A data-driven based fault diagnosis method for radio frequency circuit board

Abstract

The application discloses a kind of data-driven based on radio frequency circuit board fault diagnosis method, by pre-determined fault, and input excitation signal obtains multiple groups of time domain data, then the time domain data is doubly filtered to obtain frequency domain data, then the frequency domain data is extracted, and a fault type of fault feature is obtained, then for all fault types, fault type is regarded as label, corresponding fault feature is regarded as training sample, input to classifier, train the classifier until convergence, finally, for a piece of radio frequency circuit board, the same data processing is used, and its fault feature is sent into the trained classifier that is trained to converge, to obtain fault type, complete fault diagnosis.The application does not need to add additional internal measuring point, only through the input and output port of radio frequency circuit board carries out fault diagnosis, by applying excitation to radio frequency circuit board and collecting radio frequency output signal data, these data are analyzed using machine learning model, so that the fault type of radio frequency circuit board is accurately identified.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of fault diagnosis, and more specifically relates to a data-driven radio frequency circuit board fault diagnosis method which does not need to add additional measuring points, but uses the input and output ports of the radio frequency circuit board for fault diagnosis. BACKGROUND

[0002] Radio frequency circuit boards are widely used in communication, radar, medical equipment and automotive electronics, etc. The internal structure mainly includes key components such as radio frequency amplifiers, π attenuators and filters. Due to the high operating frequency, large power and complex electromagnetic field environment of radio frequency circuit boards, they are more prone to component aging, damage, and nonlinear and uncertain faults than traditional analog circuits. If these faults are not isolated and located in time, it may lead to the overall failure of the radio frequency circuit board. Therefore, it is particularly important to establish a fault diagnosis method for radio frequency circuit boards.

[0003] Current radio frequency circuit board fault diagnosis methods usually include disassembling the circuit board to expose the internal structure, and directly measuring the voltage, current and waveform information at both ends of the components to assess the aging or damage of the components. However, due to the high frequency characteristics of radio frequency circuits, adding measuring points directly inside the radio frequency circuit board may affect the impedance and electromagnetic field of the circuit. In addition, this method is not suitable for large-scale fault diagnosis and requires high professional skills of technicians.

[0004] Therefore, it is urgent to develop a non-invasive radio frequency circuit board fault diagnosis method that does not require adding measuring points inside the radio frequency circuit board, and can utilize the input and output ports of the radio frequency circuit board to achieve automated fault diagnosis without the need for professional operation. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and provide a data-driven radio frequency circuit board fault diagnosis method for batch automated fault detection of radio frequency circuit boards without adding additional internal measuring points, and only through the input and output ports of the radio frequency circuit board for fault diagnosis. Specifically, by applying excitation to the radio frequency circuit board and collecting radio frequency output signal data, a machine learning model is used to analyze these data to accurately identify the fault type of the radio frequency circuit board.

[0006] To achieve the above-mentioned application purpose, the data-driven radio frequency circuit board fault diagnosis method of the present application comprises the following steps:

[0007] (1) Model the radio frequency circuit board, select important components for monitoring, and introduce a fault model on the radio frequency circuit board, and introduce a predetermined fault on the radio frequency circuit board according to the fault model, i.e. select a fault type;

[0008] (2) For the selected fault type, select the sweep signal as the excitation signal f in (t) Inject into the input end of the RF circuit board and collect the time domain signal f from the output end of the RF circuit board out (t) Obtain the time domain data f(n) and repeat the above process to obtain multiple sets of time domain data f(n) to form a systematic data set, where t is time and n is the sample point number sequence;

[0009] (3) Double filter processing is performed on the collected set of time domain data f(n)

[0010] 3.1) Construct a pure noise time domain signal s with the same length as the time domain data f(n) extended (n)

[0011] 3.1.1) Extract the envelope signal from the time domain data f(n)

[0012] a) Calculate the discrete Fourier transform to obtain the frequency domain data F(k):

[0013]

[0014] Where N is the length of the discrete Fourier transform and k represents the frequency point;

[0015] b) Calculate the frequency response of the Hilbert transform

[0016]

[0017] H(k) is the transformation matrix;

[0018] c) Calculate the signal after Hilbert transform

[0019]

[0020] d) Calculate the envelope A(n):

[0021]

[0022] 3.1.2) Extract the pure noise data s(n)

[0023] Determine a threshold a, for the data points in the time domain data f(n) that satisfy A(n) < a are determined as pure noise data, thus extracting the pure noise data s(n), n1 < n < n2, 0 < n1 < n2 < n0, where n0 is the sample point number of the time domain data f(n);

[0024] 3.1.3) Generate a pure noise time domain signal s with the same length extended (n)

[0025] The pure noise data s(n) is periodically extended to generate a pure noise time-domain signal s of equal length. extended (n):

[0026] s extended (n)=s n ((nmodT)+n1)

[0027] Where n = 0, 1, 2, ..., n0-1;

[0028] 3.2) Adaptive Filtering

[0029] a) For a pure noise time-domain signal s extended (n) is filtered to obtain noise data.

[0030]

[0031] Where w(n) is the coefficient vector of the adaptive filter, w(n) = [w0(n), w1(n), ..., w M-1 (n)] T M is the order of the adaptive filter, w m (n), m=0,1,2,…M-1 are the coefficients of the adaptive filter, s extended (n)=[s extended (n),s extended (n-1),…,s extended (n-M+1)] T ;

[0032] b) Calculate the error signal e(n):

[0033]

[0034] c) Employing the Least Mean Square Error (LMS) criterion algorithm, the mean square value of the error signal is minimized by iteratively adjusting the adaptive filter coefficients:

[0035] J(n)=E[e 2 (n)]

[0036] Where E[·] represents the expected value, and the coefficients of the adaptive filter are iteratively adjusted as follows:

[0037] w(n+1)=w(n)+μs extended (n)e(n)

[0038] Where μ is the step size factor, which controls the rate and stability of the adaptive filter coefficient update;

[0039] d) the error signal e(n) is used as the adaptive filtered time domain data

[0040] 3.3) Fourier transform

[0041] the adaptive filtered time domain data is Fourier transformed to obtain frequency domain data

[0042]

[0043] where N is the length of the discrete Fourier transform and k represents the frequency point;

[0044] 3.4) Sliding maximum filter

[0045] the frequency domain data is subjected to a sliding maximum filter, for each frequency point k, the sliding window contains W frequency domain samples, the output of the maximum filter is the frequency domain data is the maximum value in the window:

[0046]

[0047] where the value range of k needs to ensure that k-m is within the effective frequency range, i.e. 0≤k-m<N;

[0048] (4) the frequency domain data is subjected to feature extraction,

[0049] 4.1) Extract key indicators describing the frequency domain waveform as sample features

[0050] 4.2) Extract features representing the differences between different fault types, i.e. extract inter-sample features;

[0051] 4.2.1) Construct a "base spectrum map" for each fault type

[0052] a) Calculate the mean and residual

[0053] First, randomly select L pieces of frequency domain data after secondary filtering from this fault type The mean of a single piece of frequency domain data is denoted as x i , i = 1, 2, …, L, and the mean of the means of the L pieces of frequency domain data is denoted as The residual v i of the mean of the i-th piece of frequency domain data x i is:

[0054]

[0055] b) Calculate the standard deviation σ using the Bessel formula x:

[0056]

[0057] c) using the Ljite test to remove gross errors in the frequency domain data, that is, if there is a certain frequency domain data residual absolute value greater than 3 times the standard deviation, that is, |v i |>3σ x , this strip of frequency domain data is a gross error, and is removed, and steps a) is returned to reselect, calculate and remove until there is no gross error in the L strips of frequency domain data selected randomly after the second filtering;

[0058] d) taking the mean value of the mean value of the L strips of frequency domain data without gross errors as the "base frequency spectrum" of this fault type;

[0059] 4.2.2) Extract the Pearson correlation coefficient: for the "base frequency spectrum" of the cth fault type, denoted as and calculate the Pearson correlation coefficient with all the "base frequency spectra" of the fault types , to obtain C correlation features r cj , j = 1, 2, …, C:

[0060]

[0061] where x ci is the ith strip of the L strips of frequency domain data corresponding to the cth fault type without gross errors, and x ji is the jth strip of the L strips of frequency domain data corresponding to the jth fault type without gross errors;

[0062] The C correlation features r cj are the inter-sample features of the cth fault type;

[0063] (5) Splice the sample features and inter-sample features of the cth fault type to obtain the fault features of the cth fault type. For all fault types, take the fault type as the label and the corresponding fault feature as the training sample, input it into the classifier, and train the classifier until it converges.

[0064] (6) Fault diagnosis

[0065] For a piece of RF circuit board, select a sweep signal as the excitation signal f in (t) is injected into the input end of the RF circuit board, and the time domain signal f out(t) to obtain time domain data f(n), repeat the above process to obtain multiple sets of time domain data f(n) to form a systematic data set, then process according to the method of steps (3), 4.1), 4.2.1) to obtain sample features and "base frequency spectrum", and compare it with the "base frequency spectrum" of all fault types Pursue the Pearson correlation coefficient calculation to obtain C correlation features, then splice the sample features and the sample features to obtain the fault features, and send them into the trained classifier of step (5) to obtain the fault type, and complete the fault diagnosis.

[0066] The application purpose of the application is achieved as follows:

[0067] The data-driven RF circuit board fault diagnosis method of the application obtains multiple sets of time domain data by predefining faults and inputting excitation signals, then performs double filtering processing on the time domain data to obtain frequency domain data, then performs feature extraction on the frequency domain data to obtain fault features of a fault type, then takes the fault type as a label and the corresponding fault features as training samples for all fault types, inputs them into a classifier, trains the classifier until it converges, finally, adopts the same data processing on an RF circuit board to obtain its fault features and inputs them into the trained classifier to obtain the fault type, thereby completing the fault diagnosis. The application does not need to add additional internal measuring points, but only needs to perform fault diagnosis through the input and output ports of the RF circuit board, applies excitation to the RF circuit board and collects RF output signal data, analyzes these data by using a machine learning model, and thereby accurately identifies the fault type of the RF circuit board. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 is a flowchart of a specific embodiment of the data-driven RF circuit board fault diagnosis method of the application;

[0069] Figure 2 is a schematic diagram of the adaptive filtering in the application;

[0070] Figure 3 is a specific example of the double filtering processing in the application, wherein the first and second graphs are time domain and frequency domain graphs respectively; the third and fourth graphs are time domain and frequency domain graphs after the first filtering respectively, and the fifth graph is a frequency spectrum graph after the sliding maximum filtering. DETAILED DESCRIPTION

[0071] The specific embodiments of the application are described below in conjunction with the accompanying drawings, so that those skilled in the art can better understand the application. It should be particularly noted that in the following description, when the detailed description of known functions and designs may obscure the main content of the application, these descriptions will be omitted here.

[0072] Figure 1 is a flow chart of one embodiment of the present application based on data-driven RF circuit board fault diagnosis method.

[0073] In this embodiment, as shown in Figure 1 the present application based on data-driven RF circuit board fault diagnosis method includes the following steps:

[0074] Step S1: ADS modeling and fault type selection

[0075] Modeling the RF circuit board, selecting important components for monitoring, and introducing a fault model on the RF circuit board, introducing a predetermined fault on the RF circuit board according to the fault model, i.e. selecting a fault type.

[0076] In this embodiment, the RF fault circuit board is modeled by using ADS (Advanced Design System).

[0077] Step S2: Injecting excitation signal and collecting multiple sets of time domain data f(n)

[0078] For the selected fault type, a sweep signal is selected as the excitation signal f in (t) is injected into the input end of the RF circuit board, and the time domain signal f out (t) is collected from the output end of the RF circuit board, obtaining the time domain data f(n), repeating the above process to obtain multiple sets of time domain data f(n), thereby forming a systematic data set, wherein t is time and n is the sample point number sequence;

[0079] The entire fault diagnosis system consists of three main parts: host computer, RF transceiver terminal platform and RF fault board (RF circuit board with predetermined faults introduced). The host computer is responsible for controlling the RF transceiver terminal platform to generate the (RF) excitation signal f in (t) and inject it into the RF fault board. The RF signal output by the RF fault board is then collected and transmitted to the host computer for data display and fault diagnosis.

[0080] The host computer controls the RF terminal board to inject the excitation signal into the RF fault board through the serial port, in this embodiment, a sweep signal of 1.68-1.69 GHz is used as the excitation signal f in (t). By adjusting the resistance value and plugging the plug cap, the short circuit, open circuit and parameter drift of the main components (such as amplifiers and attenuators) in the RF circuit board are simulated. Only the original time domain signal is collected from the output end of the RF circuit board, and the fault type number F0-F13 is marked.

[0081] Step S3: Double filtering the collected time-domain data f(n)

[0082] The collected time-domain data is double filtered. First, the time-domain data is preliminarily filtered in the time domain; then, the data is converted to the frequency domain by Fourier transform, and the frequency-domain data is secondarily filtered. Compared with single filtering, this "double filtering" method can more flexibly select appropriate filtering schemes in the time domain and the frequency domain, thereby effectively reducing the noise amplitude of the frequency-domain waveform.

[0083] The number of collected points is n0, and the time-domain data is represented as f(n), 0<n<n0, containing multiple periods. The information of these data mainly falls into two categories: one is pure noise data s(n), n1<n<n2, 0<n1<n2<n0, and the other contains both noise and effective data. Therefore, first, a pure noise time-domain signal s(n) with the same length as the original data, 0<n<n0-1, needs to be constructed, and the specific steps are as follows.

[0084] Step S3.1: Construct a pure noise time-domain signal s extended (n)

[0085] Step S3.1.1: Extract the envelope signal of the time-domain data f(n)

[0086] a) Calculate the discrete Fourier transform to obtain the frequency-domain data F(k):

[0087]

[0088] Where N is the length of the discrete Fourier transform, and k represents the frequency point;

[0089] b) Calculate the frequency response of the Hilbert transform

[0090]

[0091] H(k) is the transformation matrix;

[0092] c) Calculate the signal after the Hilbert transform

[0093]

[0094] d) Calculate the envelope A(n):

[0095]

[0096] Step S3.1.2: Take out the pure noise data s(n)

[0097] A threshold a is determined, and the data points in the time domain data f(n) satisfying A(n) < a are determined as pure noise data, so that the pure noise data s(n) is extracted, n1 < n < n2, 0 < n1 < n2 < n0, wherein n0 is the number of sampling points of the time domain data f(n).

[0098] Step S3.1.3: generating an equal-length pure noise time domain signal s extended (n)

[0099] The pure noise data s(n) is periodically extended to generate an equal-length pure noise time domain signal s extended (n):

[0100] s extended (n) = s n ((n mod T) + n1)

[0101] wherein n = 0, 1, 2, … n0-1.

[0102] Step S3.2: adaptive filtering

[0103] In this embodiment, as shown in Figure 2 , the adaptive filtering process is:

[0104] a) filtering the pure noise time domain signal s extended (n) to obtain noise data

[0105]

[0106] wherein w(n) is a coefficient vector of the adaptive filter, w(n) = [w0(n), w1(n), …, wM-1(n)], M is the order of the adaptive filter, w(n), m = 0, 1, 2, … M-1 is the coefficient of the adaptive filter, s(n) = [s(n), s(n-1), …, s(n-M+1)]. M-1 T ; m extended extended extended extended T ;

[0107] b) calculating an error signal e(n):

[0108]

[0109] c) using the least mean square error criterion (LMS) algorithm, the mean square value of the error signal is minimized by iteratively adjusting the adaptive filter coefficients:

[0110] J(n) = E[e​​​​​​2 (n)]

[0111] wherein E[·] denotes the expected value, the iterative adjustment of the adaptive filter coefficients is:

[0112] w(n+1)=w(n)+μs extended (n)e(n)

[0113] wherein μ is a step factor, controlling the rate and stability of the adaptive filter coefficient update;

[0114] d) taking the error signal e(n) as the adaptive filtered time domain data

[0115] Step S3.3: Fourier transform

[0116] to the adaptive filtered time domain data a Fourier transform is performed, resulting in frequency domain data

[0117]

[0118] wherein N is the length of the discrete Fourier transform and k represents the frequency point;

[0119] Step S3.4: sliding maximum filter

[0120] In the frequency domain, a sliding maximum filter method is used for secondary filtering, which, on the basis of retaining the peak characteristics in the frequency domain, makes the time domain waveform smoother.

[0121] to the frequency domain data a sliding maximum filter is performed, for each frequency point k, the sliding window contains W frequency domain samples, the output of the maximum filter is the frequency domain data is the maximum value in the window:

[0122]

[0123] wherein the value range of k needs to ensure that k-m is within the effective frequency range, i.e. 0≤k-m<N.

[0124] In this embodiment, Figure 3 the processing process of the time domain data collected for the fault F0 through "double filtering" is shown, wherein the first and second graphs respectively show the original time domain graph and the frequency domain graph; the third and fourth graphs show the time domain graph and the frequency domain graph after the first filtering; the fifth graph shows the frequency spectrum graph after the sliding maximum filtering.

[0125] Step S4: feature extraction is performed on the frequency domain data ​

[0126] Step S4.1: Extracting key indicators describing the frequency domain waveform as sample features

[0127] In this embodiment, the key indicators describing the sample frequency domain waveform include maximum value, range, mean value, variance, rise time width, fall time width, bandwidth, peak factor, margin, waveform factor, etc., wherein:

[0128] a) The rise time width is the difference of the abscissa (n) when the signal reaches (20%) to (80%) of the range.

[0129] [Rise Time Width = |n 20 -n 80 |

[0130] where n 80 represents the abscissa at which the signal reaches the range (80%), and n 20 represents the abscissa at which the signal reaches the range.

[0131] b) The fall time width is the difference of the abscissa when the signal reaches (80%) to (20%) of the range.

[0132] [Fall Time Width = |n 80 -n 20 |

[0133] where n 80 represents the abscissa at which the signal reaches the range (80%), and n 20 represents the abscissa at which the signal (x(n)) reaches the range (20%).

[0134] c) The bandwidth is the difference of the abscissa when the signal reaches (50%) of the range to (50%) of the next range.

[0135] [Bandwidth = |n 50 -n2 50 |

[0136] where n 50 represents the abscissa at which the signal reaches the range (50%), and n2 50 represents the abscissa at which the signal next reaches the range (50%).

[0137] d) Peak factor (Peak Factor)

[0138]

[0139] where (RMS) is the root mean square value of the signal:

[0140] e) Waveform Factor

[0141]

[0142] f) Crest Factor

[0143]

[0144] Step S4.2: Extracting features representing the differences between different fault types, i.e., extracting inter-sample features;

[0145] Step S4.2.1: Constructing a "base frequency spectrum" for each fault type

[0146] a) First, randomly select L pieces of frequency domain data after secondary filtering from the fault type Let the mean of each piece of frequency domain data be x i , i = 1, 2,..., L, and let the mean of the means of the L pieces of frequency domain data be Then the residual v i of the mean of the i-th piece of frequency domain data x i is:

[0147]

[0148] b) Calculate the standard deviation σ x using the Bessel formula:

[0149]

[0150] c) Use the Lait test method to remove gross errors from the frequency domain data, i.e., if there is a frequency domain data whose residual absolute value is greater than 3 times the standard deviation, i.e., |v i | > 3σ x , this piece of frequency domain data is a gross error and is removed, and the process returns to step a) to reselect, calculate and remove until there are no gross errors in the L pieces of frequency domain data after secondary filtering.

[0151] d) Take the mean of the means of the L pieces of frequency domain data without gross errors as the "base frequency spectrum" of the fault type.

[0152] In this embodiment, L = 10, and taking fault type F0 as an example, the specific results are shown in Table 1:

[0153]

[0154]

[0155] Table 1

[0156] Step S4.2.2: Extracting the Pearson correlation coefficient: for the "base frequency spectrum" of the c-th fault type, denoted as and the "base frequency spectrum" of all fault types Perform the Pearson correlation coefficient calculation to obtain C correlation features r cj , j = 1, 2, …, C:

[0157]

[0158] Wherein, x ci is the i-th of the L ′ frequency domain data of the c-th fault type without containing gross errors, x ji is the j-th of the L ′ frequency domain data of the j-th fault type without containing gross errors.

[0159] The C correlation features r cj are the inter-sample features of the c-th fault type.

[0160] In this embodiment, when extracting features, the specific features of fault F0 are selected for detailed description, as shown in Table 2.

[0161]

[0162]

[0163] Table 2

[0164] In this embodiment, as shown in Table 1, the fault types include 38 sample features feature1-38 and 14 inter-sample features, i.e., Pearson correlation coefficients pearson1-14.

[0165] Step S5: Training the classifier

[0166] The sample features of the c-th fault type are spliced with the inter-sample features to obtain the fault features of the c-th fault type. For all fault types, the fault type is taken as a label, and the corresponding fault features are taken as training samples, which are input to the classifier. The classifier is trained until convergence.

[0167] Step S6: Fault diagnosis

[0168] For a piece of radio frequency circuit board, a sweep signal is selected as an excitation signal f in (t) is injected into the input end of the radio frequency circuit board, and a time domain signal f out(t) to obtain time domain data f(n), repeating the above process to obtain multiple sets of time domain data f(n) to form a systematic data set, then processing according to the method of steps S3, S4.2, S4.2.1 to obtain sample features and "base frequency spectrograms", and comparing them with "base frequency spectrograms" of all fault types Pursuing Pearson correlation coefficient calculation to obtain C correlation features, then splicing sample features and sample-to-sample features to obtain fault features, feeding them into the trained classifier of step S5 to obtain fault types, and completing fault diagnosis.

[0169] Table 3 shows the diagnostic effect of the present application.

[0170] Fault type Test sample Predicted accurate sample Accuracy rate F0 40 37 0.925 F1 40 39 0.975 F2 40 40 1 F3 40 39 0.975 F4 40 40 1 F5 40 40 1 F6 40 40 1 F7 40 40 1 F8 40 40 1 F9 40 40 1 F10 40 39 0.975 F11 40 40 1 F12 40 40 1 F13 40 40 1

[0171] In the present embodiment, for the fault prediction of 14 fault types F0-F13, 10 fault types all reached 100%, and the other four were also above 92.5%, achieving the expected effect.

[0172] Although the above describes the illustrative specific embodiments of the present application in order to facilitate the understanding of the present application by those skilled in the art, it should be clear that the present application is not limited to the scope of the specific embodiments, and for those skilled in the art, as long as various changes are within the spirit and scope of the present application as defined and determined by the appended claims, all the inventions utilizing the concept of the present application are within the scope of protection.

Claims

1. A data-driven method for diagnosing radio frequency circuit board faults, characterized in that, Includes the following steps: (1) Model the radio frequency circuit board, select important components for monitoring, and introduce a fault model on the radio frequency circuit board. Based on the fault model, introduce a predetermined fault on the radio frequency circuit board, i.e., select a fault type. (2) For the selected fault type, the frequency sweep signal is selected as the excitation signal. Injected into the input terminal of the RF circuit board, and time-domain signal acquired from the output terminal of the RF circuit board. Time-domain data obtained Repeat the above process to obtain multiple sets of time-domain data. This forms a systematic dataset, where t is time and n is the number of sampling points. (3) A set of time-domain data collected Perform dual filtering; 3.1) Construct a length and time domain data The same pure noise time-domain signal ; 3.1.1) For time-domain data Extract the envelope signal; a) Calculate the Discrete Fourier Transform to obtain the frequency domain data. : ； Where N is the length of the discrete Fourier transform, and k represents the frequency point; b) Calculate the frequency response of the Hilbert transform. : ； The transformation matrix; c) Calculate the signal after Hilbert transform. : ； d) Calculate the envelope : ； 3.1.2) Extract pure noise data ; Determine a threshold For satisfying Time domain data The data points in the data were identified as pure noise data, and thus the pure noise data was extracted. ,in, For time domain data The number of sampling points; 3.1.3) Generate a time-domain signal of equal length containing pure noise. ; Pure noise data Periodically extended, generating a time-domain signal of equal length consisting of pure noise. : ； in, =0,1,2,… ; 3.2) Adaptive filtering; a) For pure noise time-domain signals Filtering is performed to obtain noise data. : ； in, Let's consider the coefficient vector of the adaptive filter. , It is the order of the adaptive filter. These are the coefficients of the adaptive filter. ; b) Calculate the error signal : ； c) Employing the minimum mean square error criterion algorithm, the mean square value of the error signal is minimized by iteratively adjusting the adaptive filter coefficients: ； in, To represent the expected value, the adaptive filter coefficients are iteratively adjusted as follows: ； in, It is the step size factor, which controls the rate and stability of the adaptive filter coefficient update; d) The error signal Time-domain data after adaptive filtering ; 3.3) Fourier Transform; For the time-domain data after adaptive filtering Perform a Fourier transform to obtain the frequency domain data. : ； Where N is the length of the discrete Fourier transform, and k represents the frequency point; 3.4) Sliding maximum value filtering; For frequency domain data Perform sliding maximum filtering; for each frequency domain point k, the sliding window contains... The frequency domain data is obtained by filtering a sample in the frequency domain and then filtering the maximum value. The maximum value within the window: ； The value of k must be within the effective frequency range to ensure that km is within the effective frequency range, i.e. ; (4) Frequency domain data Perform feature extraction; 4.1) Extract key indicators describing the frequency domain waveform as sample features; 4.2) Extracting features that characterize the differences between different fault types, i.e., extracting inter-sample features; 4.2.1) Construct a "baseline spectrum" for each fault type; a) First, randomly select L frequency domain data points that have undergone secondary filtering from this fault type. The mean of a single frequency domain data point is denoted as follows: The mean of the L frequency domain data points is denoted as Then the mean of the i-th frequency domain data residual : ； b) Calculate the standard deviation using Bessel's formula. : ； c) Use the Wright test to remove gross errors from the frequency domain data. That is, if the absolute value of the residual in any frequency domain data exceeds three times the standard deviation, then... The frequency domain data has a gross error and should be removed. Return to step a) to reselect, calculate and remove data until L randomly selected frequency domain data after double filtering are free of gross errors. d) The mean of the L frequency domain data points that do not contain gross errors This serves as the "baseline spectrum" for this type of fault; 4.2.2) Extracting the Pearson correlation coefficient: For the "basic spectrum" of the c-th fault type, denoted as... And compared with the "baseline spectrum" of all fault types. The Pearson correlation coefficient was calculated to obtain... One relevant feature : ； in, For the c-th fault type, excluding gross errors In the frequency domain data strip, For the j-th fault type, excluding gross errors In the frequency domain data strip; One relevant feature That is, the inter-sample characteristics of the c-th fault type; (5) Concatenate the sample features of the c-th fault type with the features between samples to obtain the fault features of the c-th fault type. For all fault types, use the fault type as the label and the corresponding fault features as training samples, input them into the classifier, and train the classifier until convergence. (6) Fault diagnosis; For an RF circuit board, a frequency sweep signal is selected as the excitation signal. Injected into the input terminal of the RF circuit board, and time-domain signal acquired from the output terminal of the RF circuit board. Time-domain data obtained Repeat the above process to obtain multiple sets of time-domain data. This forms a systematic dataset. Then, it is processed according to steps (3), 4.1), and 4.2.1) to obtain sample features and a "baseline spectrum diagram". These are then compared with the "baseline spectrum diagrams" of all fault types. The Pearson correlation coefficient was calculated to obtain... The relevant features are then concatenated with the sample features and the inter-sample features to obtain the fault features. These features are then fed into the classifier that has been trained and converged in step (5) to obtain the fault type and complete the fault diagnosis.

2. The data-driven RF circuit board fault diagnosis method according to claim 1, characterized in that, The sample features include maximum value, range, mean, variance, rising edge width, falling edge width, bandwidth, peak factor, margin, and waveform factor.

Images