Valve inner leakage signal feature processing method and system based on XGBoost model

Abstract

The invention provides a valve inner leakage signal feature processing method and system based on an XGBoost model, and relates to the technical field of data processing. According to the method provided by the invention, the importance score of the feature parameters of each type included in the sample is determined according to the times that the feature parameters of different types are used as split nodes in the construction process of the model tree of the neural network model; the number of the feature parameters is deleted for the first time, a first sample set comprising samples of different numbers and types of feature parameters is determined according to importance score sorting, a target feature type is determined according to a first evaluation index of the first sample set, and processing of the valve inner leakage signal features is completed. The method can effectively reduce the types and number of the feature parameters included in the sample, thereby avoiding the overfitting of the neural network model, reducing the model training time, and improving the model training efficiency and precision.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a valve internal leakage signal feature processing method and system based on an XGBoost model. Background Art

[0002] As an important part of the gas pipeline, the safety of the pipeline system has always been the focus of the industry. Once the valve has internal leakage, it will lead to abnormal operation of the gas pipeline and may cause serious environmental pollution, property loss and even casualties. Therefore, it is very important to find and solve the problem of internal leakage of the valve in time. The traditional internal leakage detection method usually monitors the safety of the valve based on parameters such as pressure, temperature and flow. However, this method has the problems of poor accuracy and real-time performance.

[0003] In the related art, a neural network model is usually trained to achieve safety detection of valves through the trained neural network model. However, in the training process of the neural network model, since each training sample has multiple types of characteristic parameters included in different internal leakage signals, and there are many types of characteristic parameters, the trained neural network model is overfitted, and the use of many types of characteristic parameters as training samples will result in a long training time for the neural network model, resulting in low model training efficiency and poor accuracy.

[0004] Therefore, there is an urgent need for a valve internal leakage signal feature processing method based on the XGBoost model to process multiple feature parameters included in the training samples, avoid overfitting of the neural network model, reduce the model training time, and improve the model training efficiency and accuracy. Summary of the invention

[0005] The embodiment of the present invention provides a valve internal leakage signal feature processing method and system based on the XGBoost model, which can quickly and accurately process multiple types of feature parameters included in the training samples, thereby avoiding overfitting of the neural network model, reducing the training time of the neural network model, and thus improving the model training efficiency and accuracy.

[0006] To achieve the above object, the embodiments of the present invention adopt the following technical solutions:

[0007] In a first aspect, a method for processing valve internal leakage signal features based on an XGBoost model is provided, the method comprising: obtaining a sample set, the sample set comprising multiple internal leakage signal samples, each internal leakage signal sample comprising multiple types of feature parameters; normalizing the multiple internal leakage signal samples so that the values ​​of the same type of feature parameters included in the multiple internal leakage signal samples are within a preset threshold range; inputting the multiple internal leakage signal samples into a neural network model for training to obtain an importance score of each type of feature parameter, wherein the importance score of each type of feature parameter is the number of times each type of feature parameter is used as a split node in the process of constructing a model tree of the neural network model; sorting the importance scores of each type of feature parameter from large to small, and determining the first number of feature parameters ranked first as first feature parameters; generating multiple first sample sets, each first sample set comprising multiple first samples, each The first sample includes multiple first feature parameters sorted from large to small according to the importance score, the first samples corresponding to different first sample sets include different types and quantities of first feature parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first feature parameters; multiple first sample sets are input into a neural network model for training to obtain first evaluation indicators corresponding to each first sample set, and the first evaluation indicators include root mean square error, mean absolute percentage error, determination coefficient and training time; a target sample set is determined according to the first evaluation indicators corresponding to each first sample set, the target sample set is one of the multiple first sample sets, the root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to a first threshold, the determination coefficient of the target sample set is greater than or equal to a second threshold, and the training time of the target sample set is less than or equal to a third threshold.

[0008] In a possible implementation of the first aspect, the types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

[0009] In a possible implementation of the first aspect, the method further includes: inputting the sample set into a neural network model configured with different model training parameters to obtain a second evaluation indicator corresponding to each neural network model, the model training parameters including feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation indicators including root mean square error, mean absolute error, mean absolute percentage error and determination coefficient; according to the second evaluation indicators corresponding to each neural network model, determining the model training parameters configured for the target neural network model as the target model training parameters, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to the fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to the fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to the sixth threshold.

[0010] In a possible implementation manner of the first aspect, the method further includes: inputting the target sample set into a target neural network model for training to obtain a trained neural network model.

[0011] In a possible implementation manner of the first aspect, a formula for determining a root mean square error RMSE is:

[0012]

[0013] The formula for determining the mean absolute error MAE is:

[0014]

[0015] The formula for determining the mean absolute percentage error MAPE is:

[0016]

[0017] Coefficient of determination R 2 The formula for determining is:

[0018]

[0019] Where n is the number of samples; y i is the actual value of the i-th sample; is the predicted value of the i-th sample; The average value of the actual values ​​of the multiple samples included in each sample set.

[0020] In a possible implementation manner of the first aspect, the neural network model is an XGBoost model.

[0021] The beneficial effects of the present invention are as follows: the method provided by the present invention determines the importance score of each type of feature parameter included in the sample according to the number of times different types of feature parameters are used as split nodes in the process of constructing the model tree of the neural network model, and then performs an initial reduction in the number of feature parameters based on the importance score without affecting the model accuracy, and determines a first sample set of samples including feature parameters of different numbers and types based on the importance score sorting, determines the target feature type based on the first evaluation index of the first sample set, and completes the processing of the valve internal leakage signal feature, which can effectively reduce the types and quantities of feature parameters included in the sample, thereby avoiding overfitting of the neural network model, reducing the model training time, and thus improving the model training efficiency and accuracy.

[0022] In a second aspect, an embodiment of the present invention provides a valve internal leakage signal feature processing system based on an XGBoost model, the system comprising: an acquisition unit, used to acquire a sample set, the sample set comprising multiple internal leakage signal samples, each internal leakage signal sample comprising multiple types of feature parameters; a processing unit, used to normalize the multiple internal leakage signal samples so that the values ​​of the same type of feature parameters included in the multiple internal leakage signal samples are within a preset threshold range; a training unit, used to input the multiple internal leakage signal samples into a neural network model for training, and obtain an importance score of each type of feature parameter, wherein the importance score of each type of feature parameter is the number of times the feature parameter of each type is used as a split node in the process of constructing a model tree of the neural network model; a determination unit, used to sort the importance scores of the feature parameters of each type from large to small, and determine the first number of feature parameters ranked first as the first feature parameters; a generation unit, used to generate multiple first sample sets, each first sample The set includes multiple first samples, each first sample includes multiple first feature parameters sorted from large to small according to the importance score, the first samples corresponding to different first sample sets include different types and quantities of first feature parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first feature parameters; the training unit is also used to input the multiple first sample sets into the neural network model for training to obtain the first evaluation index corresponding to each first sample set, the first evaluation index includes the root mean square error, the mean absolute percentage error, the determination coefficient and the training time; the determination unit is also used to determine the target sample set according to the first evaluation index corresponding to each first sample set, the target sample set is one of the multiple first sample sets, the root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to the first threshold, the determination coefficient of the target sample set is greater than or equal to the second threshold, and the training time of the target sample set is less than or equal to the third threshold.

[0023] In a possible implementation of the second aspect, the types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

[0024] In a possible implementation of the second aspect, the training unit is further used to input the sample set into a neural network model configured with different model training parameters to obtain a second evaluation indicator corresponding to each neural network model, the model training parameters include feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation indicators include root mean square error, mean absolute error, mean absolute percentage error and determination coefficient; the determination unit is further used to determine the model training parameters configured for the target neural network model as the target model training parameters according to the second evaluation indicators corresponding to each neural network model, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to the fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to the fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to the sixth threshold.

[0025] According to a third aspect, an electronic device is provided, comprising a memory and one or more processors; the memory is coupled to the processor; wherein the memory stores computer program code, the computer program code comprises computer instructions, and when the computer instructions are executed by the processor, the electronic device executes a method as in any implementation of the first aspect.

[0026] According to a fourth aspect, a computer-readable storage medium is provided, comprising computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the method in any implementation of the first aspect.

[0027] According to a fifth aspect, a computer program product is provided. When the computer program product is run on a computer, the computer is enabled to execute the method in any implementation of the first aspect.

[0028] It can be understood that the beneficial effects that can be achieved by the system of the second aspect, the electronic device of the third aspect, the computer-readable storage medium of the fourth aspect, and the computer program product of the fifth aspect provided above can be referred to the beneficial effects in the first aspect and any possible design method thereof, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of the hardware structure of an electronic device shown in an embodiment of the present invention;

[0030] Figure 2 This is a flow chart of a method for processing valve internal leakage signal characteristics based on an XGBoost model according to an embodiment of the present invention;

[0031] Figure 3 A schematic diagram of a first evaluation index of a plurality of first sample sets shown in an embodiment of the present invention;

[0032] Figure 4 A schematic diagram of a second evaluation index of multiple model training parameters shown in an embodiment of the present invention;

[0033] Figure 5 A schematic diagram of another second evaluation index of multiple model training parameters shown in an embodiment of the present invention;

[0034] Figure 6 A schematic diagram of a second evaluation index of a plurality of model training parameters according to an embodiment of the present invention;

[0035] Figure 7 A schematic diagram of a second evaluation index of a plurality of model training parameters according to an embodiment of the present invention;

[0036] Figure 8 A schematic diagram of a cluster analysis result shown in an embodiment of the present invention;

[0037] Fig. 9 The figure is a schematic diagram of the hardware structure of a processing system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The technical solution in the embodiment of the present invention will be described below in conjunction with the accompanying drawings in the embodiment of the present invention. Among them, in the description of the present invention, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; "and / or" in the present invention is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. And, in the description of the present invention, unless otherwise specified, "multiple" means two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items.

[0039] In addition, in order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first", "second", etc. are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not necessarily limit the difference.

[0040] Meanwhile, in the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way for easy understanding.

[0041] As an important part of the gas pipeline, the safety of the pipeline system has always been the focus of the industry. Once the valve has internal leakage, it will lead to abnormal operation of the gas pipeline and may cause serious environmental pollution, property loss and even casualties. Therefore, it is very important to find and solve the problem of internal leakage of the valve in time. The traditional internal leakage detection method usually monitors the safety of the valve based on parameters such as pressure, temperature and flow. However, this method has the problems of poor accuracy and real-time performance.

[0042] In the related art, a neural network model is usually trained to achieve safety detection of valves through the trained neural network model. However, in the training process of the neural network model, since each training sample has multiple types of characteristic parameters included in different internal leakage signals, and there are many types of characteristic parameters, the trained neural network model is overfitted, and the use of many types of characteristic parameters as training samples will result in a long training time for the neural network model, resulting in low model training efficiency and poor accuracy.

[0043] Therefore, there is an urgent need for a valve internal leakage signal feature processing method based on the XGBoost model to process multiple feature parameters included in the training samples, avoid overfitting of the neural network model, reduce the model training time, and improve the model training efficiency and accuracy.

[0044] In view of this, an embodiment of the present invention provides a valve internal leakage signal feature processing method based on the XGBoost model, the method comprising: obtaining a sample set, the sample set comprising multiple internal leakage signal samples, each internal leakage signal sample comprising multiple types of feature parameters; normalizing the multiple internal leakage signal samples so that the values ​​of the same type of feature parameters included in the multiple internal leakage signal samples are within a preset threshold range; inputting the multiple internal leakage signal samples into a neural network model for training to obtain an importance score of each type of feature parameter, wherein the importance score of each type of feature parameter is the number of times each type of feature parameter is used as a split node in the process of constructing a model tree of the neural network model; sorting the importance scores of each type of feature parameter from large to small, and determining the first number of feature parameters ranked first as the first feature parameters; generating multiple first sample sets, each first sample set comprising multiple first samples, Each first sample includes multiple first feature parameters sorted from large to small according to the importance score, the first samples corresponding to different first sample sets include different types and quantities of first feature parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first feature parameters; multiple first sample sets are input into a neural network model for training to obtain first evaluation indicators corresponding to each first sample set, and the first evaluation indicators include root mean square error, mean absolute percentage error, determination coefficient and training time; a target sample set is determined according to the first evaluation indicators corresponding to each first sample set, the target sample set is one of the multiple first sample sets, the root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to a first threshold, the determination coefficient of the target sample set is greater than or equal to a second threshold, and the training time of the target sample set is less than or equal to a third threshold.

[0045] The method provided by the present invention determines the importance score of each type of feature parameter included in the sample according to the number of times different types of feature parameters are used as split nodes in the process of constructing the model tree of the neural network model, and then performs an initial reduction on the number of feature parameters based on the importance score without affecting the model accuracy. The first sample set of samples including feature parameters of different numbers and types is determined based on the importance score sorting, and the target feature type is determined based on the first evaluation index of the first sample set to complete the processing of the valve internal leakage signal feature. The method can effectively reduce the types and quantities of feature parameters included in the sample, thereby avoiding overfitting of the neural network model, reducing the model training time, and thus improving the model training efficiency and accuracy.

[0046] In some embodiments, the method for determining the intersection-and-parallel ratio threshold value of pipeline magnetic flux leakage internal detection provided by the embodiments of the present invention can be executed by an intersection-and-parallel ratio threshold value processing system 100 for pipeline magnetic flux leakage internal detection (hereinafter referred to as processing system 100). As an example, the processing system 100 can be any electronic device 200 with data processing capabilities, such as a general-purpose computer, a personal computer, a laptop computer, a switch or a tablet computer, etc. The specific implementation method of the purchase list processing system 100 is not limited here.

[0047] Figure 1 The hardware structure diagram of the electronic device provided by the embodiment of the present invention is shown. The electronic device 200 includes a processor 210, a memory 220 and a communication interface 230.

[0048] The processor 210 may include one or more processing cores. The processor 210 uses various interfaces and lines to connect various parts in the electronic device 200, and executes various functions and processes data of the electronic device 200 by running or executing instructions, programs, code sets or instruction sets stored in the memory 220, and calling data stored in the memory 220. Optionally, the processor 210 can be implemented in at least one hardware form of a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA).

[0049] The memory 220 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory 220 includes a non-transitory computer-readable storage medium. The memory 220 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 220 may include a program storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a sample acquisition function, a model training function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.

[0050] The communication interface 230 is used to communicate with other devices, equipment or communication networks, such as data storage devices, image processing equipment or Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0051] In physical implementation, the above-mentioned components (such as processor 210, memory 220 and communication interface 230) can be components in the same device (such as a laptop). Alternatively, at least two of the components can be set in the same device, that is, as different components in one device, such as a deployment method similar to devices or components in a distributed system.

[0052] It is to be understood that the structure illustrated in this embodiment does not constitute a specific limitation on the electronic device 200. In other embodiments of the present invention, the electronic device 200 may include more or fewer components than those illustrated, or combine certain components, or separate certain components, or arrange the components differently. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0053] The following describes a valve internal leakage signal feature processing method based on an XGBoost model provided by an embodiment of the present invention in conjunction with the accompanying drawings.

[0054] Figure 2 A flow chart of a valve internal leakage signal feature processing method based on an XGBoost model provided by an embodiment of the present invention. Optionally, the method can be Figure 1 The electronic device 200 shown is executed, that is, executed by the processing system 100. The method may include the following steps:

[0055] S1. Obtain a sample set, where the sample set includes multiple inner leakage signal samples, and each inner leakage signal sample includes multiple types of characteristic parameters.

[0056] In one possible implementation, the types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectral centroid, frequency band energy, harmonic energy, spectral broadening, spectral kurtosis, spectral slope and spectral flatness.

[0057] It can also be understood that each inner leakage signal sample includes a time domain characteristic parameter and a frequency domain characteristic parameter of the inner leakage signal, wherein the time domain characteristic parameters include mean, absolute mean, standard deviation, variance, peak value, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor and peak-to-valley factor, and the frequency domain characteristic parameters include main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

[0058] It should be noted that the types of the above characteristic parameters are only exemplary, and the embodiments of the present invention do not impose any particular restrictions on the specific types and quantities of the characteristic parameters.

[0059] S2. Normalize the multiple inner leakage signal samples so that the values ​​of the same characteristic parameters included in the multiple inner leakage signal samples are within a preset threshold range.

[0060] Specifically, the processing system first deletes the extreme values ​​in the sample set, and then normalizes the data, scaling the values ​​of the same type of feature parameters to the same range to maintain the consistency of the data processing process.

[0061] S3. Input multiple inner leakage signal samples into the neural network model for training to obtain the importance score of each type of feature parameter.

[0062] The importance score of each type of feature parameter is the number of times each type of feature parameter is used as a split node in the process of constructing the model tree of the neural network model.

[0063] Specifically, the importance score (F-score) is an indicator that measures the importance of a feature in model construction, and is usually used to characterize the contribution of a feature parameter to the model's predictive ability. The number of times a feature parameter is used as a split node in the construction of the model tree of a neural network model can also be understood as the frequency of use of a feature parameter in the construction of the model tree of a neural network model. Therefore, the importance score is used to characterize the importance of a feature parameter to the predictive ability of a neural network model.

[0064] Exemplarily, the characteristic parameters mean, absolute mean, standard deviation, variance, peak value, root mean square value, waveform factor, peak factor, impulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness are input into the neural network model, and the importance scores obtained are mean (128), absolute mean (8), standard deviation (39), variance (82), peak value (268), root mean square value (6 7), waveform factor (2), peak factor (130), impulse factor (12), skewness (16), kurtosis factor (18), marginal factor (1), shape factor (4), peak-valley factor (2), main frequency (205), spectrum centroid (165), band energy (82), harmonic energy (32), spectrum broadening (58), spectrum kurtosis (60), spectrum slope (24) and spectrum flatness (14). Thus, the importance score of the peak value is 268 and the importance score of the main frequency is 205. This indicates that the peak value and main frequency contribute more to the accuracy of the prediction ability of the neural network model than other feature parameters, while the importance score of the marginal factor is 1 and the importance score of the peak-valley factor is 2, indicating that the marginal factor and the peak-valley factor contribute less to the accuracy of the prediction ability of the neural network model.

[0065] In a possible implementation, the neural network model is an XGBoost model.

[0066] S4. Sort the importance scores of the feature parameters of each category from large to small, and determine the first number of feature parameters ranked first as the first feature parameters.

[0067] Specifically, the first number is determined according to a preset importance score threshold, and a feature parameter having an importance score greater than the importance score threshold is determined as the first feature parameter.

[0068] Exemplarily, when the importance score threshold is 20, the first characteristic parameter among the characteristic parameters shown in the above example is mean, standard deviation, variance, peak value, root mean square value, peak factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis and spectrum slope. The importance scores of these types of characteristic parameters are greater than the importance score threshold.

[0069] S5. Generate multiple first sample sets, each of which includes multiple first samples, and each of which includes multiple first feature parameters sorted in descending order of importance scores.

[0070] The first samples corresponding to different first sample sets include different types and quantities of first characteristic parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first characteristic parameters.

[0071] Exemplarily, when the first feature parameters are A, B, C and D, the importance scores of A, B, C and D are arranged in descending order, and the processing system generates four first sample sets, namely, first sample set 1, first sample set 2, first sample set 3 and first sample set 4, wherein the first sample set 1 includes a plurality of first samples a, each of which includes first feature parameters A, B, C and D, the first sample set 2 includes a plurality of first samples b, each of which includes first feature parameters A, B and C, the first sample set 3 includes a plurality of first samples c, each of which includes first feature parameters A and B, and the first sample set 4 includes a plurality of first samples d, each of which includes first feature parameter A. It should be noted that the above is only an exemplary description, the number of first sample sets is the same as the number of first feature parameters, and the feature parameters ranked last are deleted in order according to the order of importance scores, such as (A, B, C, D), (A, B, C), (A, B) and (A).

[0072] S6. Input multiple first sample sets into the neural network model for training to obtain a first evaluation index corresponding to each first sample set, where the first evaluation index includes a root mean square error, a mean absolute percentage error, a determination coefficient, and a training time.

[0073] In some embodiments, the formula for determining the root mean square error RMSE is:

[0074]

[0075] The formula for determining the above mean absolute percentage error MAPE is:

[0076]

[0077] Coefficient of determination R 2 The formula for determining is:

[0078]

[0079] Where n is the number of samples; y i is the actual value of the i-th sample; is the predicted value of the i-th sample; The average value of the actual values ​​of the multiple samples included in each sample set.

[0080] For example, see Figure 3 , Figure 3Schematic diagram of a first evaluation index corresponding to a plurality of first sample sets shown in the present invention, (a) shows the root mean square error RMSE of the first sample set corresponding to different numbers of feature parameters, (b) shows the mean absolute percentage error MAPE of the first sample set corresponding to different numbers of feature parameters, (c) shows the determination coefficient R of the first sample set corresponding to different numbers of feature parameters 2 , (d) shows the training time of the first sample set corresponding to different numbers of feature parameters.

[0081] S7. Determine a target sample set according to the first evaluation index corresponding to each first sample set.

[0082] The target sample set is one of the multiple first sample sets. The root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to the first threshold, the determination coefficient of the target sample set is greater than or equal to the second threshold, and the training time of the target sample set is less than or equal to the third threshold.

[0083] Combination Figure 4 , the root mean square error of the first sample set with 9 features is the minimum value among the root mean square errors of multiple first sample sets, and the average absolute percentage error of the first sample set with 9 features is less than or equal to the first threshold, the determination coefficient is greater than or equal to the second threshold, and the training time is less than or equal to the third threshold. Therefore, the processing system determines that the first sample set with 9 features is the target sample set. The types of characteristic parameters of each first sample included in the target sample set are peak value, main frequency, spectrum centroid, peak factor, mean, band energy, variance, root mean square and spectrum kurtosis.

[0084] It can be seen from the above S1-S7 that the method provided by the embodiment of the present invention determines the importance score of each type of feature parameter included in the sample according to the number of times different types of feature parameters are used as split nodes in the process of constructing the model tree of the neural network model, and then performs an initial reduction in the number of feature parameters based on the importance score without affecting the model accuracy, and determines the first sample set of samples including feature parameters of different numbers and types based on the importance score sorting, determines the target feature type based on the first evaluation index of the first sample set, and completes the processing of the valve internal leakage signal feature, which can effectively reduce the types and quantities of feature parameters included in the sample, thereby avoiding overfitting of the neural network model, reducing the model training time, and thus improving the model training efficiency and accuracy.

[0085] In some embodiments, the method provided by the embodiment of the present invention further includes the following steps:

[0086] Input the sample set into a neural network model configured with different model training parameters to obtain a second evaluation index corresponding to each neural network model, where the model training parameters include feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation index includes root mean square error, mean absolute error, mean absolute percentage error and determination coefficient;

[0087] According to the second evaluation index corresponding to each neural network model, the model training parameters configured by the target neural network model are determined as the target model training parameters, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to the fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to the fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to the sixth threshold.

[0088] The formula for determining the above mean absolute error MAE is:

[0089]

[0090] Where n is the number of samples; y i is the actual value of the i-th sample; is the predicted value of the i-th sample; The average of the actual values ​​of the multiple samples included in each sample set.

[0091] The determination method of the root mean square error, the mean absolute percentage error and the determination coefficient is the same as that in the above embodiment and will not be repeated here.

[0092] For details, see Figure 4-Figure 7 , Figure 4 The second evaluation index is obtained by inputting the target sample set into the neural network model configured with different feature sampling ratios. Figure 5 The second evaluation index is obtained by inputting the target sample set into the neural network model configured with different learning rates. Figure 6 The second evaluation index is obtained by inputting the target sample set into the neural network model configured with different maximum depths. Figure 7 The second evaluation index is obtained by inputting the target sample set into the neural network model configured with different sample sampling ratios, which is Figure 4-Figure 7 It can be obtained that the target model training parameters are feature sampling ratio of 0.7, learning rate of 0.2, maximum depth of 4, and sample sampling ratio of 0.6. At this time, the prediction accuracy of the neural network model is high, and it has good generalization performance, which effectively avoids overfitting.

[0093] Optionally, the method provided in the embodiment of the present invention further includes the following steps:

[0094] The target sample set is input into the target neural network model for training to obtain a trained neural network model.

[0095] In order to illustrate the beneficial effects of the present application, combined with the above examples, a cluster analysis is performed on the 9 types of characteristic parameters included in the target sample set. First, the original acoustic emission signal data set is divided into 6 working conditions according to the internal leakage flow, namely 6 clusters: 0-5L / Min, 5L / Min-10L / Min, 10L / Min-15L / Min, 15L / Min-20L / Min, 20L / Min-25L / Min and 25L / Min-30L / Min. Each cluster contains 50 groups of valve internal leakage acoustic emission signals, a total of 300 groups. The clustering results of each characteristic parameter are shown in the figure below. Figure 8 As shown, from Figure 8 It can be seen that the data points of each of the 9 characteristic parameters are divided into different clusters, and the points in each cluster show a certain clustering trend in the characteristic value. Among them, the clustering effect of the peak is the best, the points in each cluster are tightly clustered, and there is also a good degree of separation between clusters, indicating that the clustering algorithm has found the inherent structure of the data on this characteristic parameter. From the clustering effect of the 9 features, it can be seen that the clustering trend of each characteristic parameter has a high degree of correlation with the valve internal leakage flow, which can well distinguish different groups in the signal data, and show different sensitivities for different degrees of internal leakage flow, and different clusters correspond to different internal leakage rate ranges, which can help the establishment of the subsequent valve internal leakage rate quantification model.

[0096] The above mainly introduces the scheme of the embodiment of the present invention from the perspective of the method. It is understandable that in order to realize the above functions, the processing system includes at least one of the hardware structure and software modules corresponding to the execution of each function. It should be easily appreciated by those skilled in the art that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the embodiments of the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of the present invention.

[0097] The embodiment of the present invention can divide the processing system into functional units according to the above method example. For example, the processing system can be divided into functional units corresponding to various functions, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. It should be noted that the division of units in the embodiment of the present invention is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0098] For example, Fig. 9 The hardware structure diagram of a processing system provided by an embodiment of the present invention is shown. The processing system 100 comprises: an acquisition unit 110, which is used to acquire a sample set, the sample set comprises a plurality of inner leakage signal samples, each inner leakage signal sample comprises a plurality of types of characteristic parameters;

[0099] The processing unit 120 is used to perform normalization processing on the multiple inner leakage signal samples so that the values ​​of the same characteristic parameters included in the multiple inner leakage signal samples are within a preset threshold range;

[0100] A training unit 130 is used to input a plurality of inner leakage signal samples into a neural network model for training, and obtain an importance score of each type of feature parameter, wherein the importance score of each type of feature parameter is the number of times each type of feature parameter is used as a split node in the process of constructing a model tree of the neural network model;

[0101] A determining unit 140 is used to sort the importance scores of the feature parameters of each category from large to small, and determine the first number of feature parameters ranked first as first feature parameters;

[0102] A generating unit 150 is used to generate a plurality of first sample sets, each of which includes a plurality of first samples, each of which includes a plurality of first characteristic parameters sorted in descending order according to importance scores, the first samples corresponding to different first sample sets include different types and quantities of first characteristic parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first characteristic parameters;

[0103] The training unit 130 is further used to input the plurality of first sample sets into the neural network model for training, and obtain a first evaluation index corresponding to each first sample set, wherein the first evaluation index includes a root mean square error, a mean absolute percentage error, a determination coefficient, and a training time;

[0104] The determination unit 140 is further configured to determine a target sample set according to the first evaluation index corresponding to each first sample set, the target sample set being one of the multiple first sample sets, the root mean square error of the target sample set being the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set being less than or equal to a first threshold, the determination coefficient of the target sample set being greater than or equal to a second threshold, and the training time of the target sample set being less than or equal to a third threshold;

[0105] The determination unit 140 is further configured to determine the types of the plurality of first characteristic parameters included in the samples of the target sample set as target characteristic types.

[0106] Optionally, the types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

[0107] Optionally, the training unit 130 is also used to input the sample set into a neural network model configured with different model training parameters to obtain a second evaluation indicator corresponding to each neural network model, where the model training parameters include feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation indicators include root mean square error, mean absolute error, mean absolute percentage error and determination coefficient; the determination unit 140 is also used to determine the model training parameters configured for the target neural network model as the target model training parameters according to the second evaluation indicators corresponding to each neural network model, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to the fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to the fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to the sixth threshold.

[0108] It should be understood that the specific description of the above optional manners can refer to the above method embodiments, which will not be repeated here. In addition, the explanation of any of the processing systems 100 provided above and the description of the beneficial effects can refer to the above corresponding method embodiments, which will not be repeated here.

[0109] The embodiment of the present invention further provides a computer-readable storage medium, in which at least one computer instruction is stored, and the at least one computer instruction is loaded and executed by a processor to implement the methods of the above embodiments. For the explanation of the relevant contents and the description of the beneficial effects in any of the above-mentioned computer-readable storage media, reference can be made to the above-mentioned corresponding embodiments, which will not be repeated here.

[0110] The embodiment of the present invention further provides a chip. The chip integrates a control circuit and one or more ports for implementing the functions of the above-mentioned processing system 100. Optionally, the functions supported by the chip can be referred to above and will not be described in detail here.

[0111] Those skilled in the art will appreciate that all or part of the steps of the above embodiments can be implemented by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a random access memory, etc. The above-mentioned processing unit or processor can be a central processing unit, a general-purpose processor, a specific circuit structure (application specific integrated circuit, ASIC), a microprocessor (digital signal processor, DSP), a field programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof.

[0112] The embodiment of the present invention also provides a computer program product including instructions, when the instructions are run on a computer, the computer executes any one of the methods in the above embodiments. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., an SSD), etc.

[0113] It should be noted that the above-mentioned devices for storing computer instructions or computer programs provided in the embodiments of the present invention, such as but not limited to the above-mentioned memory, computer-readable storage medium and communication chip, etc., all have non-transitory. Those skilled in the art should be aware that in one or more of the above-mentioned examples, the functions described in the embodiments of the present invention can be implemented with hardware, software, firmware or any combination thereof. When implemented using software, these functions can be stored in a computer-readable storage medium or transmitted as one or more instructions or codes on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein the communication medium includes any medium that is convenient for transmitting a computer program from one place to another. The storage medium can be any available medium that a general or special-purpose computer can access.

[0114] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A valve internal leakage signal feature processing method based on XGBoost model, characterized in that: The method comprises: Acquire a sample set, the sample set comprising a plurality of inner leakage signal samples, each inner leakage signal sample comprising a plurality of types of characteristic parameters; Normalizing the multiple inner leakage signal samples so that the values ​​of the same characteristic parameters included in the multiple inner leakage signal samples are within a preset threshold range; Inputting the plurality of inner leakage signal samples into a neural network model for training, and obtaining an importance score of each type of characteristic parameter, wherein the importance score of each type of characteristic parameter is the number of times the characteristic parameter of each type is used as a split node in the process of constructing a model tree of the neural network model; The importance scores of the feature parameters of each category are sorted from large to small, and the feature parameters with the first number of the top scores are determined as the first feature parameters; Generate multiple first sample sets, each of which includes multiple first samples, each of which includes multiple first feature parameters sorted in descending order of importance scores, the first samples corresponding to different first sample sets include different types and quantities of first feature parameters, and the first samples corresponding to the same first sample set include the same type and quantity of first feature parameters; Inputting the plurality of first sample sets into a neural network model for training to obtain a first evaluation index corresponding to each first sample set, wherein the first evaluation index includes a root mean square error, a mean absolute percentage error, a determination coefficient, and a training time; A target sample set is determined according to a first evaluation indicator corresponding to each first sample set, the target sample set is one of the multiple first sample sets, the root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to a first threshold, the determination coefficient of the target sample set is greater than or equal to a second threshold, and the training time of the target sample set is less than or equal to a third threshold.

2. The method according to claim 1, characterized in that The types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak value, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

3. The method according to claim 2, characterized in that The method further comprises: Inputting the sample set into a neural network model configured with different model training parameters, obtaining a second evaluation index corresponding to each neural network model, wherein the model training parameters include feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation index includes root mean square error, mean absolute error, mean absolute percentage error and determination coefficient; According to the second evaluation index corresponding to each neural network model, the model training parameters configured by the target neural network model are determined as the target model training parameters, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to the fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to the fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to the sixth threshold.

4. The method according to claim 3, characterized in that: The method further comprises: The target sample set is input into the target neural network model for training to obtain a trained neural network model.

5. The method according to claim 2, characterized in that: The formula for determining the root mean square error RMSE is: The formula for determining the mean absolute error MAE is: The formula for determining the mean absolute percentage error MAPE is: Coefficient of determination R 2 The formula for determining is: Where n is the number of samples; y i is the actual value of the i-th sample; is the predicted value of the i-th sample; The average of the actual values ​​of the multiple samples included in each sample set.

6. The method according to claim 5, characterized in that The neural network model is an XGBoost model.

7. A valve internal leakage signal feature processing system based on XGBoost model, characterized in that: The system comprises: An acquisition unit, configured to acquire a sample set, wherein the sample set includes a plurality of inner leakage signal samples, and each inner leakage signal sample includes a plurality of types of characteristic parameters; a processing unit, configured to perform normalization processing on the plurality of inner leakage signal samples so that the values ​​of the same characteristic parameters included in the plurality of inner leakage signal samples are within a preset threshold range; A training unit, used for inputting the plurality of inner leakage signal samples into a neural network model for training, and obtaining an importance score of each type of characteristic parameter, wherein the importance score of each type of characteristic parameter is the number of times the characteristic parameter of each type is used as a split node in the process of constructing a model tree of the neural network model; A determining unit, configured to sort the importance scores of the feature parameters of each category from large to small, and determine the first number of feature parameters ranked first as first feature parameters; A generating unit, configured to generate a plurality of first sample sets, each of which includes a plurality of first samples, each of which includes a plurality of first characteristic parameters sorted in descending order according to importance scores, the first samples corresponding to different first sample sets include different types and quantities of first characteristic parameters, and the first samples corresponding to the same first sample set include the same types and quantities of first characteristic parameters; The training unit is further used to input the multiple first sample sets into the neural network model for training to obtain a first evaluation index corresponding to each first sample set, wherein the first evaluation index includes a root mean square error, a mean absolute percentage error, a determination coefficient, and a training time; The determination unit is further used to determine a target sample set according to a first evaluation indicator corresponding to each first sample set, the target sample set is one of the multiple first sample sets, the root mean square error of the target sample set is the minimum value of the root mean square errors of the multiple first sample sets, the mean absolute percentage error of the target sample set is less than or equal to a first threshold, the determination coefficient of the target sample set is greater than or equal to a second threshold, and the training time of the target sample set is less than or equal to a third threshold.

8. The system according to claim 7, characterized in that The types of characteristic parameters include mean, absolute mean, standard deviation, variance, peak value, root mean square value, waveform factor, peak factor, pulse factor, skewness, kurtosis factor, marginal factor, shape factor, peak-to-valley factor, main frequency, spectrum centroid, frequency band energy, harmonic energy, spectrum broadening, spectrum kurtosis, spectrum slope and spectrum flatness.

9. The system according to claim 8, characterized in that The training unit is further used to input the sample set into a neural network model configured with different model training parameters to obtain a second evaluation index corresponding to each neural network model, wherein the model training parameters include feature sampling ratio, learning rate, maximum depth and sample sampling ratio, and the second evaluation index includes root mean square error, mean absolute error, mean absolute percentage error and determination coefficient; The determination unit is further used to determine the model training parameters configured for the target neural network model as the target model training parameters according to the second evaluation index corresponding to each neural network model, wherein the root mean square error of the target neural network model is the minimum value of the root mean square errors of multiple neural network models, the mean absolute error of the target neural network model is less than or equal to a fourth threshold, the mean absolute percentage error of the target neural network model is less than or equal to a fifth threshold, and the determination coefficient of the target neural network model is greater than or equal to a sixth threshold.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; Wherein, the processor is configured to execute the instructions to implement the valve internal leakage signal feature processing method based on the XGBoost model as described in any one of claims 1-6.

Images