A method, apparatus and medium for predicting remaining life of an electric machine

Abstract

The application relates to the field of motor life, and discloses a method, device and medium for predicting the residual life of a motor, the method comprising: determining a first frequency domain signal corresponding to partial discharge in the motor; wherein the first frequency domain signal comprises a target characteristic peak corresponding to the partial discharge; determining a first predicted mileage value corresponding to the motor based on the target characteristic peak in the first frequency domain signal; and determining the residual life of the motor based on the first predicted mileage value. The frequency domain signal has strong anti-interference capability, and the target characteristic peak is extremely sensitive to early aging, microscopic defects and other degradation phenomena of motor insulating materials. The scheme combines the frequency domain signal and the target characteristic peak of the frequency domain signal to predict the motor life, can avoid interference misjudgment of the time domain signal and information loss of a single feature, can accurately associate the whole life cycle state of the motor, and can significantly improve the precision and accuracy of residual life prediction.

Description

Technical Field

[0001] This invention relates to the technical field of motor life, and in particular to a method, apparatus and medium for predicting the remaining life of a motor. Background Technology

[0002] Currently, there are many problems and technical shortcomings in the field of motor life prediction. Traditional motor life prediction methods mostly rely on macroscopic parameters such as motor operating time and ambient temperature. These parameters have a weak correlation with the actual aging state of the motor insulation system, making it difficult to accurately reflect the true loss of internal insulation. This results in low accuracy of the prediction results and fails to provide a reliable basis for motor maintenance and replacement.

[0003] Therefore, there is an urgent need to develop a method for predicting motor lifespan in order to overcome the technical bottlenecks of existing technologies, which have poor perception of insulation status and insufficient prediction accuracy. Summary of the Invention

[0004] To overcome the problem of low reliability in motor life measurement in the prior art, the present invention provides a method, apparatus and medium for predicting the remaining life of a motor.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for predicting the remaining life of an electric motor, comprising: Determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein, the first frequency domain signal includes the target characteristic peak corresponding to the partial discharge; Based on the target feature peak in the first frequency domain signal, determine the first predicted mileage value corresponding to the motor; Based on the first predicted mileage value, the remaining lifespan of the motor is determined.

[0006] In one embodiment, determining the first frequency domain signal corresponding to the partial discharge in the motor includes: Determine the first radio frequency current value corresponding to the partial discharge in the motor; The first radio frequency current value is converted into a first frequency domain signal.

[0007] In one embodiment, the first radio frequency current value corresponding to the partial discharge is measured in the following manner: Using the connection point of the U, V, and W three-phase windings of the motor as the neutral point, the neutral point is connected to a grounding element, and a radio frequency current measuring device is connected in series between the neutral point and the grounding element. The first radio frequency current value is obtained by reading the detection value of the radio frequency current measuring device.

[0008] In one embodiment, determining the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal includes: Read multiple first frequency domain data points from the first frequency domain signal; wherein, the first frequency domain data points include frequency and amplitude corresponding to the frequency; Input the plurality of first frequency domain data points into the motor life prediction model; The first predicted mileage value is obtained by identifying the information of the target characteristic peak through the motor life prediction model.

[0009] In one embodiment, the motor life prediction model is trained using the following method: Obtain a training dataset under accelerated aging test conditions; wherein, the training dataset includes a second frequency domain signal corresponding to the partial discharge of the motor, and a cumulative mileage value corresponding to the second frequency domain signal; The second frequency domain data points in the second frequency domain signal are input into the model to be trained until the difference between the second predicted driving value output by the model to be trained and the remaining driving mileage value corresponding to the cumulative driving mileage value is less than a preset threshold, thus obtaining the motor life prediction model.

[0010] In one embodiment, determining the remaining lifespan of the motor based on the first predicted mileage value includes: Based on preset conversion rules, the first predicted mileage value is processed to obtain the second predicted mileage value.

[0011] Secondly, the present invention also provides an apparatus for predicting the remaining life of an electric motor, the apparatus comprising: The first frequency domain signal acquisition module is used to determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein, the first frequency domain signal includes a target characteristic peak corresponding to the partial discharge; The first predicted mileage value acquisition module is used to determine the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal. The motor remaining life acquisition module is used to determine the remaining life of the motor based on the first predicted mileage value.

[0012] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor, when executing the program, implements a method for predicting the remaining life of a motor.

[0013] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer program, which, when executed, implements a method for predicting the remaining lifespan of an electric motor.

[0014] Fifthly, a computer program product includes a computer program that, when executed by a processor, implements the steps of a method for predicting the remaining life of an electric motor.

[0015] The beneficial effects of this invention are: Compared to prediction methods based on time-domain signals, this scheme achieves the following significant technical advantages by employing a frequency-domain analysis approach: Enhance anti-interference capability and improve signal quality: By converting the partial discharge signal to the frequency domain for analysis, it is possible to effectively filter out the strong electromagnetic noise and mechanical vibration inherent in the operation of the motor, providing a high signal-to-noise ratio signal basis for life prediction. Keenly capture early degradation characteristics: Target characteristic peaks in the frequency domain (such as specific resonant frequencies) are extremely sensitive to degradation phenomena such as early aging and microscopic defects in motor insulation materials; minute changes in parameters such as amplitude, center frequency, and half-width can directly and quantitatively map the lifespan consumption state. Achieving accurate and quantitative lifetime prediction: By focusing on target feature peaks that are strongly correlated with lifetime and deriving the "first predicted mileage value" based on their quantification parameters, this method avoids misjudgments caused by the mixing of multiple features in the time domain signal, significantly improves the accuracy and precision of remaining lifetime prediction, and provides accurate data support for predictive maintenance. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for predicting the remaining lifespan of an electric motor in one embodiment of the present invention. Figure 2 This is a flowchart illustrating the steps for determining the first frequency domain signal corresponding to partial discharge in a motor in one embodiment of the present invention. Figure 3 This is a flowchart illustrating the steps for measuring the first radio frequency current value corresponding to partial discharge in one embodiment of the present invention. Figure 4 This is a flowchart illustrating the step of determining the first predicted mileage value of the motor based on the target feature peak in the first frequency domain signal in one embodiment of the present invention. Figure 5 This is a structural block diagram of a virtual device in one embodiment of the present invention; Figure 6 This is an internal structural diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0017] The following embodiments are further explanations and supplements to the present invention and do not constitute any limitation on the present invention.

[0018] The terms "first" and "second" in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising" and any variations thereof are intended to cover non-exclusive protection. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. The term "multiple" in this application can mean at least two, for example, two, three, or more, and is not limited by the embodiments of this application.

[0019] Currently, the mainstream power motors used in new energy vehicles are permanent magnet synchronous motors, permanent magnet brushless DC motors, and asynchronous motors. The following factors can cause partial discharge in these motors: 1. Defects in stator insulation materials: Defects such as bubbles, cracks, and impurities in the insulation material lead to concentrated electric fields and increased voltage, resulting in partial discharge. 2. Structural design problems: Unreasonable motor designs, such as uneven insulation layer thickness or sharp angles in the insulation structure, can lead to uneven electric field distribution and partial discharge. 3. Excessive voltage: The operating voltage of the motor exceeds its rated voltage, increasing the electric field strength experienced by the insulation system and making it prone to partial discharge. 4. Temperature and humidity effects: The thermal expansion coefficient and hygroscopicity of the insulation material can cause stress in the insulation structure during operation, reducing insulation performance and easily triggering partial discharge. 5. Charge accumulation: During motor operation, static charge may accumulate on the surface of the stator insulation. When the charge accumulates to a certain level, partial discharge will occur.

[0020] Partial discharge in motors poses the following hazards: it causes stator insulation aging, leading to the decomposition of insulation materials, structural damage, and accelerated insulation aging; it causes faults, as partial discharge can expand insulation defects, leading to internal motor faults such as insulation breakdown and winding short circuits; it degrades performance, affecting motor operation by increasing noise and reducing efficiency; and it poses safety hazards, as severe partial discharge can lead to motor fires, equipment damage, and other safety accidents.

[0021] Therefore, early detection and monitoring of weak discharges inside the motor, and measurement of partial discharge characteristics, will undoubtedly help in identifying early signs of insulation faults, which is of great significance for ensuring the safe operation of the motor and extending its service life. Thus, this solution predicts the motor's lifespan based on its discharge characteristics.

[0022] In existing technologies, motor life prediction methods mostly rely on macroscopic parameters such as motor running time and ambient temperature. These parameters have a weak correlation with the actual aging state of the motor insulation system, making it difficult to accurately reflect the true loss of the internal insulation of the motor. This results in low accuracy of the prediction results and fails to provide a reliable basis for motor maintenance and replacement.

[0023] To facilitate understanding of the technical solutions provided in the embodiments of this application, the design concept of the embodiments of this application will be introduced first below: Therefore, this application provides a method for predicting the remaining life of a motor, see [link to relevant documentation]. Figure 1 ,include: Step 100: Determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein the first frequency domain signal includes the target characteristic peak corresponding to the partial discharge.

[0024] In this embodiment, frequency domain signal analysis can "disassemble" complex, mixed, and difficult-to-distinguish signals, clearly identifying and tracing specific components directly related to insulation aging. Furthermore, frequency domain signals have strong anti-interference capabilities; for electromagnetic environmental noise, frequency domain analysis suppresses noise through spectrum separation and characteristic frequency extraction. Frequency domain signals can also accurately identify mechanical damage (bearings / rotors) and electrical anomalies (windings / air gaps). Therefore, this embodiment uses the first frequency domain signal to predict motor lifespan.

[0025] In the embodiment, the target characteristic peak refers to the one or several resonance peaks that are most relevant to and sensitive to the aging process of motor insulation in the frequency domain signal spectrum. In the partial discharge frequency domain signal, the specific resonance peak selected as the indicator of motor insulation aging has a strong correlation with the change of quantitative parameters such as amplitude and frequency with the life consumption of the motor, and is the cornerstone of life prediction.

[0026] Step 200: Based on the target feature peak in the first frequency domain signal, determine the first predicted mileage value corresponding to the motor.

[0027] In this embodiment, the first predicted mileage value is determined based on the fact that the target characteristic peak in the first frequency domain signal can well characterize the partial discharge situation.

[0028] When using a model to predict motor life using target characteristic peaks, the model can learn the dynamic transformation laws of parameters such as the position (frequency coordinates), width (half-peak width / full-peak width), amplitude, and number of target characteristic peaks (e.g., amplitude growth rate, peak position shift trend, peak width expansion). This allows the model to establish a mapping relationship between "characteristic parameters - insulation state - remaining life," thereby achieving accurate prediction from frequency domain signal characteristics to motor life, avoiding misjudgments caused by relying on only a single parameter, and improving the robustness of the prediction.

[0029] Among them, the roles of different parameters of the target characteristic peak are: (1) Position: Partial discharge of different insulation defects will excite signals of specific frequencies. The peak position shift can reflect the change of defect type and indirectly reflect the life consumption stage; (2) Width: The peak width is related to the duration and energy distribution of partial discharge. In the early stage of insulation aging, the discharge energy is concentrated and the peak width is narrow. As aging intensifies, the internal defects of insulation increase and the discharge process is dispersed, and the peak width gradually widens. The peak width change rate can quantify the life decay rate; (3) Amplitude: The amplitude directly corresponds to the intensity of partial discharge. The better the insulation performance, the lower the discharge intensity and the smaller the amplitude; when insulation deterioration accelerates (such as the appearance of microcracks), the discharge intensity increases sharply and the amplitude increases significantly. The amplitude threshold can be used as a key indicator for life warning; (4) Quantity (number of characteristic peaks): In the early stage, the insulation defect is single and there are only 1-2 characteristic peaks in the frequency domain. As life is consumed, the defect expands (such as the increase of air gaps and crack bifurcation), which will excite multiple discharge signals of different frequencies. The increase in the number of characteristic peaks can reflect the degree of defect spread.

[0030] Step 300: Determine the remaining lifespan of the motor based on the first predicted mileage value.

[0031] In this embodiment, the first predicted mileage value (the calculated mileage that the motor can still travel) is used to determine the remaining lifespan of the motor. When the motor is manufactured, a standard motor has a total mileage value to indicate that the motor can travel to that total mileage value. However, when each motor is in use, the aging condition of the motor will be different due to different usage scenarios and conditions. Based on the aging condition of each motor, a first predicted mileage value relative to the standard motor can be obtained, which corresponds to the remaining lifespan of the motor.

[0032] In one embodiment, see Figure 2 In step 100, determining the first frequency domain signal corresponding to the partial discharge in the motor includes the following two steps: Step 110: Determine the first radio frequency current value corresponding to the partial discharge in the motor.

[0033] In this embodiment, since frequency domain signals cannot be directly measured and are obtained based on time domain signals, this solution characterizes partial discharge by measuring the first radio frequency current value of the motor because: partial discharge is a rapid movement of charge, which inevitably forms current pulses and generates high-frequency current components in the power and ground lines; current signals are easy to measure, have little impact on the system, and are safe to operate; compared to voltage signals, current signals are less affected by external electromagnetic interference and can more directly capture discharge pulses.

[0034] Step 120: Convert the first radio frequency current value into a first frequency domain signal.

[0035] In this embodiment, converting the first radio frequency current value into a first frequency domain signal is equivalent to converting a time domain signal into a frequency domain signal. Methods that can be used include Fast Fourier Transform (FFT), Discrete Fourier Transform (DFT), or Short-Time Fourier Transform (STFT). The specific transformation process is existing technology and will not be described in detail here.

[0036] In one embodiment, see Figure 3 In step 110, the first radio frequency current value corresponding to the partial discharge is measured in the following manner, including the following steps: In step 111, the connection point of the U, V, and W phase windings of the motor is taken as the neutral point, and the neutral point is connected to the grounding element. A radio frequency current measuring device is connected in series between the neutral point and the grounding element.

[0037] In this embodiment, ideally, the three-phase (UVW) currents of the vehicle motor are balanced, and the potential of the motor's star point Y (neutral point) to ground is 0. In this case, when there is no partial discharge inside the motor, the current on the neutral point's grounding wire is 0. When a partial discharge occurs, only the radio frequency current generated by the arc discharge flows through the neutral point's grounding wire. In actual motor operation, due to the motor manufacturing process, the temperature rise differences between different coils, and the differences in the control of phase currents by the electronic control terminal, it is actually difficult to ensure that the motor maintains three-phase balance under all operating conditions and at all times. At this time, the potential of the motor's neutral point to ground will not be absolutely 0. Even if no partial discharge occurs, a grounding current will flow through the neutral point's grounding wire. At this time, the frequency of this grounding current will be consistent with the frequency of the motor's phase current. Generally, the frequency of the phase current of a vehicle motor is in the order of kJ / kF, which is much lower than the frequency of the current generated by the arc discharge (in the order of mJ / kF). Therefore, by monitoring the spectrum of the discharge current, we can filter out or ignore low-frequency signals and focus on the frequency and amplitude of high-frequency signals. Furthermore, as the main insulation of the motor ages, the discharge characteristic peaks tend to trend towards higher frequencies. Therefore, by detecting the star-point grounding current and analyzing its spectral characteristics, the partial discharge condition and insulation aging status of the motor can be determined. Studies have shown that the closer the energy of the motor stator winding discharge pulse is to the low-frequency range (<100kHz), the better the insulation performance; the closer it is to the high-frequency range (>50MHz), the more severe the insulation aging.

[0038] In this embodiment, the current flowing through the grounding wire of the neutral point is monitored using a radio frequency current measuring device. Specifically, the neutral point is connected to and grounded by a grounding element; the radio frequency current measuring device is a high-frequency current transformer or a high-frequency current clamp, and the grounding element is a grounding resistor, a grounding reactor, or a grounding transformer.

[0039] In step 112, the first radio frequency current value is obtained by reading the detection value of the radio frequency current measuring device.

[0040] In the embodiments, the radio frequency current value is used to characterize the degree of partial discharge of the motor.

[0041] In this embodiment, the radio frequency current is converted into a frequency domain signal by a spectrum analyzer or an oscilloscope with fast Fourier transform capability. The device includes the following two connection and matching usage scenarios: Scenario 1: If a high-frequency current transformer is used to measure the radio frequency current value, then a spectrum analyzer should be used to perform a fast Fourier transform. Scenario 2: If a high-frequency current clamp is used to measure the radio frequency current value, then an oscilloscope with fast Fourier transform function should be used to perform fast Fourier transform.

[0042] In one embodiment, see Figure 4 In step 200, determining the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal includes the following steps: Step 210: Read multiple first frequency domain data points from the first frequency domain signal; wherein, the first frequency domain data points include frequency and amplitude corresponding to the frequency.

[0043] In this embodiment, when predicting the remaining lifespan of the motor, after acquiring the frequency domain signal, a corresponding spectrum is obtained; then, the sampling frequency range and sampling frequency interval on the spectrum are determined, and multiple predetermined sampling frequency points are obtained; based on the spectrum, multiple frequency domain amplitudes corresponding to the multiple predetermined sampling frequency points are obtained.

[0044] In the embodiments, studies have shown that the closer the energy of the motor stator winding discharge pulse is to the low frequency band (<100kHz), the better the insulation performance; the closer it is to the high frequency band (>50MHz), the more severe the insulation aging. Therefore, by transforming the time-domain signal into a frequency-domain signal, when determining the sampling frequency range, signals in the low frequency band (<1MHz) can be ignored, and we only record and save frequency-domain signals with frequencies higher than 1MHz. This method is simpler, but it should be noted that regardless of whether the current transformer or the current clamp used to match the oscilloscope is coupled on the star point Y grounding line, its bandwidth must be wide enough to ensure that high-frequency signals are not filtered out.

[0045] In one embodiment, when the model is used, the sampling frequency range in the frequency domain graph of the frequency domain signal is 10MHz-210MHz, and the sampling frequency interval is 0.5MHz. Therefore, each first frequency domain signal contains 401 sampling frequency points. The corresponding frequency domain amplitude is then obtained from the spectrum graph based on these sampling frequency points. Thus, it includes 401 corresponding sampling frequency points and frequency domain amplitudes; that is, each first frequency domain data point includes one sampling frequency point and a corresponding frequency domain amplitude. It can be understood that whether it is the data for model training or the data obtained when acquiring the corresponding frequency domain signal during model application, the same sampling frequency range and sampling frequency interval can be used, and each corresponding frequency domain signal contains 401 sampling frequency points.

[0046] Step 220: Input the plurality of first frequency domain data points into the motor life prediction model.

[0047] Step 230: Identify the information of the target characteristic peak through the motor life prediction model to obtain the first predicted mileage value.

[0048] In this embodiment, the above multiple first frequency domain data points are all input into the motor life prediction model to obtain the predicted first mileage value (the remaining driving mileage of the motor), thereby determining the remaining life of the motor.

[0049] In one embodiment, the motor life prediction model is trained using the following method: Step 410: Obtain the training dataset under accelerated aging test conditions; wherein the training dataset includes the second frequency domain signal corresponding to the partial discharge of the motor, and the cumulative mileage value corresponding to the second frequency domain signal.

[0050] In this embodiment, the acquisition of the motor training dataset includes collecting training data under accelerated aging conditions, such as a preset total service life of 300,000 kilometers, which is equivalent to the motor running 60,000 kilometers under accelerated aging test conditions. The cumulative mileage value is the actual mileage already driven by the motor.

[0051] In this embodiment, the training dataset is designed so that the motor runs for 60,000 kilometers under accelerated aging test conditions, which is equivalent to 300,000 kilometers under normal operating conditions. The 60,000 kilometers is divided into 60 segments at 1000km intervals (corresponding to 60 sets of data in the training dataset). At the beginning of each segment, the radio frequency current value corresponding to partial discharge of the motor is collected. The signal represented by the radio frequency current value is converted into a frequency domain signal. In the spectrum of the corresponding frequency domain signal, the sampling frequency range is from 10MHz to 210MHz, with a sampling interval of 0.5MHz, and the frequency domain amplitude corresponding to 401 frequency points is collected. That is, each set of data includes the radio frequency current value, the corresponding cumulative mileage value, and the corresponding 401 frequency domain data points (each frequency domain data point includes the frequency and the amplitude corresponding to the frequency). The 60 sets of data in the training dataset are divided into 50 training sets and 10 test sets to train the model.

[0052] Accelerated aging is understood as significantly increasing the "working intensity" of a motor by raising the temperature, increasing the load, and intensifying vibration, causing it to accumulate damage equivalent to long-term normal use in a short period of time. This allows for efficient and rapid verification of product lifespan and reliability. This is a key technology for ensuring product quality under cost and time constraints.

[0053] The principle behind accelerated aging of electric motors is to apply stresses (such as temperature, voltage, and mechanical load) exceeding normal operating conditions in a laboratory environment, thereby accelerating the physicochemical processes inside the motor (such as insulation degradation and mechanical wear), and thus simulating the aging effects of long-term actual operation in a shorter time. The core basis of this method is the Arrhenius equation and the inverse power-law model. The former describes the exponential effect of temperature on the rate of chemical reactions (the aging rate may double for every 10°C increase in temperature), while the latter is applicable to the acceleration of mechanical or voltage stresses (such as increased load leading to accelerated wear). Since the relevant principles and usage are already known, they will not be elaborated upon here.

[0054] In this embodiment, the training dataset of the model is obtained based on accelerated aging test conditions; in another parallel embodiment, the training dataset of the model can also be obtained based on the condition that the motor is running normally.

[0055] In one embodiment, the training dataset includes a second frequency domain signal corresponding to the partial discharge of the motor, and the cumulative mileage value corresponding to the second frequency domain signal.

[0056] In another embodiment, the training dataset includes a second frequency domain signal and a second radio frequency current value corresponding to partial discharge of the motor, as well as the cumulative mileage value corresponding to the second frequency domain signal. In this scheme, the second frequency domain signal is converted from the second radio frequency current value. The purpose of using the second radio frequency current value during training is to provide a multi-task learning strategy, which can significantly improve the accuracy and robustness of lifespan prediction; that is, the second radio frequency current serves as an additional supervisory signal, improving the model's generalization ability. The second radio frequency current can also be used in model training, enabling cross-validation of prediction results and state monitoring. For example, if the current value predicted by the model deviates significantly from the actual measured current value, it may indicate a completely new fault mode that the model has not learned, which can trigger an alarm to remind the user to perform manual inspection, thus increasing the system's reliability.

[0057] Step 420: Input the second frequency domain data points in the second frequency domain signal into the model to be trained until the difference between the second predicted driving value output by the model to be trained and the remaining driving mileage value corresponding to the cumulative driving mileage value is less than a preset threshold, and obtain the motor life prediction model.

[0058] In the embodiment, the training process of the model is characterized; since the second predicted driving value output by the model is the remaining driving range of the motor, it needs to be compared with the remaining driving range value; correspondingly, the sum of the remaining driving range value and the cumulative driving range value is the total mileage of the motor.

[0059] In one embodiment, step 300, determining the remaining lifespan of the motor based on the first predicted mileage value, includes the following steps: Based on preset conversion rules, the first predicted mileage value is processed to obtain the second predicted driving value.

[0060] In this embodiment, since step 410 specifies that the training dataset used is the motor under the "accelerated aging test" condition, the first predicted mileage value obtained corresponds to the converted mileage of the motor under the "accelerated aging test" condition. For example, if the preset total driving life of the motor is 300,000 kilometers, it is equivalent to the motor running 60,000 kilometers under the accelerated aging test condition. Therefore, the mileage obtained by the training data used by the motor life prediction model is the converted mileage under the total mileage of 60,000 kilometers (the first predicted mileage value). Therefore, the first predicted mileage value needs to be processed by the preset conversion rule to obtain the second predicted driving value corresponding to 300,000 kilometers, which is the remaining life of the motor.

[0061] In this embodiment, the model construction and training, corresponding to step 410, includes the following two cases: Example 1: The steps for obtaining the neural network model for predicting motor life include: Obtain the training dataset, which includes multiple sets of data; each set of data contains multiple frequency domain data points, the corresponding true value of radio frequency current, the corresponding cumulative driving mileage value (mileage already driven by the motor), and a preset total motor mileage value (total motor lifespan). An initial model is constructed and configured as a multi-output structure with two output terminals, which are used to output the predicted RF current value and the predicted remaining mileage value, respectively. Model training uses multiple frequency domain amplitudes and total motor mileage as input features, and the true value of radio frequency current and the true value of remaining mileage as training targets to train the initial model, resulting in a trained neural network model for predicting motor life.

[0062] The process of training the initial model is performed iteratively in the following manner: Traversal steps: Traverse multiple sets of data in the training dataset; For each set of data to be iterated over, perform the following sub-steps: (a1) Forward propagation step: Multiple frequency domain data points and the total motor mileage value in the set of data are used as input feature vectors and input into the initial model (for example, the total lifespan value is used as an additional feature node and concatenated with the frequency domain amplitude vector and input into the network); the input feature vectors are propagated forward layer by layer through the input layer and at least one hidden layer of the model, where each neuron in each layer performs a weighted summation of its input and applies a bias, and then generates the output of the neuron through a non-linear activation function; finally, the output layer of the model calculates two output values ​​based on its input, which are used as the radio frequency current prediction value and the remaining mileage prediction value, respectively; (b1) Loss calculation steps: Calculate the first difference between the predicted value of the radio frequency current and the corresponding true value of the radio frequency current in the current data group, and calculate the second difference between the predicted value of the remaining mileage and the corresponding true value of the remaining mileage in the current data group (obtained by subtracting the accumulated mileage from the total motor mileage); Based on the first difference and the second difference, calculate a scalar form of joint loss value according to the predefined joint loss function; (c1) Gradient backpropagation steps: Starting from the output layer, calculate the partial derivatives of the joint loss value with respect to all weight parameters and bias parameters in the model; the calculation of the partial derivatives uses the chain rule, taking the partial derivative of the joint loss value with respect to the input of the output layer as the starting point, and propagating back layer by layer through the hidden layers towards the input layer to calculate the partial derivative of the joint loss value with respect to each parameter in each layer. (d1) Parameter update steps: Based on the partial derivatives calculated in step (c1), the gradient descent optimization algorithm is used to adjust and update the values ​​of all weight parameters and bias parameters in the model in the opposite direction of the partial derivatives and with a step size determined by a predefined learning rate. The process involves iteratively executing the traversal steps until all data sets have been traversed once or multiple times, and the joint loss value is lower than the predetermined convergence threshold or the predetermined maximum number of iterations is reached, thereby obtaining a trained neural network model for predicting motor lifespan.

[0063] For Example 1, the corresponding network structure is a multi-layer feedforward framework: Input layer, which receives a feature vector composed of the signal amplitude corresponding to M frequency domain data points and 1 total motor life value, with M+1 nodes; Hidden layer, each layer contains multiple neurons, which perform nonlinear transformation on the input through an activation function, with 2 hidden layers and 9 neurons in this project; Output layer, which outputs the results of the task, with the number of nodes determined by the number of output targets, with 2 nodes in this project, simultaneously obtaining the predicted value of radio frequency current and the predicted value of remaining mileage.

[0064] Normalize the data, including the normalization formula. and the inverse normalization formula It is used for preprocessing and postprocessing, and does not directly participate in the model training steps. During training, the input features and training target are normalized values.

[0065] During the forward propagation process, data is processed from the input layer through the hidden layers and then passed to the output layer to obtain the predicted value. Each neuron receives the output from the previous layer, performs a weighted summation, and processes it through an activation function, sequentially using the net input function. , And, activation function: , in: x It is an input vector (consisting of two parts, such as multiple frequency domain data points and a total motor mileage value). It is the first input connected i One element, Neuron j Net input, It connects neuron i to neuron i. j The weight, Neuron j The bias, f It is an activation function. Neuron j The output (corresponding to the two prediction values ​​of the output layer, the RF current prediction value and the remaining range prediction value). 。

[0066] After obtaining the predicted values ​​through the above steps, loss calculation is performed using the loss function: , Where L is the joint loss value, This is the predicted value for RF current. This is the actual value of the radio frequency current. This is the predicted value of the remaining mileage. This represents the actual remaining mileage.

[0067] During gradient backpropagation, the output layer error is calculated: , in, For the output layer error term, For predicted values, For the true value, This is the derivative of the activation function.

[0068] Perform hidden layer error calculation: , in, For hidden layer error terms, As weight, For the next layer of error, This is the derivative of the activation function.

[0069] Perform weight gradient calculation: , in, For the weight gradient, For the error term, Enter the activation value.

[0070] Perform bias gradient calculation: , in, For the bias gradient, This is the error term.

[0071] Parameter updates include weight updates: , in, For the updated weights, For the old weight, For learning rate, This represents the weight gradient.

[0072] Perform bias update: , in, For the updated bias, For the old offset, For learning rate, This is the bias gradient.

[0073] Example 2: The steps for obtaining the neural network model for predicting motor life include: Obtain the training dataset, which includes multiple sets of data; each set of data contains multiple frequency domain data points, the corresponding cumulative mileage value (mileage already driven by the motor), and a preset total motor mileage value (total motor lifespan). Build an initial model, configured to include one output for outputting the remaining mileage prediction; Model training uses multiple frequency domain data points as input features and the actual remaining mileage as the training target to supervise the training of the initial model, resulting in a trained neural network model for predicting motor life.

[0074] The process of training the initial model is performed iteratively in the following manner: Traversal steps: Traverse multiple sets of data in the training dataset; For each set of data to be iterated over, perform the following sub-steps: (a2) Forward propagation step: Multiple frequency domain data points and the total motor mileage value in the set of data are used as input feature vectors and input into the initial model (e.g., the total lifespan value is used as an additional feature node and concatenated with the frequency domain amplitude vector into the network); the input feature vector is propagated forward layer by layer through the input layer and at least one hidden layer of the model, where each neuron in each layer performs a weighted summation of its input and applies a bias, and then produces the output of the neuron through a non-linear activation function; finally, the output layer of the model calculates a single output value based on its input as the remaining mileage prediction value; (b2) Loss calculation steps: Calculate the difference between the predicted remaining mileage and the actual remaining mileage in the current data set (obtained by subtracting the accumulated mileage from the total motor mileage); based on this difference, calculate a scalar loss value according to a predefined loss function (such as mean square error MSE); (c2) Gradient backpropagation steps: Starting from the output layer, calculate the partial derivatives of the loss value with respect to all weight parameters and bias parameters in the model; the calculation of partial derivatives uses the chain rule, taking the partial derivative of the loss value with respect to the input of the output layer as the starting point, and propagating back layer by layer through the hidden layers towards the input layer to calculate the partial derivative of the loss value with respect to each parameter in each layer. (d2) Parameter update steps: Based on the partial derivatives calculated in step (c2), the gradient descent optimization algorithm is used to adjust and update the values ​​of all weight parameters and bias parameters in the model in the opposite direction of the partial derivatives and with a step size determined by a predefined learning rate. The process involves iteratively executing traversal steps until all data sets have been traversed once or multiple times, and the loss value is lower than the predetermined convergence threshold or the predetermined maximum number of iterations is reached, thereby obtaining a trained neural network model for predicting motor lifespan.

[0075] The present invention also provides a device for predicting the remaining life of an electric motor, see [link to device]. Figure 5 The device includes a first frequency domain signal acquisition module 510, a first predicted mileage value acquisition module 520, and a motor remaining life acquisition module 530. The first frequency domain signal acquisition module 510 is used to determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein, the first frequency domain signal includes a target characteristic peak corresponding to the partial discharge; The first predicted mileage value acquisition module 520 is used to determine the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal. The motor remaining life acquisition module 530 is used to determine the remaining life of the motor based on the first predicted mileage value.

[0076] In one embodiment, the first frequency domain signal acquisition module 510 is specifically used for: Determine the first radio frequency current value corresponding to the partial discharge in the motor; convert the first radio frequency current value into a first frequency domain signal.

[0077] In one embodiment, the first radio frequency current value corresponding to the partial discharge is measured in the following manner: taking the connection point of the U, V, and W phase windings of the motor as the neutral point, connecting the neutral point to a grounding element and grounding it, and connecting a radio frequency current measuring device in series between the neutral point and the grounding element; obtaining the first radio frequency current value by reading the detection value of the radio frequency current measuring device.

[0078] In one embodiment, the first predicted mileage value acquisition module 520 is specifically used for: Read multiple first frequency domain data points from the first frequency domain signal; wherein, the first frequency domain data points include frequency and amplitude corresponding to the frequency; input the multiple first frequency domain data points into the motor life prediction model; identify the information of the target characteristic peak through the motor life prediction model to obtain the first predicted mileage value.

[0079] In one embodiment, the motor life prediction model is trained using the following method: Obtain a training dataset under accelerated aging test conditions; wherein, the training dataset includes a second frequency domain signal corresponding to the partial discharge of the motor, and a cumulative mileage value corresponding to the second frequency domain signal; The second frequency domain data points in the second frequency domain signal are input into the model to be trained until the difference between the second predicted driving value output by the model to be trained and the remaining driving mileage value corresponding to the cumulative driving mileage value is less than a preset threshold, thus obtaining the motor life prediction model.

[0080] In one embodiment, the motor remaining life acquisition module 530 is specifically used for: Based on preset conversion rules, the first predicted mileage value is processed to obtain the second predicted driving value.

[0081] An electronic device, which may be a server, has an internal structure diagram as shown below. Figure 6 As shown, the electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data for a preset correction factor used to determine the first electrical charge required for a transitional operating condition, and for a preset table used to determine the length of time required for the range extender to experience the transitional operating condition. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for predicting the remaining lifespan of a motor.

[0082] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0083] In one embodiment, an electronic device is provided, including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor, when executing the program, implements the steps of a method for predicting the remaining life of a motor. Determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein, the first frequency domain signal includes the target characteristic peak corresponding to the partial discharge; Based on the target feature peak in the first frequency domain signal, determine the first predicted mileage value corresponding to the motor; Based on the first predicted mileage value, the remaining lifespan of the motor is determined.

[0084] In one embodiment, the processor, when executing a computer program, also performs the following steps: The determination of the first frequency domain signal corresponding to the partial discharge in the motor includes: Determine the first radio frequency current value corresponding to the partial discharge in the motor; The first radio frequency current value is converted into a first frequency domain signal.

[0085] In one embodiment, when the processor executes the computer program, it further performs the following steps: the first radio frequency current value corresponding to the partial discharge is measured by the following method: Using the connection point of the U, V, and W three-phase windings of the motor as the neutral point, the neutral point is connected to a grounding element, and a radio frequency current measuring device is connected in series between the neutral point and the grounding element. The first radio frequency current value is obtained by reading the detection value of the radio frequency current measuring device.

[0086] In one embodiment, when the processor executes the computer program, it further implements the following steps: determining the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal includes: Read multiple first frequency domain data points from the frequency domain signal; wherein, the first frequency domain data points include a frequency and the amplitude corresponding to the frequency; Input the plurality of first frequency domain data points into the motor life prediction model; The first predicted mileage value is obtained by identifying the information of the target characteristic peak through the motor life prediction model.

[0087] In one embodiment, when the processor executes the computer program, it further performs the following steps: the motor life prediction model is trained using the following method: Obtain a training dataset under accelerated aging test conditions; wherein, the training dataset includes a second frequency domain signal corresponding to the partial discharge of the motor, and a cumulative mileage value corresponding to the second frequency domain signal; The second frequency domain data points in the second frequency domain signal are input into the model to be trained until the difference between the second predicted driving value output by the model to be trained and the remaining driving mileage value corresponding to the cumulative driving mileage value is less than a preset threshold, thus obtaining the motor life prediction model.

[0088] In one embodiment, the processor, when executing the computer program, further implements the following step: determining the remaining lifespan of the motor based on the first predicted mileage value, including: Based on preset conversion rules, the first predicted mileage value is processed to obtain the second predicted driving value.

[0089] The present invention also provides a computer-readable storage medium for storing a computer program, which, when executed, implements a method for predicting the remaining lifespan of a motor. The specific execution steps are described in relation to an electronic device and will not be repeated here.

[0090] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of a method for predicting the remaining lifespan of a motor. The specific execution steps are described in relation to an electronic device and will not be repeated here.

[0091] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0092] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this disclosure can be embodied in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the invention can also be implemented as a computer program product contained in one or more computer-readable media, which contains computer-readable program code. Computer-readable storage media can be, for example, but not limited to—electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof.

[0093] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0094] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for predicting the remaining life of an electric motor, characterized in that, include: A first frequency domain signal of partial discharge in the motor is determined; wherein, the first frequency domain signal includes a target characteristic peak corresponding to the partial discharge; Based on the target characteristic peak in the first frequency domain signal, a first predicted mileage value is determined; the target characteristic peak includes the resonance peak in the spectrum of the first frequency domain signal that is associated with the aging process and lifespan consumption of the motor insulation. Determining the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal includes: Read multiple first frequency domain data points from the first frequency domain signal; wherein, the first frequency domain data points include frequency and amplitude corresponding to the frequency; Input the plurality of first frequency domain data points into the motor life prediction model; The first predicted mileage value is obtained by identifying the information of the target characteristic peak through the motor life prediction model. Based on the first predicted mileage value, the remaining lifespan of the motor is determined.

2. The method for predicting the remaining life of an electric motor according to claim 1, characterized in that, The determination of the first frequency domain signal corresponding to the partial discharge in the motor includes: Determine the first radio frequency current value corresponding to the partial discharge in the motor; The first radio frequency current value is converted into a first frequency domain signal.

3. The method for predicting the remaining life of an electric motor according to claim 2, characterized in that, The first radio frequency current value corresponding to the partial discharge is measured in the following way: Using the connection point of the U, V, and W three-phase windings of the motor as the neutral point, the neutral point is connected to a grounding element, and a radio frequency current measuring device is connected in series between the neutral point and the grounding element. The first radio frequency current value is obtained by reading the detection value of the radio frequency current measuring device.

4. The method for predicting the remaining life of an electric motor according to claim 1, characterized in that, The motor life prediction model was trained using the following method: Obtain a training dataset under accelerated aging test conditions; wherein, the training dataset includes a second frequency domain signal corresponding to the partial discharge of the motor, and the cumulative mileage value corresponding to the second frequency domain signal; The second frequency domain data points in the second frequency domain signal are input into the model to be trained until the difference between the second predicted mileage value output by the model to be trained and the remaining mileage value corresponding to the cumulative mileage value is less than a preset threshold, thus obtaining the motor life prediction model.

5. The method for predicting the remaining life of an electric motor according to claim 4, characterized in that, Determining the remaining lifespan of the motor based on the first predicted mileage value includes: Based on preset conversion rules, the first predicted mileage value is processed to obtain the second predicted mileage value.

6. A device for predicting the remaining life of an electric motor, characterized in that, The device includes: The first frequency domain signal acquisition module is used to determine the first frequency domain signal corresponding to the partial discharge in the motor; wherein, the first frequency domain signal includes a target characteristic peak corresponding to the partial discharge; the target characteristic peak includes a resonance peak in the spectrum of the first frequency domain signal that is associated with the aging process and life consumption of the motor insulation; The first predicted mileage value acquisition module is used to determine a first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal; the step of determining the first predicted mileage value corresponding to the motor based on the target feature peak in the first frequency domain signal includes: Read multiple first frequency domain data points from the first frequency domain signal; wherein, the first frequency domain data points include frequency and amplitude corresponding to the frequency; Input the plurality of first frequency domain data points into the motor life prediction model; The first predicted mileage value is obtained by identifying the information of the target characteristic peak through the motor life prediction model. The motor remaining life acquisition module is used to determine the remaining life of the motor based on the first predicted mileage value.

7. An electronic device comprising a memory, a processor, and a program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the method for predicting the remaining life of an electric motor as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed, is used to implement the method for predicting the remaining life of an electric motor as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting the remaining life of an electric motor as described in any one of claims 1 to 5.

Images