Fan abnormal sound intelligent discrimination detection method and system based on support vector machine

Abstract

The invention relates to an intelligent fan abnormal sound distinguishing and detecting method and system based on a support vector machine. The method comprises the steps that a fan sound signal and a fan rotating speed signal are collected; identifying the fan sound signal and intercepting a stable section sound signal; a weighting processing is carried out on the sound signal in the stable section so as to obtain sound features reflecting loudness perception of human ears; the fan abnormal sound feature extraction and screening feature extraction analysis method for the sound signals subjected to A weighting processing comprises the steps that a frequency range is divided into 1 / 12 octave of a finer interval, 15 constant bandwidth energy for inspecting the sound energy condition in a fixed bandwidth and an average frequency spectrum for eliminating noise introduced due to instantaneous change; and inputting the extracted and screened fan abnormal sound feature data into a fan abnormal sound intelligent discrimination model of a support vector machine so as to carry out classification prediction on the normal sound and the abnormal sound of the fan. Intelligent detection of the running state of the fan is achieved.

Description

Technical Field

[0001] The invention relates to a method and system for intelligently distinguishing and detecting abnormal noise of a fan based on a support vector machine, and belongs to the field of fan detection. Background Art

[0002] In the manufacturing and quality control process of fans, abnormal noise detection of fans is a key link to ensure product quality and performance. The traditional manual listening detection method has been widely used for a long time. The inspectors rely on their experience and hearing to judge whether the fan has abnormal noise. However, this manual method has many limitations. On the one hand, manual detection is highly dependent on the professional skills, mental state and experience level of the inspectors. It is difficult for different inspectors to maintain completely consistent judgment standards, which can easily lead to subjective differences and misjudgments in the test results. On the other hand, with the continuous expansion of the scale of fan production and the increasing requirements for production efficiency, the low efficiency of manual detection has gradually become a bottleneck restricting the optimization of production processes and the improvement of production capacity. It cannot meet the needs of accurate and efficient detection of abnormal noise of fans in large-scale and rapid production environments, and the traceability is poor. Summary of the invention

[0003] The present invention provides a method and system for intelligently distinguishing and detecting abnormal noise of a fan based on a support vector machine, aiming to solve at least one of the technical problems existing in the prior art.

[0004] The technical solution of the present invention relates to a method and system for intelligently distinguishing and detecting abnormal fan noise based on a support vector machine. The method according to the present invention comprises the following steps:

[0005] S100, collecting fan sound signals and fan speed signals;

[0006] S200, identifying the fan sound signal and intercepting the steady-state sound signal; performing A-weighting processing on the steady-state sound signal to obtain a sound feature that reflects the loudness perception of the human ear;

[0007] S300, extracting and screening abnormal fan sound features from the sound signal after A-weighting processing;

[0008] Among them, the analysis methods for extracting fan abnormal sound features include the 1 / 12 octave analysis method that divides the frequency range into finer intervals, the 15 constant bandwidth energy analysis method that examines the sound energy within a fixed bandwidth, and the average spectrum analysis method used to eliminate the noise introduced by instantaneous changes;

[0009] S400, the extracted and filtered fan abnormal sound feature data is input into the fan abnormal sound intelligent discrimination model of the support vector machine to classify and predict the normal sound and abnormal sound of the fan.

[0010] Further, in step S200:

[0011] The A-weighting process attenuates low-frequency sounds and enhances high-frequency sounds, wherein the frequency response function H of the A-weighting is A (f) is expressed as follows:

[0012]

[0013] Wherein, f represents frequency.

[0014] Further, in step S300,

[0015] The center frequency f of the 1 / 12 octave band ci According to a specific geometric series distribution, it is expressed as follows:

[0016] f ci =f0·2 i / 12

[0017] Where f0 is the initial frequency; i is the octave number;

[0018] The frequency range of each octave is determined by the upper frequency f li and the lower frequency f ui Decide, which is expressed as follows:

[0019] f li =f ci 2 -1 / 24

[0020] f ui =f ci 2 1 / 24

[0021] In the formula, the upper and lower limit frequencies are based on the center frequency;

[0022] Wherein, the octave frequency bandwidth is expressed as follows:

[0023] BW i =f ui -f li ;

[0024] Among them, the energy E i The integral of the signal power spectrum density X(f) is expressed as follows:

[0025]

[0026] Or, energy E i The discrete spectrum is represented as follows:

[0027]

[0028] Where P(fk)=|X(fk)|2 is the power value of the spectrum; Δf = fs / N is the spectrum resolution, and N is the number of points of the FFT.

[0029] Further, in the step S300,

[0030] The 15-constant bandwidth energy is set as the energy between the frequency range f0 to f0 + 15 Hz, where f0 is the center frequency of the frequency band, and the energy Ef0 of the frequency band is calculated by accumulating the power of the frequency points within this frequency band;

[0031] Among them, according to the spectrum power P(f) = ∣X(f)∣ 2 , the energy is expressed as follows:

[0032]

[0033] In the formula, Δf is the frequency resolution of the spectrum.

[0034] Further, in the step S300,

[0035] The acquisition of the average spectrum of the signal includes the following steps:

[0036] Window the given signal x(t), which is expressed as follows:

[0037] x w [n] = x[n] · w[n]

[0038] In the formula, n is the window serial number, and w[n] represents the window function; x w [n] represents the windowed signal;

[0039] Perform Fourier transform on each windowed signal to obtain the spectrum S k (f) of each window, where k is the window number and f is the frequency; the amplitude of the S k (f) spectrum is expressed as follows:

[0040]

[0041] In the formula, L is the window length, x w [n] is the windowed signal data, and f is the frequency range 0 ≤ f < Fs, where Fs represents the sampling frequency;

[0042] The average spectrum of the signal is expressed as follows:

[0043]

[0044] In the formula, N is the total number of time windows; S k (f) is the spectrum of the kth time window.

[0045] Furthermore, the fan abnormal noise intelligent discrimination model of the support vector machine is a nonlinear vector machine model, and uses a Gaussian kernel function as the kernel function; the hyperparameters of the model include a penalty factor that controls the model's tolerance for misclassified samples in the training data, and kernel function parameters used to determine the range of action of the Gaussian kernel function.

[0046] Further, the step S400 includes:

[0047] S410, using Lagrangian mathematical function to expand and transform the model, the calculation of the optimal hyperplane is expressed as follows:

[0048]

[0049] In the formula, margin represents the classification interval; ω represents the spatial weight; b represents the regression bias term; y i represents the sample label, x i represents sample characteristics, i represents the i-th sample, and N represents the total number of samples;

[0050] S420, introducing the Lagrangian function L(ω, b, a) into the inequality, which is expressed as follows:

[0051]

[0052] Where a represents the Lagrange multiplier; a i represents the Lagrange multiplier of the i-th sample;

[0053] S430, introducing slack variables into the model, which are expressed as follows:

[0054]

[0055] In the formula, ξ i represents the slack variable;

[0056] After introducing the slack variables, the objective function of the nonlinear model is expressed as follows:

[0057]

[0058] Where C is the penalty factor.

[0059] Further, in step S400,

[0060] The basic kernel mathematical function is introduced into the model, and its classification function f(x) is expressed as follows:

[0061]

[0062] In the formula, K(x i ,x) represents the basis kernel function; b* Indicates the classification limit.

[0063] The technical solution of the present invention also relates to a computer-readable storage medium on which program instructions are stored, and the above-mentioned method is implemented when the program instructions are executed by a processor.

[0064] The technical solution of the present invention also relates to a fan abnormal noise intelligent discrimination and detection system based on a support vector machine, wherein the system comprises a computer device, and the computer device comprises the above-mentioned computer-readable storage medium.

[0065] The beneficial effects of the present invention are as follows:

[0066] In order to overcome the shortcomings of traditional manual listening detection, the present invention proposes a fan abnormal sound intelligent discrimination detection method and system based on support vector machine. The machine learning algorithm of support vector machine (SVM) is used in the fan abnormal sound detection. The present invention analyzes the sound signal collected when the fan is running, uses professional signal processing technology to extract key information that can effectively characterize the fan sound characteristics, and uses the extracted features as input data of the SVM model, and uses the trained SVM model to automatically predict and classify the subsequent fan data. The present invention constructs the SVM model input by extracting the abnormal sound features, accurately extracts the abnormal sound features to help SVM focus on the essence of the abnormal sound, has the advantages of strong anti-interference and reduced misjudgment, and uses SVM to automatically predict and classify the fan data to determine the abnormal sound, realize data-driven objective judgment, get rid of manual dependence and reduce human errors, effectively improve detection efficiency, and realize data traceability and quality control enhancement, can fully record the detection data including the original sound signal, abnormal sound features, prediction results and time, and can trace the quality of a specific fan or batch, trace back to locate the root cause when there is a problem, and assist in quality improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a basic flow chart of the method according to the present invention.

[0068] Figure 2 Detailed description of the invention is given in Figure 4. It is a flow chart of model training and optimization according to the method of the present invention. DETAILED DESCRIPTION

[0069] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in combination with the embodiments and drawings to fully understand the purpose, scheme and effect of the present invention.

[0070] It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to another feature, or it may be indirectly fixed or connected to another feature. The singular forms "a", "said" and "the" used herein are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art. The terms used in this specification are intended only to describe specific embodiments and are not intended to limit the invention. The term "and / or" used herein includes any combination of one or more of the related listed items.

[0071] It should be understood that, although the term first, second, third etc. may be adopted to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish the same type of elements from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as" etc.) provided herein is only intended to better illustrate embodiments of the present invention, and unless otherwise required, the scope of the present invention will not be limited.

[0072] Reference Figure 1 to Figure 2 In some embodiments, the fan abnormal noise intelligent discrimination detection method based on support vector machine according to the present invention at least includes the following steps:

[0073] S100, collecting fan sound signals and fan speed signals;

[0074] S200, identifying the fan sound signal and intercepting the steady-state sound signal; performing A-weighting processing on the steady-state sound signal to obtain a sound feature that reflects the loudness perception of the human ear;

[0075] S300, extracting and screening abnormal fan sound features from the sound signal after A-weighting processing;

[0076] Among them, the analysis methods for extracting fan abnormal sound features include the 1 / 12 octave analysis method that divides the frequency range into finer intervals, the 15 constant bandwidth energy analysis method that examines the sound energy within a fixed bandwidth, and the average spectrum analysis method used to eliminate the noise introduced by instantaneous changes;

[0077] S400, the extracted and filtered fan abnormal sound feature data is input into the fan abnormal sound intelligent discrimination model of the support vector machine to classify and predict the normal sound and abnormal sound of the fan.

[0078] The support vector machine-based intelligent discrimination detection method and system for fan abnormal noise of the present invention use a machine learning algorithm that introduces a support vector machine in the detection of fan abnormal noise, extracts key information that can effectively characterize the fan sound characteristics and uses it as input data for the SVM model, and uses the trained SVM model to automatically predict and classify subsequent fan data to determine abnormal noise, thereby achieving data-driven objective judgment.

[0079] In some embodiments, when collecting fan sound signals, the present invention places the fan in an existing soundproof box to reduce the interference of environmental noise, and uses a high-sensitivity microphone to collect sound signals during operation. The microphone is fixedly installed perpendicular to the fan, 5 cm above the center of the fan, and points to the center of the fan. The system sampling frequency is set to 48kHz to ensure that the high-frequency components of the signal are fully captured, and the analog signal is converted into a digital signal through a high-precision A / D conversion module to ensure the integrity and accuracy of the data. Furthermore, the system also synchronously collects fan speed pulse signals, where 2 pulses are recorded per revolution, so as to achieve precise alignment of the sound signal with the mechanical state of the fan in subsequent analysis. After the system completes the collection of the fan sound signal, the collected data is transmitted to the data preprocessing unit for further processing.

[0080] In some embodiments, the present invention identifies and intercepts a sound signal segment of the fan in a stable operating state from the collected fan sound signal. It should be noted that during the start and stop process of the fan, the sound signal will be affected by a variety of unstable factors, such as starting current shock, speed change, etc. The present invention uses a signal of a stable speed segment to better reflect the sound characteristics of the fan during normal operation, which is beneficial to subsequent abnormal sound detection.

[0081] In some embodiments, the present invention performs A-weighting processing on the obtained fan sound signal in the stable section, so that the processed signal is more consistent with the human ear's subjective perception of sound loudness, which helps to extract sound features that are more practical and relevant. It should be noted that the A-weighting processing method adopted by the present invention can simulate the human ear's perception of the loudness of sounds of different frequencies and weight them. Specifically, based on the equal loudness curve, in acoustic measurement, by weighting the different frequency components of the sound signal, the measurement results are more consistent with the subjective perception of the human ear. Furthermore, the A-weighting processing method of the present invention attenuates low-frequency sounds to a certain extent and appropriately enhances high-frequency sounds, thereby reflecting the characteristics of the human ear in loudness perception.

[0082] Specifically, the A-weighted frequency response function H of the present invention is A (f) is expressed as follows:

[0083]

[0084] Wherein, f represents frequency, and its unit is Hz.

[0085] In some embodiments, the present invention extracts and screens the fan abnormal sound features of the sound signal after A-weighting processing. Specifically, in the detection of fan abnormal sound, feature extraction and screening play a vital role. The fan abnormal sound feature extraction of the present invention mainly includes 1 / 12 octave, 15 constant bandwidth energy, average spectrum, sound pressure level and loudness.

[0086] In an application embodiment, for the 1 / 12 octave in the feature extraction of fan abnormal sound, it should be noted that the 1 / 12 octave of the present invention is an analysis method that divides the frequency range into finer intervals. In the detection of fan abnormal sound, the 1 / 12 octave divides the audible frequency range (usually 20Hz-20kHz) into smaller frequency bands. Compared with the traditional octave analysis, the method of the present invention can more finely reveal the characteristic changes of sound at different frequencies. This is very important for detecting fan abnormal sound, because the abnormal sound of the fan may appear in a specific narrow frequency range. The present invention uses 1 / 12 octave to more accurately locate these abnormal frequency areas.

[0087] Specifically, the center frequency f of the 1 / 12 octave band is ci According to a specific geometric series distribution, it is expressed as follows:

[0088] f ci =f0·2 i / 12

[0089] Wherein, f0 is the initial frequency, and in the specific embodiment of the present invention, f0 is set to 20 Hz; i is the serial number of the octave.

[0090] The frequency range of each octave is determined by the upper and lower frequency limits f li and f ui Decide, which is expressed as follows:

[0091] f li =f ci 2 -1 / 24

[0092] f ui =f ci 2 1 / 24

[0093] In the formula, the upper and lower limit frequencies are based on the center frequency and are proportional to 2 ±1 / 24 Calculated to ensure that the range width of each frequency band is symmetrical on a logarithmic scale. The octave frequency bandwidth is expressed as follows:

[0094] BW i =f ui -f li

[0095] Energy is usually the integral (or accumulation in the case of discrete sampling) of the signal's power spectral density (PSD):

[0096]

[0097] Furthermore, the energy is represented using a discrete spectrum (the power spectrum P(fk) calculated by FFT) as follows:

[0098]

[0099] Where P(fk)=|X(fk)| 2 is the power value of the spectrum; Δf=fs / N is the spectrum resolution, and N is the number of FFT points.

[0100] In an application embodiment, for the 15-Hz constant bandwidth energy in the feature extraction of fan abnormal sound, it should be noted that the 15-Hz constant bandwidth energy analysis is to examine the energy of the sound within a frequency interval with a fixed bandwidth of 15 Hz. For fan abnormal sound detection, using the above fixed bandwidth analysis method can more accurately capture the energy changes within a specific frequency range. Different faults or abnormal conditions of the fan may cause energy accumulation or changes within a specific 15 Hz bandwidth interval.

[0101] Specifically, the 15 constant bandwidth energy is set as the energy in the frequency range from f0 to f0+15 Hz, where f0 is the center frequency of the frequency band, and the energy Ef0 of the frequency band can be calculated by accumulating the power of the frequency points in the frequency band.

[0102] Where, spectral power P(f) = |X(f)| 2 , the energy calculation formula is as follows:

[0103]

[0104] Where Δf is the frequency resolution of the spectrum, that is, the interval between each point in the spectrum. If the spectrum is calculated by FFT (Fast Fourier Transform), then Δf is determined by the sampling rate and the number of FFT points, that is, Δf = fs / N.

[0105] In an application embodiment, regarding the average spectrum in the extraction of fan abnormal sound features, it should be noted that in fan abnormal sound detection, calculating the average spectrum of the signal is an important step in analyzing the frequency characteristics of the signal. The average spectrum can eliminate the noise introduced by instantaneous changes, obtain more stable and accurate spectrum characteristics, and effectively reveal the abnormal frequencies that may exist during the operation of the fan. Compared with traditional manual detection methods, the automated method based on spectrum analysis adopted by the present invention can more accurately identify fan abnormal sounds, especially when noise sources with relatively concentrated frequencies appear.

[0106] Specifically, windowing is performed on the given signal x(t), with the window length set to 0.2 s and the overlapping ratio set to 80%, to ensure the continuity of the spectrum and at the same time reduce the spectrum leakage caused by signal truncation.

[0107] x w [n] = x[n] · w[n]

[0108] In the formula, n is the window serial number, and w[n] is the window function.

[0109] Perform Fourier transform on each windowed signal to obtain the spectrum S k (f) of each window, where k is the window number and f is the frequency. The amplitude calculation formula of the spectrum is:

[0110]

[0111] In the formula, L is the window length, x w [n] is the signal data after windowing, and f is the frequency range of 0 ≤ f < Fs, where Fs is the sampling frequency.

[0112]

[0113] In the formula, N is the total number of time windows; S k (f) is the spectrum of the kth time window.

[0114] In an application embodiment, for other abnormal sound feature indicators in the extraction of fan abnormal sound features, the principle of the abnormal sound feature indicators is shown in Table I. It should be noted that due to the large number of extracted features, not all features can effectively distinguish normal fan sounds from abnormal abnormal sounds, and some features have small changes in the two types of samples or lack significant regularity. In order to ensure that the input features of the model have a high ability to distinguish and reduce the potential interference of redundant information on model performance, feature screening becomes a key link. Based on rich practical experience and in-depth understanding of the laws of fan abnormal sound characteristics, the present invention systematically observes and analyzes the extracted features, and preferentially retains features that show significant differences between normal samples and abnormal samples. On the contrary, features with low discrimination, that is, features with no obvious changes in normal and abnormal states or highly overlapping numerical distributions, will be eliminated, thereby optimizing the input dimension of the model and improving classification performance.

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122] Table I

[0123] In some embodiments, the present invention inputs the extracted and screened fan abnormal sound feature data into a support vector machine (SVM) model. Before input, the feature data is organized into a format suitable for SVM model processing, and the feature set is labeled according to the artificial listening results as the target label of the model. It can be understood that the SVM model is a binary classification model based on statistical learning theory, which searches for an optimal hyperplane in the feature space, separates samples of different categories as much as possible, and maximizes the minimum distance between the two categories of samples and the hyperplane. For the problem of fan abnormal sound detection, the present invention regards the normal sound samples and abnormal sound samples of the fan as two different types of samples, and learns their distribution patterns in the feature space through the SVM model, thereby realizing the classification and discrimination of new samples.

[0124] Specifically, in order to ensure that the SVM model can classify nonlinear data, the present invention first uses the Lagrangian mathematical function to expand and transform the SVM model, and the calculation formula of the optimal hyperplane is shown in formula (1).

[0125]

[0126] In formula (1), margin represents the classification interval; ω represents the spatial weight; b represents the regression bias term; y i represents the sample label, x i represents sample characteristics, i represents the i-th sample, and N represents the total number of samples;

[0127] Then, the Lagrangian function is introduced into the inequality, and its calculation formula is shown in formula (2).

[0128]

[0129] In formula (2), a represents the Lagrange multiplier; a i represents the Lagrange multiplier of the ith sample.

[0130] In order to classify nonlinear data, the slack variable is introduced into the SVM model, and its calculation formula is shown in formula (3).

[0131]

[0132] In formula (3), ξ i represents the slack variable.

[0133] After introducing the slack variables, the objective function calculation formula of the nonlinear SVM model is shown in formula (4).

[0134]

[0135] In formula (4), C is the penalty factor.

[0136] In addition, in order to ensure the linear inseparable data processing capability and global search performance of the SVM model, the base kernel mathematical function is also applied to the SVM model. At this time, the classification function calculation formula is shown in formula (5).

[0137]

[0138] In formula (5), K(x i ,x) represents the basis kernel function; b * Indicates the classification limit.

[0139] It can be understood that from the calculation formula of the above SVM model, the classification performance of SVM is directly affected by the kernel function parameters and the penalty factor parameters, so it is particularly important to accurately select the kernel function type and the penalty factor parameter value. Furthermore, the penalty factor C is a user-set parameter, and the penalty factor C is used to adjust the influence of the slack variable penalty term. When C takes a larger value, the optimization problem is more concerned with classification accuracy, and the interval width can be sacrificed, that is, the interval can be narrower, so that fewer classification errors are made; when C takes a smaller value, the interval is wider, there are more sample data in the interval, and the classification errors can be more.

[0140] It should be noted that common base kernel functions include Gaussian kernel function, polynomial kernel function, linear kernel function, etc., among which different base kernel functions are specifically shown in Table II below.

[0141]

[0142] Table II

[0143] In some embodiments, the present invention trains and optimizes the fan abnormal noise intelligent discrimination model of the support vector machine (SVM). In the fan abnormal noise detection, the training of the SVM (support vector machine) model is the core step of the entire detection process, and its purpose is to build a classifier that can effectively distinguish normal fan sounds from abnormal noises. Figure 2 , the model training and optimization of the present invention includes the following steps:

[0144] In order to improve the generalization ability and classification performance of the support vector machine (SVM) model in fan abnormal sound detection, the original data is first normalized. Through linear normalization technology, the abnormal sound feature values ​​of the fan can be mapped to a unified interval. It should be noted that the normalization processing of the present invention is to avoid uneven distribution of model weights due to differences in feature value scales, while improving the training stability and optimization efficiency of the SVM model based on distance calculation. At the same time, the normalization processing of the present invention can reduce the impact of noise, enhance the model's ability to learn the essential laws of the data, and provide consistent input features for subsequent training and prediction.

[0145] After normalization, the data set is randomly shuffled and divided into a training set and a test set at a ratio of 80% and 20%. The training set is used to train the model parameters, and the test set is used to adjust and optimize the model parameters and evaluate the final performance.

[0146] Then, for the selection of a suitable SVM model type, the present invention adopts a nonlinear SVM model and selects a Gaussian kernel function (RBF kernel) as the kernel function. It should be noted that the RBF kernel function can map data to a high-dimensional space, realize linear separability in the space, and has good adaptability to the complex and nonlinearly distributed sound features in fan abnormal sound detection. The expression of the Gaussian kernel function is as follows:

[0147]

[0148] Furthermore, during the model training process, two important hyperparameters need to be set, namely the penalty factor C and the kernel function parameter σ. The penalty factor C controls the model's tolerance for misclassified samples in the training data. A larger C value will make the model pay more attention to reducing the training error, while a smaller C value can enhance the generalization ability of the model. The kernel function parameter σ determines the scope of the RBF kernel. A larger γ value corresponds to a more complex decision boundary, while a smaller σ value makes the model smoother.

[0149] After completing the initial parameter setting, the SVM model is constructed using the training set data, and the performance indicators of the model, including the confusion matrix and accuracy, are calculated using the test set. The confusion matrix can clearly show the classification results of the model, including the number of correctly classified and misclassified samples, while the accuracy is the core evaluation indicator used to measure the overall classification performance of the model. If the accuracy of the test set is low or there are obvious misclassified samples in the confusion matrix, the model needs to be optimized by adjusting the values ​​of C and σ. For example, when the model is overfitting, the value of C can be reduced to enhance the model's generalization ability for unknown data; when the decision boundary of the model is too simple, the value of σ can be appropriately increased to improve the nonlinear mapping ability of the kernel function.

[0150] By adjusting parameters multiple times and repeating training and testing, we finally determined a set of parameter configurations that can achieve high accuracy and good generalization ability on the test set as the final model. The trained SVM model can then be used in the actual fan noise detection task to ensure that it has accurate and reliable classification performance on unknown data.

[0151] It can be understood that after obtaining the optimal parameters, the trained support vector machine (SVM) model together with its corresponding parameters (such as penalty factor C, kernel function parameter σ, etc.) will be saved to ensure that the model can be directly loaded and used in subsequent detection processes, avoiding time and resource consumption caused by repeated training.

[0152] In actual detection tasks, the newly collected fan sound data must first be subjected to standardized preprocessing and feature extraction to generate a data feature format consistent with the training phase, thereby ensuring a high degree of consistency between the input features and the model's expected format. After feature extraction, the extracted features are directly input into the saved SVM model. Based on its support vector, kernel function, and optimized parameters, the model of the present invention performs nonlinear mapping on the input data and calculates its position in a high-dimensional space. Through the decision function of the model, it quickly determines whether the input sample belongs to the normal fan sound or abnormal sound category, thereby achieving accurate classification prediction of the fan.

[0153] The fan abnormal noise intelligent discrimination detection method and system based on support vector machine of the embodiment of the present invention are verified by blind side. Specifically, after the support vector machine (SVM) model training is completed and saved, three independent blind test verifications are performed to comprehensively evaluate the classification ability and generalization performance of the model in practical applications. Among them, the blind test data is completely independent of the training set and the test set, and strictly simulates the real detection scene to ensure the scientificity and objectivity of the evaluation results.

[0154] Furthermore, in each blind test, the newly collected fan sound data is first preprocessed and feature extracted to generate a feature format consistent with the training stage. Subsequently, these features are input into the trained SVM model for classification prediction. The results of the three blind tests are shown in Tables 1, 2, and 3. According to the table data, the constructed SVM model not only has a high classification accuracy, but also shows good robustness and generalization ability, and can effectively cope with complex scenarios in actual industrial detection. The successful verification of the blind test results further proves the scientificity and rationality of the feature selection and parameter optimization strategies used in the model design, and provides a reference for the practical application of the model in industrial fan abnormal sound detection.

[0155]

[0156] Table 1

[0157]

[0158] It should be appreciated that the method steps in the embodiments of the present invention can be implemented or implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer readable memory. The method can use standard programming techniques. Each program can be implemented in a high-level process or object-oriented programming language to communicate with a computer system. However, if necessary, the program can be implemented in an assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed ASIC for this purpose.

[0159] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0160] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, an RSM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention can also include the computer itself.

[0161] The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on the display.

[0162] The above is only a preferred embodiment of the present invention. The present invention is not limited to the above implementation. As long as the technical effect of the present invention is achieved by the same means, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical scheme and / or implementation method may have various modifications and changes.

Claims

1. Intelligent identification and detection method of abnormal fan noise, characterized in that: The method comprises the following steps: S100, collecting fan sound signals and fan speed signals; S200, identifying the fan sound signal and intercepting the steady-state sound signal; performing A-weighting processing on the steady-state sound signal to obtain a sound feature that reflects the loudness perception of the human ear; S300, extracting and screening abnormal fan sound features from the sound signal after A-weighting processing; Among them, the analysis methods for extracting fan abnormal sound features include the 1 / 12 octave analysis method that divides the frequency range into finer intervals, the 15 constant bandwidth energy analysis method that examines the sound energy within a fixed bandwidth, and the average spectrum analysis method used to eliminate the noise introduced by instantaneous changes; S400, the extracted and filtered fan abnormal sound feature data is input into the fan abnormal sound intelligent discrimination model of the support vector machine to classify and predict the normal sound and abnormal sound of the fan.

2. The method according to claim 1, characterized in that In step S200: The A-weighting process attenuates low-frequency sounds and enhances high-frequency sounds, wherein the frequency response function H of the A-weighting is A (f) is expressed as follows: Wherein, f represents frequency.

3. The method according to claim 1, characterized in that In the step S300, The center frequency f of the 1 / 12 octave band ci According to a specific geometric series distribution, it is expressed as follows: f ci =f0·2 i / 12 Where f0 is the initial frequency; i is the octave number; The frequency range of each octave is determined by the upper frequency f li and the lower frequency f ui Decide, which is expressed as follows: f li =f ci ·2 -1 / 24 f ui =f ci ·2 1 / 24 In the formula, the upper and lower limit frequencies are based on the center frequency; Wherein, the octave frequency bandwidth is expressed as follows: BW i =f ui -f li ; Among them, the energy E i The integral of the signal power spectrum density X(f) is expressed as follows: Or, energy E i The discrete spectrum is represented as follows: Where P(fk)=|X(fk)| 2 is the power value of the spectrum; Δf=fs / N is the spectrum resolution, and N is the number of FFT points.

4. The method according to claim 1, characterized in that: In the step S300, 15 Constant bandwidth energy is set as the energy in the frequency range f0 to f0+15 Hz, where f0 is the center frequency of the band. The energy Ef0 of the band is calculated by accumulating the power of the frequency points in the band. Among them, according to the spectrum power P(f)=|X(f)| 2 , energy E f0 It is expressed as follows: Where Δf is the frequency resolution of the spectrum.

5. The method according to claim 1, characterized in that In the step S300, The average spectrum acquisition of the signal includes the following steps: The given signal x(t) is windowed, which is expressed as follows: x w [n]=x[n]·w[n] Where n is the window number, w[n] represents the window function; x w [b] indicates the windowed signal; Perform Fourier transform on each windowed signal to obtain the spectrum S of each window k (f), where k is the window number and f is the frequency; S k (f) The amplitude of the spectrum is expressed as follows: where L is the window length, and x w [n] is the windowed signal data, f is the frequency range 0 ≤ f < Fs, where Fs represents the sampling frequency; Average spectrum of the signal It is expressed as follows: Where N is the total number of time windows; S k (f) is the spectrum of the kth time window.

6. The method according to claim 1, characterized in that The fan abnormal noise intelligent discrimination model of the support vector machine is a nonlinear vector machine model, and adopts a Gaussian kernel function as a kernel function; the hyperparameters of the model include a penalty factor for controlling the model's tolerance for misclassified samples in the training data, and a kernel function parameter for determining the range of action of the Gaussian kernel function.

7. The method according to claim 1, characterized in that The step S400 includes: S410, using Lagrangian mathematical function to expand and transform the model, the calculation of the optimal hyperplane is expressed as follows: In the formula, margin represents the classification interval; ω represents the spatial weight; b represents the regression bias term; y i represents the sample label, x i represents sample characteristics, i represents the i-th sample, and N represents the total number of samples; S420, introducing the Lagrangian function L(ω, b, a) into the inequality, which is expressed as follows: Where a represents the Lagrange multiplier; a i represents the Lagrange multiplier of the i-th sample; S430, introducing slack variables into the model, which are expressed as follows: In the formula, ξ i represents the slack variable; After introducing the slack variables, the objective function of the nonlinear model is expressed as follows: Where C is the penalty factor.

8. The method according to claim 7, characterized in that In the step S400, The basic kernel mathematical function is introduced into the model, and its classification function f(x) is expressed as follows: In the formula, K(x i ,x) represents the basis kernel function; b * Indicates the classification limit.

9. A computer-readable storage medium having program instructions stored thereon, wherein the program instructions are executed by a processor to implement the method according to any one of claims 1 to 8.

10. Fan abnormal sound intelligent identification and detection system, characterized by: include: A computer device comprising a computer readable storage medium according to claim 9.

Images