Motor fault prediction method and device, electronic equipment and storage medium

Abstract

The application discloses a motor fault prediction method and device, electronic equipment and storage medium. The method comprises the following steps: obtaining a historical current signal of a target motor in a first time period, performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; for each target component, a target prediction model is used to predict the input target component to obtain a prediction component of the target component in a second time period; a plurality of prediction components are reconstructed to obtain a prediction current signal in the second time period; a standard current signal is obtained, and a target clustering algorithm is used to process the standard current signal and the prediction current signal to obtain a clustering result corresponding to the prediction current signal; and a fault prediction result of the target motor in the second time period is determined based on the clustering result. Based on the technical scheme of the application, the accuracy of motor fault prediction can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer application, and in particular to a motor fault prediction method and device, an electronic device and a storage medium. BACKGROUND

[0002] As a common power device, the motor has the advantages of long service life and good stability, and has been widely used in the fields of aerospace, medical instruments and the like. However, due to factors such as load mutation and harsh working environment, the motor often fails and is damaged. Once the motor is damaged, it will seriously affect the normal operation of the system. Taking a medical surgical robot as an example, the joints of the medical surgical robot are usually driven by motors. When the motor is working normally, the joints of the robot can complete rotation or movement under the driving of the motor, and the combination of the movements of the joints enables the medical surgical robot to complete complex surgical actions. When the motor fails, the movement of the corresponding joint cannot be controlled, thereby affecting the normal work of the surgical robot, affecting the surgical process, and even endangering the patient's life in serious cases. Therefore, how to predict the fault of the motor before the fault occurs has always been a hot issue in the field of fault detection.

[0003] In related technologies, the vibration signal of the motor is usually predicted by a deep network learning model, and the motor fault is predicted by analyzing the predicted vibration signal. However, the accuracy of the prediction result is often unsatisfactory, that is, the accuracy of the current common motor fault prediction method is poor. SUMMARY

[0004] The present application provides a motor fault prediction method, device, electronic device and storage medium to solve the technical problem of poor accuracy of the current common motor fault prediction method.

[0005] According to an aspect of the present application, a motor fault prediction method is provided, wherein the method comprises:

[0006] Obtaining a historical current signal of a target motor in a first time period, and performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component;

[0007] For each target component, the target component is predicted by a target prediction model to obtain a predicted component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which the crossover rate and the mutation rate are associated with the number of genetic iterations;

[0008] The plurality of predicted components are reconstructed to obtain a predicted current signal in the second time period;

[0009] obtain a standard current signal, and process the standard current signal and the predicted current signal through a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal;

[0010] determine a fault prediction result of the target motor in the second time period based on the clustering result.

[0011] According to another aspect of the present application, there is provided a fault prediction device for a motor, wherein the device comprises:

[0012] an empirical mode decomposition module configured to obtain a historical current signal of a target motor in a first time period, and perform empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component;

[0013] a component prediction module configured to, for each target component, predict the input target component through a target prediction model to obtain a predicted component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which a crossover rate and a mutation rate are associated with a genetic iteration number;

[0014] a component reconstruction module configured to reconstruct a plurality of the predicted components to obtain a predicted current signal in the second time period;

[0015] a clustering processing module configured to obtain a standard current signal, and process the standard current signal and the predicted current signal through a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal;

[0016] a fault prediction module configured to determine a fault prediction result of the target motor in the second time period based on the clustering result.

[0017] According to another aspect of the present application, there is provided an electronic device, comprising:

[0018] at least one processor; and

[0019] a memory in communication connection with the at least one processor; wherein

[0020] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the fault prediction method for a motor according to any one of the embodiments of the present application.

[0021] According to another aspect of the present application, there is provided a computer readable storage medium storing computer instructions for causing a processor to implement the motor fault prediction method according to any of the embodiments of the present application when executed.

[0022] The technical solution of the embodiment of the present application comprises the following steps: obtaining a historical current signal of a target motor in a first time period, performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; for each target component, predicting the input target component by a target prediction model to obtain a predicted component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which a crossover rate and a mutation rate are associated with a genetic iteration number; reconstructing a plurality of predicted components to obtain a predicted current signal in the second time period; obtaining a standard current signal, processing the standard current signal and the predicted current signal by a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal; and determining a fault prediction result of the target motor in the second time period based on the clustering result. The present application can improve the accuracy of motor fault prediction, so that appropriate decisions can be made before motor failure occurs, and prevent uncontrollable problems caused by sudden motor failure.

[0023] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0025] Figure 1 is a flowchart of a motor fault prediction method according to an embodiment of the present application;

[0026] Figure 2 is a data example diagram of an empirical mode decomposition method according to an embodiment of the present application;

[0027] Figure 3 is a flowchart of an empirical mode decomposition current signal according to an embodiment of the present application;

[0028] Figure 4is a structural diagram of a long short-term recurrent neural network according to an embodiment of the present application;

[0029] Figure 5 is a flowchart of clustering standard current signals and predicted current signals according to an embodiment of the present application;

[0030] Figure 6 is an example diagram of clustering results according to an embodiment of the present application;

[0031] Figure 7 is a flowchart of a motor fault prediction method according to embodiment two of the present application;

[0032] Figure 8 is a flowchart of training a long short-term recurrent neural network according to an embodiment of the present application;

[0033] Figure 9 is a whole flowchart of a motor fault prediction method according to an embodiment of the present application;

[0034] Figure 10 is a structural schematic diagram of a motor fault prediction device according to embodiment three of the present application;

[0035] Figure 11 is a structural schematic diagram of an electronic device implementing a motor fault prediction method according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to make the personnel in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0037] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] Embodiment one

[0039] Figure 1 A flowchart of a motor fault prediction method is provided for embodiment one of the present application. The embodiment can be applicable to the case of predicting the fault of a motor driving a surgical robot. The method can be executed by a motor fault prediction device, which can be realized in the form of hardware and / or software, and can be configured in a computer. As shown in Figure 1 , the method comprises:

[0040] S101, obtaining a historical current signal of a target motor in a first time period, and performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component.

[0041] The target motor can be understood as an electric energy conversion device. Optionally, in a medical scenario, the target motor can be used to drive a surgical robot. Specifically, the joints of the surgical robot are driven by the target motor to complete rotation or movement, and the actions of the joints are combined to enable the surgical robot to complete complex surgical procedures. In the present embodiment, the target motor can be pre-set according to the scenario requirements, which is not limited specifically herein. Illustratively, the target motor can include a generator or a motor, etc.

[0042] The first time period can be understood as a historical time period relative to a target time point. Optionally, the first time period can be a time period determined based on the target time point and a first preset time interval. The first preset time interval can be 20s, 30s or 50s, etc. The target time point can be a real-time time point.

[0043] The historical current signal can be understood as a historical current signal.

[0044] The empirical mode decomposition (EMD) can be understood as a method of signal decomposition based on the time scale characteristics of the data itself.

[0045] The intrinsic mode components (IMF) and the residual component (RES) can be understood as the decomposition results of the empirical mode decomposition of the current signal (see Figure 2 ).

[0046] Figure 2 is a data example diagram of an empirical mode decomposition method according to the present embodiment.

[0047] wherein, Figure 2The historical current signal is respectively displayed, and six intrinsic mode components (IMF1, IMF2, IMF3, IMF4, IMF5 and IMF6) and a residual component are obtained after the historical current signal is subjected to empirical mode decomposition.

[0048] Figure 3 A flowchart for empirical mode decomposition of a current signal is provided according to an embodiment of the present application. As shown in the figure, x(t) represents the historical current signal, IMF represents the intrinsic mode component, and Res(t) represents the residual component. Figure 3

[0049] Specifically, the flowchart for empirical mode decomposition of a current signal can be as follows:

[0050] 1. Extreme points of x(t) are identified, and the minimum value points and the maximum value points in the historical current signal are located;

[0051] 2. The maximum value points and the minimum value points identified are respectively fitted by using a cubic spline interpolation method, and a first fitting curve E1 and a first fitting curve E2 are sequentially obtained;

[0052] 3. An intermediate function H(t) is calculated based on E1 and E2, and it is determined whether H(t) is an IMF; if H(t) is not an IMF, the extreme points of x(t) are re-identified, and if H(t) is an IMF, a residual term T(t) after IMF decomposition is obtained;

[0053] 4. It is determined whether T(t) is a monotonic function; if T(t) is not a monotonic function, T(t) is taken as x(t), and steps 1-3 are returned to be executed, and if T(t) is a monotonic function, T(t) is taken as Res(t).

[0054] S102. For each target component, the target component input is predicted by a target prediction model to obtain a predicted component of the target component in a second time period.

[0055] The target component can be understood as a component obtained by empirical mode decomposition. The target component can be an intrinsic mode function component or a residual component.

[0056] The target prediction model can be understood as a model with component prediction function. In the embodiment of the present application, the input data of the target prediction model is a target component. The output data of the target prediction model is a predicted component corresponding to the target component.

[0057] ​Optionally, the prediction component includes a first component corresponding to the intrinsic modal function component or a second component corresponding to the residual component. Correspondingly, the target prediction model can include a first prediction model for predicting the first component or a second prediction model for predicting the second component. Further, the first component can include a plurality of sub-components, and correspondingly, the first prediction model can also include a plurality of sub-prediction models.

[0058] The second time period can be understood as a future time period relative to the target time point corresponding to the first time period. Optionally, the second time period can be a time period determined based on the target time point and a second preset time interval. The second preset time interval can be 1s, 3s, or 5s, etc. The second preset time interval is less than the first preset time interval.

[0059] Optionally, the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm. The target genetic algorithm is a genetic algorithm in which the crossover rate and the mutation rate are associated with the number of genetic iterations.

[0060] The training sample set can be understood as a sample used to train the target prediction model.

[0061] The target genetic algorithm can be understood as a genetic algorithm in which the crossover rate and the mutation rate are associated with the number of genetic iterations.

[0062] The long short-term recurrent neural network (LSTM) can be understood as a neural network with the ability to remember long and short-term information.

[0063] The crossover rate can be understood as the ratio of the number of offspring generated by crossing to the number of individuals in the population in each generation of the genetic algorithm. The mutation rate can be understood as the percentage of the number of genes that are mutated in the population of the genetic algorithm in the total number of genes.

[0064] The number of genetic iterations can be understood as the number of real-time iterations of the target genetic algorithm. For example, the number of genetic iterations can be 1, 22, or 55, etc.

[0065] Figure 4 is a structural diagram of a long short-term recurrent neural network according to an embodiment of the present application. As shown in Figure 4 The structure of the long short-term recurrent neural network includes three gating units: a forget gate F, an input gate I, and an output gate O. The three gating units selectively store, update, and forget the structure units, thereby realizing effective transmission of information.

[0066] Wherein, the main role of the forgetting gate is to control the amount of memory information input at the last moment to the current moment memory information, which can be represented as:

[0067] F(t) = σ(W F [h(t-1), x(t)] + b F )

[0068] Wherein, F(t) is the forgetting gate, sigma is the sigmoid function, h(t) is the output of the structure unit at t moment, x(t) is the input of the structure unit at t moment, b F represents the bias value, W F represents the weight matrix.

[0069] Wherein, the expression of the sigmoid function can be: sigma = 1 / (1 + e -x ).

[0070] The main role of the input gate is to control the amount of input information at the current moment, which can be represented as:

[0071] I(t) = sigma(W I [h(t-1), x(t)] + b I )

[0072] Wherein, I(t) is the input gate, sigma is the sigmoid function, h(t) is the output of the structure unit at t moment, x(t) is the input of the structure unit at t moment, b I represents the bias value, W I represents the weight matrix.

[0073] The candidate input at the current moment can be represented as:

[0074]

[0075] Wherein, tanh represents the hyperbolic activation function, h(t) is the output of the structure unit at t moment, x(t) is the input of the structure unit at t moment, b c represents the bias value, W c represents the weight matrix.

[0076] Then the memory information C(t-1) at the last moment controlled by the forgetting gate F and the candidate input information at the current moment controlled by the input gate I can obtain the memory information C(t) at the current moment, which can be represented as:

[0077]

[0078] Wherein, C(t) represents the memory information of the structure unit at t moment, F(t) is the forgetting gate, and I(t) is the input gate.

[0079] The main function of the output gate is to control the amount of memory information at the current time, and its formula is:

[0080] O(t)=σ(W O [h(t-1),x(t)]+b O )

[0081] Wherein, O(t) represents the output gate, sigma is a sigmoid function, h(t) is the output of the structure unit at t time, x(t) is the input of the structure unit at t time, b O represents the bias value, W O represents the weight matrix.

[0082] Wherein, the relationship among h(t), O(t) and C(t) can be: h(t)=O(t) tanh(C(t)).

[0083] The LSTM model is selected as the component prediction model, so that the problem of long-term dependence defect can be avoided when a long time sequence is processed, and the accuracy of component prediction is ensured.

[0084] S103, reconstructing the plurality of predicted components to obtain a predicted current signal of the second time period.

[0085] Wherein, the predicted current signal can be understood as a predicted current signal.

[0086] S104, obtaining a standard current signal, and processing the standard current signal and the predicted current signal through a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal.

[0087] Wherein, the standard current signal can be understood as a standard current signal, and can also be understood as a normal current signal.

[0088] The target clustering algorithm can be understood as an algorithm for dividing a data set into different classes or clusters. Optionally, the target clustering algorithm can be kmeans++.

[0089] The clustering result can be understood as the clustering result of the predicted current signal. Optionally, the clustering result can be the clustering category of each data point corresponding to the predicted current signal.

[0090] Figure 5 A flowchart for clustering a standard current signal and a predicted current signal is provided according to an embodiment of the application. As Figure 5As shown, the normal current signal of the target motor is combined with the reconstructed prediction current signal generated by the GA-LSTM (i.e., the target prediction model), classified by using the kmeans++ clustering algorithm, and the average value and root mean square value of the current are selected as the clustering indicators. If the motor operating state is abnormal, the kmeans++ clustering result of the prediction current signal will deviate from the normal state center. Specifically, the distance between the data points of the current signal and the category center is calculated based on the following formula:

[0091]

[0092] wherein d represents the Euclidean distance, x represents the category center, and y represents the data point.

[0093] Figure 6 is an example diagram of a clustering result provided by an embodiment of the application. As shown, the kmeans++ algorithm can distinguish the standard current signal and the abnormal current signal, so that the target clustering algorithm can accurately identify the prediction current signal in the case of abnormality. Figure 6

[0094] S105, determining a fault prediction result of the target motor in the second time period based on the clustering result.

[0095] wherein the fault prediction result can be understood as a result of predicting whether the target motor has a fault in the second time period. Optionally, the fault prediction result can include a fault and / or no fault.

[0096] The technical scheme of the embodiment of the application, by obtaining the historical current signal of the target motor in the first time period, performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; for each target component, the target prediction model is used to predict the input target component to obtain the prediction component of the target component in the second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, the target genetic algorithm is a genetic algorithm in which the crossover rate and the mutation rate are associated with the number of genetic iterations; reconstructing a plurality of prediction components to obtain a prediction current signal in the second time period; obtaining a standard current signal, and processing the standard current signal and the prediction current signal by using a target clustering algorithm to obtain a clustering result corresponding to the prediction current signal; determining a fault prediction result of the target motor in the second time period based on the clustering result. The application can improve the accuracy of motor fault prediction, so as to make corresponding decisions before the motor fails, and prevent uncontrollable problems caused by sudden motor failure.

[0097] Embodiment two​

[0098] Figure 7 A flowchart of a motor fault prediction method provided for embodiment two of the present application, this embodiment is additionally provided for the above-mentioned embodiment of predicting the target component input by the target prediction model. As shown in the figure, the method comprises: Figure 7

[0099] S201, respectively initializing the long short-term recurrent neural network and the target genetic algorithm.

[0100] 1, initialize LSTM. According to the sample space dimension of the training sample, determine the appropriate network topology structure, determine the number of neurons of the network, and initialize the weight threshold value of each neuron node.

[0101] 2, initialize GA. According to the LSTM topology structure to be optimized, determine the number of GA population individuals, and design appropriate population number and total number of genetic iterations; select the training result of the LSTM model as the fitness function of GA; set the crossover rate and mutation rate of GA as adaptive values that change with the fitness value.

[0102] S202, predicting the input training sample by the long short-term recurrent neural network to obtain a prediction result, and determining a fitness value of the target genetic algorithm based on the prediction result.

[0103] Wherein, the fitness value can be understood as a value representing the fitness of the target genetic algorithm. In the embodiment of the present application, the fitness value can reflect the performance of the target genetic algorithm. Optionally, the closer the fitness value is to the target value, the better the performance of the target genetic algorithm.

[0104] Optionally, the determining of the fitness value of the target genetic algorithm based on the prediction result comprises:

[0105] determining the total number of samples of the training sample set, and determining the fitness value based on the prediction result, the sample label of the training sample and the total number of samples.

[0106] Wherein, the total number of samples can be understood as the total number of training samples in the training sample set. The sample label can be understood as the label of the training sample.

[0107] In the embodiment of the present application, the fitness value can be calculated based on the following formula:

[0108]

[0109] Wherein, Fit represents the fitness value, N represents the total number of samples, y represents the sample label, ​The prediction result is represented.

[0110] S203, in the case where the fitness value does not satisfy the preset condition, respectively updating the crossover rate and the mutation rate of the target genetic algorithm to obtain an updated crossover rate and an updated mutation rate.

[0111] Optionally, the preset condition can include that the fitness value is less than a target value. In the embodiment of the present application, if the fitness value is less than the target value, it indicates that the performance of the target genetic algorithm has reached the standard.

[0112] Optionally, the updating of the crossover rate of the target genetic algorithm to obtain the updated crossover rate includes:

[0113] determining an average fitness corresponding to the training sample set, and in the case where the fitness value exceeds the average fitness, taking a preset first crossover rate as the updated crossover rate; otherwise,

[0114] determining the updated crossover rate based on the first crossover rate, a preset second crossover rate, a current genetic iteration number and a preset total number of genetic iterations;

[0115] wherein the first crossover rate is greater than the second crossover rate.

[0116] wherein the average fitness can be understood as the average fitness value of all the training samples in the training sample set.

[0117] The total number of genetic iterations can be understood as the total number of iterations of the target genetic algorithm. The total number of genetic iterations can be preset according to the scene requirement, which is not limited here. Optionally, the total number of genetic iterations can include 50, 100 or 200, etc.

[0118] Relevant research has proved that the performance of the target genetic algorithm is better when the crossover rate belongs to a first fixed interval. For example, the first fixed interval can be [0.3, 0.8],

[0119] In the embodiment of the present application, the first crossover rate can be the maximum crossover rate corresponding to the first fixed interval, for example, the first crossover rate is 0.8. The second crossover rate can be the minimum crossover rate corresponding to the first fixed interval, for example, the second crossover rate is 0.3.

[0120] In the embodiment of the present application, the updated crossover rate can be calculated based on the following formula:

[0121]

[0122] wherein p c represents the updated crossover rate, pcmax represents the first crossover rate, p cmin represents the second crossover rate, Fit represents the fitness value, Fit mean represents the average fitness, j represents the current genetic iteration number, j max represents the total number of genetic iterations.

[0123] Optionally, the mutation rate of the target genetic algorithm is updated to obtain an updated mutation rate, comprising:

[0124] In the case where the fitness value exceeds the average fitness, a preset first mutation rate is taken as the updated mutation rate; otherwise,

[0125] The updated mutation rate is determined based on the first mutation rate, a preset second mutation rate, the current genetic iteration number, and a preset total number of genetic iterations.

[0126] The first mutation rate is smaller than the second mutation rate.

[0127] It is understood that for the fitness function, the smaller the fitness value of the population, the stronger the adaptability of the population, and the better the genes are more likely to be retained. According to the fitness value, the population is selected, crossed and mutated, and the fundamental purpose is to retain the evolution of good genes and eliminate inferior genes. In this step, the size of the crossover rate and the mutation rate seriously affects the performance of the genetic algorithm. In the traditional genetic algorithm, the crossover rate p c and the mutation rate p m are constants. However, for the initial genetic iteration, a larger crossover rate and a smaller mutation rate help to improve the genes of individuals with poor fitness; in the middle of the genetic iteration, a larger mutation rate helps to improve the local search ability of the algorithm; when the genetic iteration is in the later period, for individuals with strong adaptability, a smaller crossover rate and a smaller mutation rate help to retain individual genes. In summary,

[0128] Relevant research has proved that the performance of the target genetic algorithm is better when the mutation rate belongs to the second fixed interval. For example, the second fixed interval can be [0.001, 0.1],

[0129] In an embodiment of the present application, the first mutation rate can be the minimum mutation rate corresponding to the second fixed interval, and for example, the first mutation rate is 0.001. The second mutation rate can be the maximum mutation rate corresponding to the second fixed interval, and for example, the second mutation rate is 0.1.

[0130] In an embodiment of the present application, the updated mutation rate can be calculated based on the following formula:

[0131]

[0132] wherein p m represents the updated mutation rate, p mmax represents the second mutation rate, p mmin represents the first mutation rate, Fit represents the fitness value, Fit mean represents the average fitness, j represents the current genetic iteration number, j max represents the total number of genetic iterations.

[0133] S204, updating the training sample set based on the updated crossover rate and the updated mutation rate, to obtain an updated training sample set.

[0134] S205, determining target model parameters based on the updated training sample set, updating the model parameters of the long short-term recurrent neural network based on the target model parameters, to obtain an optimized prediction model.

[0135] wherein the target model parameters can be understood as a set of optimized model parameters. Optionally, the target model parameters can include target weights and / or target thresholds.

[0136] The optimized prediction model can be understood as the long short-term recurrent neural network after optimization of the model parameters.

[0137] Optionally, the determining of the target model parameters based on the updated training sample set comprises:

[0138] returning to execute the training of the long short-term recurrent neural network based on the training sample, to obtain a prediction result, and determining the fitness value of the target genetic algorithm based on the prediction result, based on the updated training sample set,

[0139] until the fitness value meets the preset condition, stopping the returning to execute operation and determining the target model parameters based on the fitness value.

[0140] Figure 8 is a flowchart for training a long short-term recurrent neural network according to an embodiment of the present application. As shown in Figure 8 the flowchart for training a long short-term recurrent neural network can be:

[0141] 1. Optimization iteration of GA, after initializing LSTM and GA, the LSTM model is trained by using the sample space (i.e. the training sample), and the training result of the model is obtained; the training result is used as the fitness value of GA, and it is judged whether the value meets the end condition of GA optimization, if the fitness value does not meet the preset requirement, the crossover rate and mutation rate are updated, the population gene of GA is selected and crossed and mutated, so as to generate new individuals, the LSTM is retrained according to the new population individuals, and a new fitness value is obtained; if the GA fitness value meets the set requirement or the iteration number of GA reaches the preset value, the LSTM node weight threshold value of GA optimization at this time is output.

[0142] 2. LSTM verification. The optimized LSTM model is trained in combination with the sample space, and in the model training process, if the training result of the neural network does not meet the expected requirement, the model training is continued; if the expected requirement is met, the corresponding LSTM model is output, which is the optimal LSTM model (i.e. the target prediction model).

[0143] S206. Train the optimization prediction model by using the training sample set before updating to obtain the target prediction model.

[0144] S207. Obtain the historical current signal of the target motor in the first time period, and perform empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component.

[0145] S208. For each target component, the target prediction model is used to predict the input target component to obtain the prediction component of the target component in the second time period.

[0146] S209. The plurality of prediction components are reconstructed to obtain the predicted current signal in the second time period.

[0147] S210. Obtain the standard current signal, and process the standard current signal and the predicted current signal by using the target clustering algorithm to obtain the clustering result corresponding to the predicted current signal.

[0148] S211. Determine the fault prediction result of the target motor in the second time period based on the clustering result.

[0149] The technical scheme of the embodiment of the application is as follows: the long short-term recurrent neural network and the target genetic algorithm are initialized respectively; the long short-term recurrent neural network is used to predict input training samples to obtain a prediction result, and the fitness value of the target genetic algorithm is determined based on the prediction result; in the case that the fitness value does not satisfy the preset condition, the crossover rate and the mutation rate of the target genetic algorithm are updated respectively to obtain an updated crossover rate and an updated mutation rate; the training sample set is updated based on the updated crossover rate and the updated mutation rate to obtain an updated training sample set; the target model parameter is determined based on the updated training sample set, the model parameter of the long short-term recurrent neural network is updated based on the target model parameter to obtain an optimized prediction model; and the target prediction model is obtained by training the optimized prediction model based on the training sample set before the update. The target genetic algorithm based on the update of the crossover rate and the mutation rate with the change of the genetic iteration number is used to optimize the long short-term recurrent neural network, so that the performance of the target prediction model obtained by training can be improved, and the accuracy of component prediction of the target prediction model can be improved.

[0150] Figure 9 A whole flowchart of a motor fault prediction method is provided according to the embodiment of the application.

[0151] As shown in Figure 9 , the whole flow of the motor fault prediction method can be:

[0152] 1. Data acquisition and processing. The historical current signal of the target motor under stable load is acquired. Generally, as the use time of the motor increases, the working performance of the motor decreases, resulting in the gradual increase of the motor current under the same load. The acquired current signal is decomposed by EMD to obtain a plurality of IMF components and a RES component.

[0153] 2. Model establishment and application. The GA-LSTM model (i.e., the target prediction model) is constructed for different IMF components and RES components obtained by decomposition. The GA-LSTM model is used to predict the IMF components and the RES components respectively, and the prediction results are integrated to obtain the predicted current signal of the motor.

[0154] 3. Prediction data reconstruction and classification. The kmeans++ method is used to classify the predicted current signal obtained by the GA-LSTM model and the standard current signal. If there is a large deviation between the kmeans++ classification result of the predicted current signal and the classification result of the standard current signal of the motor, it indicates that the target motor is about to fail.

[0155] Based on the embodiment technical scheme of the present application, motor failure can be predicted, so that corresponding decisions can be made before real failure occurs, and uncontrollable problems caused by sudden motor failure can be prevented.

[0156] Embodiment three

[0157] Figure 10 A structural schematic diagram of a motor failure prediction device provided by the third embodiment of the present application is shown in the figure. Figure 10 As shown in the figure, the device comprises an empirical mode decomposition module 310, a component prediction module 320, a component reconstruction module 330, a clustering processing module 340 and a failure prediction module 350. Among them,

[0158] The empirical mode decomposition module 310 is used to obtain the historical current signal of the target motor in the first time period, and to perform empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; the component prediction module 320 is used to predict the input target component for each target component by a target prediction model to obtain the prediction component of the target component in the second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which the crossover rate and the mutation rate are associated with the number of genetic iterations; the component reconstruction module 330 is used to reconstruct a plurality of prediction components to obtain the prediction current signal in the second time period; the clustering processing module 340 is used to obtain a standard current signal, and to process the standard current signal and the prediction current signal by a target clustering algorithm to obtain the clustering result corresponding to the prediction current signal; and the failure prediction module 350 is used to determine the failure prediction result of the target motor in the second time period based on the clustering result.

[0159] The technical scheme of the embodiment of the application comprises the following steps: obtaining a historical current signal of a target motor in a first time period, performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; for each target component, performing prediction on the input target component by using a target prediction model to obtain a prediction component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which a crossover rate and a mutation rate are associated with a genetic iteration number; reconstructing a plurality of prediction components to obtain a predicted current signal in the second time period; obtaining a standard current signal, processing the standard current signal and the predicted current signal by using a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal; and determining a fault prediction result of the target motor in the second time period based on the clustering result. The application can improve the accuracy of motor fault prediction, so that corresponding decisions can be made before motor failure occurs, and uncontrollable problems caused by sudden motor failure can be prevented.

[0160] Optionally, the target component comprises an intrinsic mode function component or a residual component, the prediction component comprises a first component corresponding to the intrinsic mode function component or a second component corresponding to the residual component, and the target prediction model comprises a first prediction model for predicting the first component or a second prediction model for predicting the second component.

[0161] Optionally, the motor fault prediction device further comprises an initialization module, a fitness value determination module, a crossover and mutation updating module, a sample set updating module, a model parameter updating module, and a target model determination module, wherein,

[0162] The initialization module is configured to initialize the long short-term recurrent neural network and the target genetic algorithm respectively before the target prediction model is used to predict the input target component;

[0163] The fitness value determination module is configured to predict a training sample input by using the long short-term recurrent neural network to obtain a prediction result, and determine a fitness value of the target genetic algorithm based on the prediction result;

[0164] The crossover and mutation updating module is configured to update a crossover rate and a mutation rate of the target genetic algorithm respectively to obtain an updated crossover rate and an updated mutation rate in a case where the fitness value does not satisfy the preset condition;

[0165] The sample set updating module is configured to update the training sample set based on the updated crossover rate and the updated mutation rate to obtain an updated training sample set;

[0166] The model parameter updating module is configured to determine target model parameters based on the updated training sample set, update model parameters of the long-short term recurrent neural network based on the target model parameters, and obtain an optimized prediction model.

[0167] The target model determining module is configured to train the optimized prediction model based on the training sample set before the update to obtain the target prediction model.

[0168] Optionally, the model parameter updating module is specifically configured to:

[0169] return to perform the operation of training the long-short term recurrent neural network based on the updated training sample set to obtain a prediction result and determining a fitness value of the target genetic algorithm based on the prediction result,

[0170] until the fitness value meets the preset condition, stop returning to perform the operation and determine target model parameters based on the fitness value.

[0171] Optionally, the fitness value determining module is specifically configured to:

[0172] determine a total number of samples of the training sample set, and determine the fitness value based on the prediction result, sample labels of the training samples, and the total number of samples.

[0173] Optionally, the crossover and mutation updating module comprises a first crossover rate updating unit and a second crossover rate updating unit, and wherein

[0174] The first crossover rate updating unit is configured to determine an average fitness corresponding to the training sample set, and in a case where the fitness value exceeds the average fitness, determine a preset first crossover rate as an updated crossover rate; otherwise,

[0175] The second crossover rate updating unit is configured to determine an updated crossover rate based on the first crossover rate, a preset second crossover rate, a current genetic iteration number, and a preset total number of genetic iterations.

[0176] The first crossover rate is greater than the second crossover rate.

[0177] Optionally, the crossover and mutation updating module further comprises a first mutation rate updating unit and a second mutation rate updating unit, and wherein

[0178] The first mutation rate updating unit is configured to, in a case where the fitness value exceeds the average fitness, determine a preset first mutation rate as an updated mutation rate; otherwise,

[0179] The second variation rate updating unit is configured to determine an updated variation rate based on the first variation rate, a preset second variation rate, a current genetic iteration number, and a preset total number of genetic iterations.

[0180] The first variation rate is less than the second variation rate.

[0181] The motor fault prediction device provided by the embodiments of the present application can perform the motor fault prediction method provided by any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0182] Embodiment four

[0183] Figure 11 A structural schematic diagram of an electronic device 10 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present application described and / or claimed in this document.

[0184] As shown in Figure 11 The electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11, wherein the memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0185] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunications networks.

[0186] The processor 11 can be various general and / or special purpose processing components having processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the failure prediction method of the electric machine.

[0187] In some embodiments, the failure prediction method of the electric machine can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded onto the RAM 13 and executed by the processor 11, one or more steps of the failure prediction method of the electric machine described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the failure prediction method of the electric machine by any other appropriate means, such as by means of firmware.

[0188] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0189] Computer programs used to implement the methods of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program, when executed, implements the functions / acts specified in the flowcharts and / or block diagrams. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0190] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0191] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0192] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0193] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0194] It should be understood that the various forms of flow shown above can be used to reorder, add or delete steps. For example, each step described in the present application can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solutions of the present application can be achieved, which is not limited herein.

[0195] The above detailed description does not constitute a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method of failure prediction of an electric machine, characterized in that, The method comprises: obtaining a historical current signal of a target motor in a first time period, and performing empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; for each target component, predicting the input target component by a target prediction model to obtain a predicted component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which a crossover rate and a mutation rate are associated with a genetic iteration number; reconstructing a plurality of predicted components to obtain a predicted current signal in the second time period; obtaining a standard current signal, and processing the standard current signal and the predicted current signal by a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal; determining a fault prediction result of the target motor in the second time period based on the clustering result; wherein the target component comprises an intrinsic mode function component or a residual component, the predicted component comprises a first component corresponding to the intrinsic mode function component or a second component corresponding to the residual component, and the target prediction model comprises a first prediction model for predicting the first component or a second prediction model for predicting the second component.

2. The method of claim 1, wherein, Before the prediction of the input target component by the target prediction model, the method further comprises: initializing the long short-term recurrent neural network and the target genetic algorithm, respectively; predicting an input training sample by the long short-term recurrent neural network to obtain a prediction result, and determining a fitness value of the target genetic algorithm based on the prediction result; in a case where the fitness value does not satisfy a preset condition, updating the crossover rate and the mutation rate of the target genetic algorithm to obtain an updated crossover rate and an updated mutation rate; updating the training sample set based on the updated crossover rate and the updated mutation rate to obtain an updated training sample set; determining a target model parameter based on the updated training sample set, updating a model parameter of the long short-term recurrent neural network based on the target model parameter to obtain an optimized prediction model; and training the optimized prediction model by using the training sample set before the update to obtain the target prediction model.

3. The method of claim 2, wherein, The determination of the target model parameter based on the updated training sample set comprises: returning to the operation of training the long short-term recurrent neural network by using the training sample to obtain a prediction result, and determining a fitness value of the target genetic algorithm based on the prediction result based on the updated training sample set, until the fitness value satisfies the preset condition, stopping the returning operation and determining the target model parameter based on the fitness value.

4. The method of claim 2, wherein, The determination of the fitness value of the target genetic algorithm based on the prediction result comprises: determining a total number of samples of the training sample set, and determining the fitness value based on the prediction result, a sample label of the training sample, and the total number of samples.

5. The method of claim 2, wherein, The updating of the crossover rate of the target genetic algorithm to obtain the updated crossover rate comprises: determining an average fitness corresponding to the training sample set, and in a case where the fitness value exceeds the average fitness, taking a preset first crossover rate as an updated crossover rate; otherwise, determining an updated crossover rate based on the first crossover rate, a preset second crossover rate, a current genetic iteration number, and a preset total number of genetic iterations; wherein the first crossover rate is greater than the second crossover rate.

6. The method of claim 5, wherein, The updating of the mutation rate of the target genetic algorithm includes: in a case where the fitness value exceeds the average fitness, taking a preset first mutation rate as an updated mutation rate; otherwise, determining an updated mutation rate based on the first mutation rate, a preset second mutation rate, a current genetic iteration number, and a preset total number of genetic iterations; wherein the first mutation rate is less than the second mutation rate.

7. A failure prediction device of an electric machine characterized by comprising: The method includes: an empirical mode decomposition module configured to obtain a historical current signal of a target motor in a first time period, and perform empirical mode decomposition on the historical current signal to obtain a plurality of intrinsic mode components and a residual component; a component prediction module configured to, for each target component, predict the input target component by a target prediction model to obtain a predicted component of the target component in a second time period, wherein the target prediction model is obtained by training a long short-term recurrent neural network based on a training sample set and a target genetic algorithm, and the target genetic algorithm is a genetic algorithm in which a crossover rate and a mutation rate are associated with a genetic iteration number; a component reconstruction module configured to reconstruct a plurality of the predicted components to obtain a predicted current signal in the second time period; a clustering processing module configured to obtain a standard current signal, and process the standard current signal and the predicted current signal by a target clustering algorithm to obtain a clustering result corresponding to the predicted current signal; a fault prediction module configured to determine a fault prediction result of the target motor in the second time period based on the clustering result; wherein the target component includes an intrinsic mode function component or a residual component, the predicted component includes a first component corresponding to the intrinsic mode function component or a second component corresponding to the residual component, and the target prediction model includes a first prediction model for predicting the first component or a second prediction model for predicting the second component.

8. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the motor fault prediction method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for causing the processor to execute the motor fault prediction method of any one of claims 1-6 when executed.

Images