage imbalance values ​​cannot be reliably predicted.

[0058] Figure 5B The performance of a multi-output regressor using the RMS values ​​of two coils of an auxiliary coil assembly attached to the stator of an electric machine is shown. Figure 5B Graph 552 plots the actual power of the motor on the x-axis and the predicted power on the y-axis. Similarly, Figure 5B Graph 554 plots the predicted voltage phase imbalance on the y-axis and the actual winding voltage imbalance on the x-axis. Figure 5BAs seen in Figures 552 and 554, the load power and winding voltage imbalance values ​​can be predicted with greater certainty when the RMS values ​​of the two coils of the auxiliary coil assembly 108 are used compared to using the RMS value of only a single coil of the auxiliary coil assembly 108.

[0059] Figure 5C The performance of a multi-output regressor using the RMS values ​​of three coils of an auxiliary coil assembly attached to the stator of an electric machine is shown. Figure 5C Graph 572 plots the actual power of the motor on the x-axis and the predicted power on the y-axis. Similarly, Figure 5C Graph 574 plots the predicted winding voltage unbalance on the y-axis and the actual winding voltage unbalance on the x-axis. Figure 5C As seen in Figures 572 and 574, when the RMS values ​​of the three coils of the auxiliary coil assembly 108 are used, the load power and winding voltage imbalance values ​​can be reliably predicted.

[0060] As described above, using RMS values ​​to predict winding voltage imbalance for a multi-phase motor 102 does not produce the most accurate results. While RMS values ​​are linearly related to winding voltage imbalance, they are also linearly related to the load 120 applied to the multi-phase motor 102. It is difficult to confidently distinguish winding voltage imbalance from varying load power, which affects the RMS value.

[0061] By analyzing the spectral information contained in the induced voltage signal rather than using only the RMS or peak value of the induced voltage signal, winding voltage imbalance in the multi-phase motor 102 can be predicted. This analysis can be performed on data acquired from a single coil of the auxiliary coil assembly 108. Figures 6A-6C Described is the use of frequency spectrum information of an induced voltage signal to predict winding voltage unbalance according to one or more examples of the present disclosure.

[0062] Figure 6A The graph generated by analyzing the spectrum information of the induced voltage signal is depicted. Specifically, Figure 6A The graph 600 shown in the figure plots the magnitude in Vs (volt-seconds) on the y-axis 602 and the frequency on the x-axis 604. The voltage of the coil is acquired at a sampling rate of approximately 10kHz. This divides the induced voltage signal into 2048 discrete frequency points, ranging from 0 to approximately 5kHz. Subsequently, principal component analysis (PCA) using singular value decomposition (SVD) is used to reduce the data from 2048 features or dimensions (corresponding to 2048 discrete frequencies) to only a few features or dimensions (typically less than 20). Each principal component is composed of generating eigenvalues ​​and eigenvectors based on the 2048 discrete frequency points. After generating the principal components, only the most important principal components (i.e., the principal components with the largest explained variance - see below) are retained for statistical analysis.

[0063] This is achieved by retaining only the principal components of the data that are most affected by load or imbalance changes. Principal components that are greatly affected by load power and imbalance changes have "larger explained variance", while principal components that are least affected tend to have "smaller explained variance". After the PCA data reduction step, the machine learning model 116 is fitted to the reduced order data set. In particular, a support vector machine (SVM) multi-output regression model is fitted (trained) to known input-output data pairs (i.e., training data sets). The data from the auxiliary coil assembly 108 can be centered and scaled. Data scaling can be performed with a minimum-maximum scaler or scaled to zero mean and unit variance. The machine learning model 116 may include a classifier, a logistic regressor, a neural network, a decision tree, or a random forest. In some embodiments, the fitting (training) of the model can be performed offline, which means not on the processor 304 of the sensor device 110. After being installed, the model can be deployed to the processor 304.

[0064] To collect training data for the machine learning model 116, the multiphase motor 102 is operated at different known loads and different known voltage phase imbalances, which can be achieved by adding series inductance. During motor operation, data is continuously collected from the coils of the auxiliary coil assembly 108. This produces a series of DFTs, where each DFT corresponds to a specific pair of imbalance values ​​and load power values.

[0065] The training dataset is used to train the machine learning model 116, enabling the machine learning model 116 to predict winding voltage imbalances in the multi-phase motor 102. The training dataset is provided with labels. If the machine learning model 116 is a single-output model, only a single label is provided to the training dataset. This label can be used to identify the measured winding voltage imbalances of the multi-phase motor 102. This label is used to train the machine learning model 116 to predict winding voltage imbalances in the multi-phase motor 102.

[0066] In the case where the machine learning model 116 is a multi-output model, two or more labels can be provided to the training dataset. These labels can be used to identify the measured load 120 and the measured winding voltage imbalance of the multi-phase motor 102. These labels are used to train the machine learning model 116 to predict the winding voltage imbalance in the multi-phase motor 102 and the load 120 connected to the multi-phase motor 102.

[0067] After the machine learning model 116 is fitted and trained with the training data, the machine learning model 116 is used to predict both the winding voltage imbalance and the load 120 operating on the multi-phase electric machine 102 based on the signals received from the single coil or multiple coils of the auxiliary coil assembly.

[0068] Figure 6BDepicted are exemplary performance graphs of a multi-output regressor for inferring both added series inductance (i.e., winding voltage imbalance) and load power provided to the multi-phase motor 102. In this embodiment, the machine learning model 116 is a multi-output model that estimates both the load and winding voltage imbalance of the multi-phase motor 102. In graph 652, using the provided training data for the PCA component, the machine learning model 116 is able to predict the load (i.e., torque, power) of the multi-phase motor 102. In graph 654, using the provided training data for the PCA component, the machine learning model 116 is able to predict the winding voltage imbalance of the multi-phase motor 102.

[0069] Figure 6C An example performance graph of a single-output regressor using a voltage signal from coil A, B, or C of the auxiliary coil assembly 108 is shown. In such an embodiment, the machine learning model 116 is a single-output model and estimates only the winding voltage imbalance, not the load 120 associated with the multi-phase motor 102. The multi-phase motor 102 is operated at different load powers and different balance / imbalance conditions (as described above, achieved with 5mH and 10mH series inductors). Only the inductance value of the series inductor is used for training, and the load power value is ignored. Figure 6C The performance graphs in were obtained using new test data. Due to the power variation, the predictions for winding voltage unbalance for each coil vary significantly, as shown in graphs 678, 680, and 682.

[0070] Figure 7 An exemplary process for detecting winding voltage imbalance in a multi-phase motor according to one or more examples of the present disclosure is shown. Process 700 may be performed by Figure 3 However, it should be appreciated that any of the following blocks may be performed in any suitable order, and that process 700 may be performed by any suitable computing device and / or controller in any environment. For example, process 700 may also be performed by Figure 1 The controller 114 is shown executing.

[0071] At block 702, the processor 304 receives a set of readings from auxiliary coils installed in the motor. For example, the set of readings may be measured by the detector 302 of the sensor device 110 associated with the auxiliary coils 108 attached to the stator 106 of the multi-phase motor 102. In some embodiments, the set of readings may be time-varying voltage data, or such data may be processed, i.e., via a fast Fourier transform.

[0072] At block 704, the processor 304 converts the set of readings into a plurality of principal components, wherein the principal components are based on the eigenvectors and eigenvalues ​​generated based on the set of readings. For example, principal component analysis (PCA) using singular value decomposition (SVD) is used to reduce the spectral data from the auxiliary coil 108 from 2048 dimensions (corresponding to 2048 discrete frequencies) to only a few dimensions (typically less than 20). Each principal component is composed by generating eigenvalues ​​and eigenvectors based on the 2048 discrete frequency points. After generating the principal components, only the most important principal components are retained for statistical analysis.

[0073] At block 706, the processor 304 inputs the plurality of principal components into the trained machine learning model to obtain a predicted winding voltage imbalance for the electric machine, wherein the predicted winding voltage imbalance indicates a voltage drop during at least one of a plurality of operating phases of the electric machine. For example, the machine learning model 116 can be used to predict winding voltage imbalance for the multi-phase electric machine 102.

[0074] At block 708, the processor 304 provides an alert based on comparing the predicted winding voltage imbalance with the expected winding voltage imbalance value. In some examples, the measured values ​​of the load and operating parameters of the multi-phase motor 102 during operation are compared with the predicted values ​​of the load and other operating parameters of the multi-phase motor 102 generated by the machine learning model 116. The alert can be a visual alert for a technician to provide necessary maintenance to the multi-phase motor 102. In some examples, the alert can be an audio alert also provided to a technician to perform necessary maintenance on the multi-phase motor 102. In some other cases, the alert can be sent to a central computing system, which can determine any remedial actions that need to be taken.

[0075] Figure 8 An exemplary process for training a machine learning model to detect winding voltage imbalance in a multi-phase motor according to one or more examples of the present disclosure is shown. Process 800 may be performed by Figure 3 However, it should be appreciated that any of the following blocks may be performed in any suitable order, and that process 800 may be performed by any suitable computing device and / or controller in any environment. For example, process 800 may also be performed by Figure 1 The controller 114 is shown executing.

[0076] At block 802, the controller 114 operates the multiphase motor 102 under various loads and voltage phase imbalances. In some embodiments, as previously described, the voltage phase imbalance can be achieved by connecting an inductor in series with one of the motor phases. These various operating conditions are important for generating a comprehensive training dataset for training the machine learning model 116. Data from each of the various operating conditions is collected from the auxiliary coil 108 using the sensor device 110.

[0077] At block 804, the processor 114 computes a discrete Fourier transform (DFT) on the data. This is performed on the induced voltage signal using a sampling rate of approximately 10 kHz. In this particular example, the spectrum includes 2048 discrete frequency points ranging from 0 to approximately 5 kHz.

[0078] At block 806, the controller 114 scales the collected data. The scaling used is a feature scaling process where each frequency is a feature in the machine learning sense. One scaling option is to scale a feature so that all its values ​​are between -1 and +1. Another option that can be used is to scale the feature so that the mean is zero (to remove the offset) and the variance is 1.

[0079] At block 808, the controller 1114 calculates principal components based on principal component analysis (PCA) using singular value decomposition (SVD) to reduce the data from 2048 dimensions (corresponding to 2048 discrete frequencies) to only a few dimensions (typically less than 20). Each principal component is composed by generating eigenvalues ​​and eigenvectors based on the 2048 discrete frequency points.

[0080] At block 810, the processor 114 retains only the most significant principal components for statistical analysis. This is achieved, for example, by retaining only the principal components of the data that are most affected by changes in load or imbalance. Principal components that are most affected by changes in load power and imbalance tend to have "larger explained variance," while principal components that are least affected tend to have "smaller explained variance."

[0081] At block 812, the processor 114 fits the machine learning model 116 to the reduced order dataset for training. For example, a support vector machine (SVM) multi-output regression model is fit (trained) to the known input-output data pairs.

[0082] At block 814 , the processor 114 deploys the machine learning model 116 to predict voltage phase imbalance of the multi-phase electric machine 102 .

[0083] At block 816, the processor 114 receives input data for operation of the multi-phase electric machine 102. The input data may be received from the sensor device 110 associated with the auxiliary coil 108 attached to the stator 106 of the multi-phase electric machine 102.

[0084] At block 818, the processor 114 calculates a DFT based on the data collected from the sensor device 110. For example, this can be achieved by using a sampling rate of approximately 10 kHz on the induced voltage signal. This divides the induced voltage signal into 2048 discrete frequency points ranging from 0 to approximately 5 kHz.

[0085] At block 820, the processor 114 scales the data. In some cases, the data set is reduced using principal component analysis before data scaling. Principal component analysis is used to reduce the 2048 discrete frequency points to the most relevant data points that are most affected by winding voltage imbalance or load operation changes.

[0086] At block 822, the processor 114 executes the machine learning model using the obtained scaled principal components to obtain a predicted winding voltage imbalance. The predicted winding voltage imbalance is compared with a threshold, and if the winding voltage imbalance is outside the threshold, an alert is generated for a technician to inspect and provide remedial measures for the multi-phase motor 102. In some embodiments, if the machine learning model 116 is a multi-output regression model, the machine learning model 116 also predicts the load 120 connected to the multi-phase motor 102.

[0087] Although the subject matter of the present disclosure has been shown and described in detail in the drawings and the foregoing description, such showing and description should be regarded as illustrative or exemplary rather than restrictive. Any statements herein describing the invention should also be regarded as illustrative or exemplary rather than restrictive, as the invention is defined by the claims. It should be understood that changes and modifications may be made by one of ordinary skill in the art within the scope of the appended claims, which may include any combination of features of the different embodiments described above.

[0088] The terms used in the claims should be interpreted to have the broadest reasonable interpretation consistent with the above description. For example, use of "a" or "the" when introducing an element should not be interpreted as excluding multiple elements. Similarly, "or" statements should be interpreted as inclusive, so the statement "A or B" does not exclude "A and B" unless it is clear from the context or the above description that only one of A and B is intended. In addition, the statement "at least one of A, B, and C" should be interpreted as one or more elements of the group of elements consisting of A, B, and C, and should not be interpreted as requiring at least one of each of the listed elements A, B, and C, regardless of whether A, B, and C are related as categories. In addition, the statement "A, B and / or C" or "at least one of A, B, or C" should be interpreted to include any single entity of the listed elements, such as A, any subset of the listed elements, such as A and B, or the entire list of elements A, B, and C.

Claims

1. A method for detecting phase voltages of a multi-phase motor, the method comprising: receiving a set of readings from an auxiliary coil mounted in the motor; converting the set of readings into a plurality of principal components, wherein the principal components are based on eigenvectors and eigenvalues ​​generated based on the set of readings; as well as The plurality of principal components are input into a trained machine learning model to obtain predicted phase voltages of the motor. 2 . The method of claim 1 , further comprising providing an alert based on comparing the predicted phase voltage to an expected voltage value. 3 . The method of claim 1 , wherein determining the plurality of principal components comprises scaling the set of readings received from the auxiliary coil.

4. The method of claim 1 , wherein converting the set of readings into the plurality of principal components comprises detecting dominant frequency components of the set of readings.

5. The method of claim 1 , wherein training the machine learning model comprises: storing the plurality of principal components in a training database; as well as Labels are provided for the stored principal components and the predicted phase voltages. 6 . The method of claim 5 , wherein the labels provided to the stored principal components and the predicted phase voltages include at least one of an operational load label and a measured voltage label. The method of claim 6 , wherein the operational load tag indicates power consumed by a load connected to the multi-phase electric machine. 8 . The method of claim 7 , wherein the predicted voltage indicates a supply voltage imbalance and / or a value outside a nominal range measured during operation of the multi-phase electric machine.

9. The method according to claim 1, further comprising: calculating a difference between the predicted phase voltage and the measured voltage; as well as The calculated difference is compared to a threshold value to determine if an imbalance exists in the multi-phase electric machine.

10. The method according to claim 1, further comprising: calculating a difference between the predicted phase voltage and the expected voltage; as well as The calculated difference is compared to a threshold value to determine the health of the rotor of the multi-phase electric machine.

11. The method according to claim 1 , further comprising: calculating a difference between the predicted phase voltage and the measured voltage; as well as The calculated difference is compared to a threshold value to determine early stator winding degradation during operation of the multi-phase electric machine.

12. A method for detecting phase voltages of a multi-phase motor, the method comprising: receiving a set of readings from an auxiliary coil mounted in the motor; calculating a root mean square value of the set of readings from the auxiliary coil installed in the motor; as well as The RMS values ​​are input into a trained machine learning model to obtain predicted phase voltages of the motor.

13. A system for detecting phase voltages of a multi-phase motor, the system comprising: an auxiliary coil installed in the multi-phase motor; as well as The controller is configured as: receiving a set of readings from the auxiliary coil mounted in the motor; converting the set of readings into a plurality of principal components, wherein the principal components are based on eigenvectors and eigenvalues ​​generated based on the set of readings; The plurality of principal components are input into a trained machine learning model to obtain predicted phase voltages of the motor.

14. The system of claim 13, wherein the controller is further configured to provide an alert based on comparing the predicted phase voltage to an expected voltage value.

15. The system of claim 13, wherein the controller is further configured to detect a dominant frequency component of the set of readings.

16. The system of claim 13, wherein the controller trains the machine learning model by: storing the plurality of principal components in a training database; and Labels are provided for the stored principal components and the predicted phase voltages. 17 . The system of claim 16 , wherein the labels provided to the stored principal components and the predicted phase voltages include at least one of an operational load label and a measured voltage label.

18. The system of claim 13, wherein the controller is further configured to: predicting a load connected to the multi-phase electric machine; and comparing a rated load associated with the multi-phase electric machine to the predicted electric machine load; and Based on determining whether the predicted load exceeds the rated load of the multi-phase electric machine, an alert is provided.

19. The system of claim 18, wherein the predicted load is used to calculate energy usage of the multi-phase electric machine.

20. The system of claim 19, wherein the energy usage of the multi-phase electric machine is used to determine electric machine efficiency.

Images