HPLC (High Performance Liquid Chromatography) dual-mode communication stability method, medium and system

Abstract

The invention provides an HPLC (High Performance Liquid Chromatography) dual-mode communication stability method, medium and system, and belongs to the technical field of dual-channel multiplexing communication, and the method comprises the following steps: firstly, collecting HPLC low-frequency / high-frequency carrier signals, working voltage / current, temperature and electromagnetic interference data in real time; stable, variable and noise components of a carrier signal are respectively obtained through empirical mode decomposition, and a correlation coefficient matrix of the variable components and interference factors is established by using multiple linear regression and singular value decomposition. And calculating a signal intensity, a signal-to-noise ratio and an error rate evaluation matrix based on the stability coefficient matrix, and constructing an abnormal feature matrix in combination with a communication protocol specification. And finally, inputting the features into a five-layer full-connection neural network, outputting fault type probability distribution through a softmax function, and taking the type corresponding to the maximum probability value as a diagnosis result. According to the invention, the technical problem of rapid and accurate diagnosis of HPLC communication stability in a complex and changeable low-voltage distribution network environment is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of dual-channel multiplexing communication, and in particular, relates to a HPLC dual-mode communication stability method, medium and system. Background Art

[0002] With the rapid advancement of smart grid construction, low-voltage power line carrier communication (HPLC) technology has been widely used in the fields of electricity consumption information collection, smart home control, demand side management, etc. HPLC technology realizes data transmission by superimposing high-frequency modulated signals on existing power lines, with the advantages of low wiring cost and wide coverage. At present, the HPLC communication system mainly adopts a single-band carrier transmission scheme, selecting a working frequency band in the 1-10MHz low frequency band or the 10-30MHz high frequency band for communication. In practical applications, due to the complex and changeable electrical environment of the low-voltage distribution network, HPLC communication faces many challenges: first, the power line channel has a strong frequency selective fading characteristic, and there are significant differences in channel conditions in different frequency bands; second, nonlinear loads such as switching power supplies and inverters will generate wide-band electromagnetic interference, which seriously affects the transmission quality of the carrier signal; third, the dynamic changes in the topology of the distribution network cause frequent fluctuations in channel characteristics, resulting in unstable communication links. Especially in scenarios such as electricity consumption information collection, the system needs to ensure the real-time and reliability of data collection, which puts higher requirements on the stability of HPLC communication.

[0003] The existing HPLC system mainly relies on manual experience to judge communication failures and calculate the stability of the HPLC system, and lacks a systematic fault diagnosis method. The common practice is to evaluate the communication quality by measuring single indicators such as signal strength and bit error rate. This method has the following problems: First, it is impossible to accurately identify the type of fault, resulting in low maintenance efficiency; second, it is difficult to adapt to the complex and changeable power line channel environment, and the fault diagnosis results are not reliable enough; third, there is a lack of effective detection of abnormalities at the communication protocol level, and it is easy to miss certain fault types. With the expansion of the scale of smart grids, these problems are becoming increasingly prominent, seriously restricting the promotion and application of HPLC technology.

[0004] The most urgent core technical problem that needs to be solved is: how to achieve rapid and accurate diagnosis of the stability of the HPLC communication system in a complex and changeable low-voltage distribution network environment. Summary of the invention

[0005] In view of this, the present invention provides a HPLC dual-mode communication stability method, medium and system, which can solve the technical problem of realizing rapid and accurate diagnosis of the stability of the HPLC communication system in a complex and changeable low-voltage power distribution network environment.

[0006] The present invention is achieved in that:

[0007] The first aspect of the present invention provides a HPLC dual-mode communication stability method, comprising the following steps:

[0008] S10, real-time data collection, all collected data are stored in the form of one-dimensional arrays, the sampling frequency is 1000 Hz, each one-dimensional array contains 1000 sampling point data; specifically collect the following data: HPLC low-frequency carrier signal data, HPLC high-frequency carrier signal data, working voltage data, working current data, ambient temperature data, electromagnetic interference data;

[0009] S20, using an empirical mode decomposition method to decompose the collected HPLC low-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC low-frequency carrier stable component data array, an HPLC low-frequency carrier variable component data array, and an HPLC low-frequency carrier noise component data array;

[0010] S30, using an empirical mode decomposition method to decompose the collected HPLC high-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC high-frequency carrier stable component data array, an HPLC high-frequency carrier variable component data array, and an HPLC high-frequency carrier noise component data array;

[0011] S40. Establish the relationship between the HPLC low-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data through a multivariate linear regression method to obtain a 3-row and 3-column HPLC low-frequency carrier variation coefficient matrix, wherein each element in the HPLC low-frequency carrier variation coefficient matrix represents the influence weight of the corresponding factor; based on the HPLC low-frequency carrier variation coefficient matrix, use a singular value decomposition method to calculate a 3-row and 3-column HPLC low-frequency carrier stability coefficient matrix;

[0012] S50. Use a multivariate linear regression method to establish the relationship between the HPLC high-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data, and obtain a 3-row and 3-column HPLC high-frequency carrier variation coefficient matrix, wherein each element in the HPLC high-frequency carrier variation coefficient matrix represents the influence weight of the corresponding factor; based on the HPLC high-frequency carrier variation coefficient matrix, use a singular value decomposition method to calculate a 3-row and 3-column HPLC high-frequency carrier stability coefficient matrix;

[0013] S60, based on the HPLC low-frequency carrier stability coefficient matrix and the HPLC high-frequency carrier stability coefficient matrix, respectively calculate three 6-row and 6-column evaluation matrices: a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, and a bit error rate evaluation matrix, wherein the signal strength evaluation matrix is ​​obtained by calculating the root mean square value of the stable component data array, the signal-to-noise ratio evaluation matrix is ​​obtained by calculating the power ratio of the stable component data array to the noise component data array, and the bit error rate evaluation matrix is ​​calculated by analyzing the number of fluctuations exceeding a preset threshold group in the variable component data array;

[0014] S70, based on the data frame structure, check code type, and timing parameter requirements defined in the communication protocol specification, construct a 12-row and 12-column dual-carrier communication protocol abnormal feature matrix, wherein the data frame structure includes an 8-byte frame header, a data segment length, and a 4-byte frame tail, and the check code type is a 32-bit cyclic redundancy check code;

[0015] S80, adopting an embedded five-layer fully connected neural network structure fault diagnosis model, inputting a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, a bit error rate evaluation matrix and a dual-carrier communication protocol abnormal feature matrix into the fault diagnosis model, the number of input layer nodes of the fault diagnosis model is all elements of the evaluation matrix and the abnormal feature matrix, the hidden layer of the fault diagnosis model uses a rectified linear unit activation function, and the output layer of the fault diagnosis model uses a normalized exponential function to output probability distribution vectors of 10 predefined fault types;

[0016] S90, input each predefined fault type and its probability distribution vector into a predefined stability detection equation, obtain the stability index of HPLC dual-mode communication and output it.

[0017] The embedded mathematical model is specifically a set of mathematical equations, including a signal quality assessment equation, a protocol anomaly detection equation, a carrier stability analysis equation, an environmental impact assessment equation, and a feature fusion equation;

[0018] The signal quality evaluation equation is used to evaluate the overall transmission quality of the carrier signal. The input includes a signal strength evaluation matrix and a signal-to-noise ratio evaluation matrix. Both matrices are 6×6 dimensions. The output is a 36-dimensional signal quality feature vector.

[0019] The protocol anomaly detection equation is used to identify fault features at the communication protocol level. The input includes a dual-carrier communication protocol anomaly feature matrix and a bit error rate assessment matrix, which are 12×12 and 6×6 dimensions respectively. The output is a 144-dimensional protocol anomaly feature vector.

[0020] The carrier stability analysis equation is used to analyze the dynamic stability of the carrier signal. The input includes the HPLC low-frequency carrier stability coefficient matrix and the HPLC high-frequency carrier stability coefficient matrix. Both matrices are 3×3 dimensions. The output is an 18-dimensional carrier stability feature vector.

[0021] The environmental impact assessment equation is used to assess the impact of the external environment on the communication system. The input includes an HPLC low-frequency carrier variation coefficient matrix and an HPLC high-frequency carrier variation coefficient matrix. Both matrices are 3×3 dimensions. The output is an 18-dimensional environmental impact feature vector.

[0022] The feature fusion equation is used to generate the features that are finally input to the neural network. The input includes the feature vectors output by the first four equations, totaling 216-dimensional features, and the output is a 100-dimensional fused feature vector.

[0023] The steps for obtaining the training data set of the fault diagnosis model specifically include simulating 10 predefined fault scenarios in a laboratory environment, collecting complete data samples under each fault type and collecting 1,000 sets of data for each fault, using data enhancement technology to expand the training set by adding Gaussian noise and random disturbances, annotating the data set to establish a correspondence between fault types and features, and dividing the training set, validation set, and test set in a ratio of 6:2:2.

[0024] The steps of training the fault diagnosis model specifically include standardizing and normalizing the collected raw data, extracting feature vectors using an embedded mathematical model, randomly initializing the weights of the neural network, training the model using a small batch gradient descent method with a batch size of 64, updating the model parameters using an Adam optimizer with a learning rate of 0.001, regularly evaluating the model performance on a validation set, stopping the training when the validation set performance has not improved for five consecutive epochs, and saving the trained optimal model parameters.

[0025] The mathematical formulas involved in each step and their principles are described in detail below:

[0026] 1. The empirical mode decomposition (EMD) method in step S20 decomposes the HPLC low-frequency carrier signal as follows:

[0027]

[0028] Where x(t) is the original HPLC low-frequency carrier signal; c i (t) is the i-th intrinsic mode function (IMF) component; r n (t) is the residual component; n is the number of decomposition levels, and n=3 in this scheme.

[0029] The specific IMF extraction process is as follows:

[0030] h1(t)=x(t)-m1(t);

[0031] h k (t) = h k-1 (t)-m k (t);

[0032] In the formula, m k (t) is h k-1 The average value of the upper and lower envelopes of (t); k is the number of iterations; when h k (t) Stop the iteration when the IMF condition is met.

[0033] The extraction method of the three components of HPLC low-frequency carrier is:

[0034] S stable (t) = r n (t);

[0035] S var (t) = c2(t);

[0036] S noise (t) = c1(t);

[0037] In the formula, S stable (t) is the stable component; S var (t) is the variable component; S noise (t) is the noise component.

[0038] 2. The multiple linear regression equation in step S40 is specifically expressed as follows:

[0039] V var =β0+β1E int +β2T env +β3V work +∈;

[0040] Where V var is the low-frequency carrier variation component; E int is the electromagnetic interference intensity; T env is the ambient temperature; V work is the operating voltage; β0, β1, β2, β3 are regression coefficients; ∈ is the random error term.

[0041] The HPLC low-frequency carrier variation coefficient matrix is ​​expressed as:

[0042]

[0043] 3. Stability coefficient matrix calculation method based on singular value decomposition (SVD):

[0044] M var =UΣVT;

[0045] M stable =U∑ -1 V T ;

[0046] Where U, V are orthogonal matrices; ∑ is a singular value diagonal matrix.

[0047] 4. The signal quality evaluation equation is specifically expressed as follows:

[0048]

[0049] Where Q(i,j) is the signal quality evaluation matrix element; N is the number of sampling points; S stable (k) is the k-th point value of the stable component; P signal ,P noise are signal power and noise power respectively; N error is the number of error frames; N total is the total number of frames; α, β, γ are weight coefficients; δ is the correction term.

[0050] 5. The protocol anomaly detection equation is specifically expressed as follows:

[0051]

[0052] Where P(i,j) is the element of the protocol anomaly feature matrix; f k (x),g k (y) is the feature extraction function; w k is the weight coefficient; m is the number of features; η is the random disturbance term.

[0053] 6. The carrier stability analysis equation is specifically expressed as follows:

[0054]

[0055] Where S(t) is the carrier stability characteristic; M stable is the stability coefficient matrix; λ1,λ2,λ3 are weighting coefficients; ξ is the damping coefficient; ω is the angular frequency; μ is the system error term.

[0056] The carrier stability matrix is ​​expressed as:

[0057]

[0058] In the formula, s ij It represents the stability coefficient of the i-th frequency band to the j-th environmental factor.

[0059] 7. The environmental impact assessment equation is specifically expressed as follows:

[0060]

[0061] Where E(t) is the environmental impact characteristic; k ij is the coupling coefficient; M var (i,j) is the element of the coefficient of variation matrix; θ ij is the time attenuation coefficient; c i is the periodic term coefficient; φ i is the phase angle; ρ is the random fluctuation term.

[0062] The environmental impact matrix is ​​expressed as:

[0063]

[0064] In the formula, e ij Represents the influence coefficient of the i-th environmental factor on the j-th carrier parameter.

[0065] 8. The feature fusion equation is specifically expressed as:

[0066]

[0067] Where F is the fused feature vector; σ is the activation function; W i is the weight matrix; X i is the input feature vector; b is the bias term; τ j is the time scale parameter; Δ j is the feature difference item; n is the number of input features; m is the number of difference items.

[0068] The feature fusion weight matrix is ​​expressed as:

[0069]

[0070] In the formula, w ij Represents the weight coefficient of the i-th output feature to the j-th input feature.

[0071] 1. Description of variables obtained in the experiment:

[0072] 1. Original signal data acquisition:

[0073] (1) HPLC low-frequency carrier signal x low (t): The carrier signal in the 1-10MHz frequency band is sampled in real time by the ADC sampler in the RF front-end module, with a sampling frequency of 1000Hz and 1000 points collected each time;

[0074] (2)HPLC high frequency carrier signal x high (t): The carrier signal in the 10-30MHz frequency band is sampled in real time by the ADC sampler in the RF front-end module, with a sampling frequency of 1000Hz and 1000 points collected each time;

[0075] (3) Operating voltage Vwork : The system operating voltage is collected through AD7606, ranging from 3.3V to 5V;

[0076] (4) Working current I work : Collect system operating current through ACS712, range 0-5A;

[0077] (5) Ambient temperature T env : Collect ambient temperature through DS18B20, the range is -55℃-125℃;

[0078] (6) Electromagnetic interference E int : Collect the environmental electromagnetic interference intensity through EM100, the range is 0-100dBμV / m.

[0079] 2. Principle and parameter description of the empirical mode decomposition (EMD) method:

[0080] 1. Basic principles of EMD:

[0081] The signal is decomposed into several intrinsic mode functions (IMFs) and a residual component through iterative screening. Each IMF must satisfy:

[0082] (1) The difference between the number of extreme points and the number of zero-crossing points does not exceed 1;

[0083] (2) The average value of the upper and lower envelopes at any time is zero.

[0084] 2. Mathematical representation of EMD decomposition process:

[0085]

[0086] Where x(t) is the original signal; c i (t) is the i-th IMF component; r n (t) is the residual component; n is the number of decomposition levels, and this scheme takes n=3.

[0087] 3.IMF extraction process:

[0088] (1) Find all local maximum and minimum points of the signal;

[0089] (2) Perform cubic spline interpolation on the maximum and minimum points to obtain the upper envelope e up (t) and the lower envelope e down (t);

[0090] (3) Calculate the mean:

[0091] (4) Subtract the mean from the original signal: h1(t) = x(t) - m1(t);

[0092] (5) Repeat the above steps until the IMF condition is met and the first IMF component c1(t) is obtained;

[0093] (6) Subtract c1(t) from the original signal and repeat the above steps for the remaining signal to obtain other IMF components.

[0094] 4. The physical meaning of the three components:

[0095] (1) Stable component S stable (t) = r n (t): represents the long-term trend of the signal;

[0096] (2) Variation component S var (t) = c2(t): represents the mid-term fluctuation of the signal;

[0097] (3) Noise component S noise (t)=c1(t): represents the high-frequency noise of the signal.

[0098] 3. Principle and parameter description of multivariate linear regression equation:

[0099] 1. Modeling the relationship between variable components and environmental factors:

[0100] V var =β0+β1E int +β2T env +β3V work +∈;

[0101] Where, V var is the variable component; E int is the electromagnetic interference intensity; T env is the ambient temperature; V work is the operating voltage; β0, β1, β2, β3 are regression coefficients; ∈ is the random error term.

[0102] 2. Regression coefficient calculation steps:

[0103] (1) Construct sample matrix X and target vector Y;

[0104] (2) Use the least squares method to solve: β = (X T X) -1 X T Y;

[0105] (3) Calculation of the coefficient of determination R 2 Evaluate the fit.

[0106] 3. Construction of coefficient of variation matrix:

[0107]

[0108] In the formula, β ij Represents the influence coefficient of the i-th environmental factor on the j-th carrier parameter.

[0109] 4. Singular value decomposition (SVD) principle and parameter description:

[0110] 1. Basic principles of SVD:

[0111] Decompose the matrix into the product of three matrices, where the middle matrix is ​​a diagonal matrix and the elements on the diagonal are singular values.

[0112] 2.SVD calculation steps:

[0113] (1) Perform SVD decomposition on the coefficient of variation matrix:

[0114] M var =U∑V T ;

[0115] Where U, V are orthogonal matrices; Σ is a singular value diagonal matrix.

[0116] (2) Calculate the stability coefficient matrix:

[0117] M stable =U∑ -1 V T .

[0118] 5. Signal quality evaluation equation principle and parameter description:

[0119] 1. Evaluation index composition:

[0120]

[0121] Where Q(i,j) is the signal quality evaluation matrix element; N is the number of sampling points; S stable (k) is the k-th point value of the stable component; P signal ,P noise are signal power and noise power respectively; N error is the number of error frames; N total is the total number of frames; α, β, γ are weight coefficients; δ is the correction term.

[0122] 2. Calculation methods of various indicators:

[0123] (1)Signal strength indicator:

[0124] (2) Signal-to-noise ratio index:

[0125] (3) Bit error rate index:

[0126] 6. Protocol anomaly detection equation principle and parameter description:

[0127] 1. Abnormal feature extraction:

[0128]

[0129] Where, P(i,j) is the element of the protocol anomaly feature matrix; f k (x),g k (y) is the feature extraction function; w k is the weight coefficient; m is the number of features; η is the random disturbance term.

[0130] 2. Feature extraction function description:

[0131] (1)f k (x) is used to extract time domain features:

[0132]

[0133] (2) g k (y) is used to extract frequency domain features:

[0134]

[0135] 7. Carrier stability analysis equation principle and parameter description:

[0136] 1. Stability evaluation model:

[0137]

[0138] Where S(t) is the carrier stability characteristic; M stable is the stability coefficient matrix; λ1,λ2,λ3 are weighting coefficients; ξ is the damping coefficient; ω is the angular frequency; r is the system error term.

[0139] 2. Stability Matrix Construction:

[0140]

[0141] In the formula, s ij It represents the stability coefficient of the i-th frequency band to the j-th environmental factor.

[0142] 8. Environmental impact assessment equation principle and parameter description:

[0143] 1. Environmental impact model:

[0144]

[0145] Where E(t) is the environmental impact characteristic; k ij is the coupling coefficient; M var(i,j) is the element of the coefficient of variation matrix; θ ij is the time attenuation coefficient; c i is the periodic term coefficient; φ i is the phase angle; ρ is the random fluctuation term.

[0146] 2. Environmental impact matrix:

[0147]

[0148] In the formula, e ij Represents the influence coefficient of the i-th environmental factor on the j-th carrier parameter.

[0149] 9. Feature fusion equation principle and parameter description:

[0150] 1. Feature fusion model:

[0151]

[0152] Where F is the fusion feature vector; σ is the activation function; W i is the weight matrix; X i is the input feature vector; b is the bias term; τ j is the time scale parameter; Δ j is the feature difference item; n is the number of input features; m is the number of difference items.

[0153] 2. Feature fusion weight matrix:

[0154]

[0155] Where, w ij Represents the weight coefficient of the i-th output feature to the j-th input feature.

[0156] 10. Equation construction principle and innovation points:

[0157] 1. Reasons for choosing EMD method:

[0158] The EMD method is adaptive and can effectively process nonlinear and non-stationary signals. It is suitable for analyzing the multi-scale characteristics of HPLC carrier signals.

[0159] 2. Multiple linear regression equation design ideas:

[0160] The linear model is used to describe the impact of environmental factors on the carrier signal, which simplifies the complex nonlinear relationship and facilitates practical application.

[0161] 3.Innovative application of SVD method:

[0162] By establishing a stable coefficient matrix through SVD decomposition, the main features of the signal are effectively extracted and the computational complexity is reduced.

[0163] 4. Design features of signal quality evaluation equation:

[0164] The three key indicators of signal strength, signal-to-noise ratio and bit error rate are comprehensively considered, and the signal-to-noise ratio is processed in logarithmic form to make the evaluation results more in line with the perception characteristics of the human ear.

[0165] 5. Innovation points of protocol anomaly detection equation:

[0166] The introduction of dual feature extraction functions in time domain and frequency domain improves the accuracy and robustness of anomaly detection.

[0167] 6. Features of carrier stability analysis equation:

[0168] Combining the differential term and the oscillation term, both the instantaneous change characteristics of the signal and the long-term stability are taken into account.

[0169] 7. Environmental impact assessment equation advantages:

[0170] The coupling coefficient and time decay term are used to accurately describe the dynamic influence process of environmental factors.

[0171] 8. Improvement of feature fusion equation:

[0172] The introduction of time scale parameters and feature difference terms improves the discrimination and generalization ability of fused features.

[0173] 11. Overall technical innovations:

[0174] 1. Multi-scale analysis: The multi-scale decomposition of the signal is realized through the EMD method, which improves the accuracy of feature extraction;

[0175] 2. Adaptive evaluation: multiple evaluation matrices are used to achieve adaptive fault detection capabilities;

[0176] 3. Feature fusion: innovatively fuses multi-source features nonlinearly to improve the robustness of the system;

[0177] 4. Real-time performance: Simplifies complex nonlinear relationships and ensures the real-time processing capabilities of the system.

[0178] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned HPLC dual-mode communication stability method.

[0179] A third aspect of the present invention provides an HPLC dual-mode communication fault detection system, comprising the above-mentioned computer-readable storage medium.

[0180] Compared with the prior art, the HPLC dual-mode communication stability method, medium and system provided by the present invention have the following beneficial effects:

[0181] First, the present invention adopts a dual-band carrier transmission scheme, communicating in the low frequency band (1-10MHz) and high frequency band (10-30MHz) at the same time, which significantly improves the anti-interference ability and communication reliability of the system. The dual-band carrier signal is decomposed by the empirical mode decomposition method to accurately extract the stable component, variable component and noise component, providing a reliable data basis for fault feature analysis.

[0182] Secondly, the present invention innovatively constructs a multi-dimensional stability evaluation system, including a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, and a bit error rate evaluation matrix. The influence weights of external environmental factors are analyzed by the multivariate linear regression method, and the stability characteristics of the carrier signal are extracted by combining the singular value decomposition technology, thus achieving a comprehensive evaluation of the communication quality.

[0183] Thirdly, the present invention designs a fault diagnosis model based on deep learning, which adopts a five-layer fully connected neural network structure and can adaptively learn complex fault feature patterns. By integrating the physical layer signal features and the protocol layer abnormal features, it can accurately identify 10 predefined fault types.

[0184] In summary, the present invention solves the technical problem of realizing rapid and accurate diagnosis of HPLC communication system faults in a complex and changeable low-voltage power distribution network environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0185] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0186] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0187] like Figure 1 As shown, it is a flow chart of a HPLC dual-mode communication stability method provided by the present invention, and the method comprises the following steps:

[0188] S10, real-time data collection, all collected data are stored in the form of one-dimensional arrays, the sampling frequency is 1000 Hz, each one-dimensional array contains 1000 sampling point data; specifically collect the following data: HPLC low-frequency carrier signal data, HPLC high-frequency carrier signal data, working voltage data, working current data, ambient temperature data, electromagnetic interference data;

[0189] S20, using an empirical mode decomposition method to decompose the collected HPLC low-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC low-frequency carrier stable component data array, an HPLC low-frequency carrier variable component data array, and an HPLC low-frequency carrier noise component data array;

[0190] S30, using an empirical mode decomposition method to decompose the collected HPLC high-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC high-frequency carrier stable component data array, an HPLC high-frequency carrier variable component data array, and an HPLC high-frequency carrier noise component data array;

[0191] S40. Establish the relationship between the HPLC low-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data through a multivariate linear regression method to obtain a 3-row and 3-column HPLC low-frequency carrier variation coefficient matrix, wherein each element in the HPLC low-frequency carrier variation coefficient matrix represents the influence weight of the corresponding factor; based on the HPLC low-frequency carrier variation coefficient matrix, use a singular value decomposition method to calculate and obtain a 3-row and 3-column HPLC low-frequency carrier stability coefficient matrix;

[0192] S50. Use a multivariate linear regression method to establish the relationship between the HPLC high-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data, and obtain a 3-row and 3-column HPLC high-frequency carrier variation coefficient matrix, wherein each element in the HPLC high-frequency carrier variation coefficient matrix represents the influence weight of the corresponding factor; based on the HPLC high-frequency carrier variation coefficient matrix, use a singular value decomposition method to calculate and obtain a 3-row and 3-column HPLC high-frequency carrier stability coefficient matrix;

[0193] S60, based on the HPLC low-frequency carrier stability coefficient matrix and the HPLC high-frequency carrier stability coefficient matrix, respectively calculate three 6-row and 6-column evaluation matrices: a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, and a bit error rate evaluation matrix, wherein the signal strength evaluation matrix is ​​obtained by calculating the root mean square value of the stable component data array, the signal-to-noise ratio evaluation matrix is ​​obtained by calculating the power ratio of the stable component data array to the noise component data array, and the bit error rate evaluation matrix is ​​calculated by analyzing the number of fluctuations exceeding a preset threshold group in the variable component data array;

[0194] S70, based on the data frame structure, check code type, and timing parameter requirements defined in the communication protocol specification, construct a 12-row and 12-column dual-carrier communication protocol abnormal feature matrix, wherein the data frame structure includes an 8-byte frame header, a data segment length, and a 4-byte frame tail, and the check code type is a 32-bit cyclic redundancy check code;

[0195] S80, adopting an embedded five-layer fully connected neural network structure fault diagnosis model, inputting a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, a bit error rate evaluation matrix and a dual-carrier communication protocol abnormal feature matrix into the fault diagnosis model, the number of nodes in the input layer of the fault diagnosis model is all elements of the evaluation matrix and the abnormal feature matrix, the hidden layer of the fault diagnosis model uses a rectified linear unit activation function, and the output layer of the fault diagnosis model uses a normalized exponential function to output probability distribution vectors of 10 predefined fault types;

[0196] S90, input each predefined fault type and its probability distribution vector into a predefined stability detection equation, obtain the stability index of HPLC dual-mode communication and output it.

[0197] The specific implementation methods of the above steps are described in detail below:

[0198] The specific implementation of step S10 is to collect key data in real time during the operation of the HPLC system. The sampling frequency is set to f s = 1000Hz, each one-dimensional array contains N = 1000 sampling points. The data collection process is as follows:

[0199] 1) Through the ADC sampler in the RF front-end module, the carrier signal in the frequency band of f1=1MHz to f2=10MHz is collected to obtain the HPLC low-frequency carrier signal data x1(t).

[0200] 2) Through the ADC sampler in the RF front-end module, the carrier signal in the frequency band of f3 = 10 MHz to f4 = 30 MHz is collected to obtain the HPLC high-frequency carrier signal data x2(t).

[0201] 3) Use the voltage sampling chip AD7606 to measure the system operating voltage V in real time work .

[0202] 4) Use Hall current sensor ACS712 to measure the system operating current I work .

[0203] 5) Use digital temperature sensor DS18B20 to detect ambient temperature T env .

[0204] 6) Collect the environmental electromagnetic interference intensity E through the electromagnetic field strength detection module EM100 int .

[0205] All collected data are stored in the form of a one-dimensional array, laying the foundation for subsequent signal analysis and fault diagnosis.

[0206] The specific implementation of step S20 is to use the empirical mode decomposition (EMD) method to decompose the collected HPLC low-frequency carrier signal x1(t) to extract the stable component, variable component and noise component of the signal. The mathematical expression of the EMD method is:

[0207]

[0208] Among them, c i (t) is the i-th intrinsic mode function (IMF) component, r n (t) is the residual component, n is the number of decomposition levels, and n=3.

[0209] The specific IMF extraction process is:

[0210] h1(t)=x1(t)-m1(t);

[0211] h k (t) = h k-1 (t)-m k (t);

[0212] Among them, m k (t) is h k-1 (t) is the average value of the upper and lower envelopes, k is the number of iterations. k (t) Stop the iteration when the IMF condition is met.

[0213] The three components finally extracted are:

[0214] S stable (t) = r3(t);

[0215] S var (t) = c2(t);

[0216] S noise (t) = c1(t);

[0217] Among them, S stable (t) is the stable component, S var (t) is the variable component, S noise (t) is the noise component.

[0218] The specific implementation of step S30 is similar to step S20, and the HPLC high-frequency carrier signal x2(t) is decomposed by the EMD method to obtain S stable (t),S var (t) and S noise (t). The mathematical expression of HPLC high frequency carrier signal decomposition is:

[0219]

[0220] The specific implementation of step S40 is to establish the HPLC low-frequency carrier variation component V var The multivariate linear regression model between the factors and the environment is expressed as follows:

[0221] V var =β0+β1E int +β2T env +β3V work +∈;

[0222] Among them, β0, β1, β2, β3 are regression coefficients, and ∈ is a random error term. The regression coefficient matrix is ​​obtained by least squares fitting:

[0223]

[0224] The matrix reflects the influence weight of each environmental factor on the change of HPLC low-frequency carrier.

[0225] The specific implementation of step S50 is similar to step S40, establishing the HPLC high-frequency carrier variation component V var The multivariate linear regression model between the HPLC high frequency carrier coefficient matrix M was obtained. var .

[0226] The specific implementation of step S60 is to calculate three evaluation matrices based on the stable component, variable component and noise component data obtained in steps S20 and S30:

[0227] 1) Signal strength evaluation matrix Q(i,j):

[0228]

[0229] Among them, α, β, γ are weight coefficients, δ is the correction term, P signal ,P noise are the signal power and noise power respectively.

[0230] 2) Signal-to-noise ratio evaluation matrix R(i,j):

[0231]

[0232] 3) Bit error rate evaluation matrix P(i,j):

[0233]

[0234] The preset thresholds include: signal strength threshold -85dBm, signal-to-noise ratio threshold 15dB, bit error rate threshold 1×10 -3 .

[0235] The specific implementation of step S70 is to construct a 12×12 abnormal feature matrix P(i, j) according to the specification requirements of the dual-carrier communication protocol, and the mathematical expression is:

[0236]

[0237] Among them, f k (x),g k (y) is the feature extraction function, w k is the weight coefficient, m is the number of features, and η is the random perturbation term. The matrix covers key features such as data frame structure, check code type, and timing parameters.

[0238] The specific implementation of step S80 is to use a fault diagnosis model with an embedded five-layer fully connected neural network structure. The input of the model is the evaluation matrix and abnormal feature matrix calculated above, totaling 216-dimensional feature vector X. The hidden layer uses the rectified linear unit (ReLU) as the activation function σ(x) = max(0, x), and the output layer uses the normalized exponential function Output the probability distribution vector Y of 10 predefined fault types.

[0239] The mathematical expression of the model is:

[0240] H1=σ(W1X+b1);

[0241] H2=σ(W2H1+b2);

[0242]

[0243] H5=σ(W5H4+b5);

[0244] Y = softmax(W6H5+b6);

[0245] Among them, W i ,b i is the weight matrix and bias vector of the i-th layer.

[0246] The model was trained using a small batch gradient descent method with a batch size of 64. The Adam optimizer was used to update the parameters with a learning rate of 0.001. When the performance of the validation set did not improve for 5 consecutive epochs, the training was stopped and the optimal model parameters were saved.

[0247] The stability detection equation in step S90 is specifically described as:

[0248]

[0249] The meanings of the variables are as follows:

[0250] 1.I stab: The final output HPLC dual-mode communication stability index has a value range of [0,1]. The closer the value is to 1, the more stable the system is.

[0251] 2.p k : The probability value of the kth fault type comes from the output layer of the neural network, with a value range of [0,1], and a total of 10 fault types;

[0252] 3.ζ k : The weight coefficient of the kth fault type, a preset constant value, ranging from [0.5, 2], reflects the impact of each fault type on system stability;

[0253] 4.ψ i : The weight coefficient of the i-th evaluation matrix (signal strength, signal-to-noise ratio, bit error rate matrix), preset constant value, value range [0.1,1];

[0254] 5. χ i : The standard deviation of the i-th evaluation matrix, which is used to characterize the fluctuation degree of the evaluation index;

[0255] 6.v j : The rate of change of the system state at the jth time point, which is the difference between two consecutive stability indices;

[0256] 7. The number of sampling points of the state change rate, the default value is 100;

[0257] Formula explanation: Numerator Calculate the weighted sum of the failure probabilities; the denominator Represents the weighted volatility of the evaluation matrix; the exponential term Reflects the temporal stability of the system state;

[0258] Here are the empirical values ​​of the two parameters: k =[0.5,0.6,0.7,0.8,1.0,1.2,1.4,1.6,1.8,2.0];ψ i =[1.0,0.8,0.5].

[0259] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned HPLC dual-mode communication stability method.

[0260] A third aspect of the present invention provides an HPLC dual-mode communication fault detection system, comprising the above-mentioned computer-readable storage medium.

[0261] Specifically, the principle of the present invention is:

[0262] 1. Dual-band carrier signal decomposition principle: The present invention uses the empirical mode decomposition (EMD) method to adaptively decompose the dual-band carrier signal. The EMD method is based on the time scale characteristics of the signal itself, and decomposes the complex signal into a finite number of intrinsic mode functions (IMFs) and a residual component through an iterative screening process. For HPLC carrier signals, the low-frequency IMF component mainly contains noise information, the intermediate-frequency IMF component reflects the dynamic change characteristics of the signal, and the residual component represents the stable trend of the signal. This decomposition method does not require a preset basis function and can effectively process nonlinear and non-stationary signals.

[0263] 2. Principle of environmental impact analysis: The relationship model between the carrier variation component and external environmental factors is established through the multivariate linear regression method. Each element of the regression coefficient matrix quantifies the influence of factors such as electromagnetic interference, ambient temperature, and operating voltage on the carrier signal. Based on this model, the impact of environmental changes on communication quality can be accurately evaluated, providing an important reference for fault diagnosis. At the same time, the singular value decomposition method is used to extract the stable characteristics of the carrier signal, and the abnormal fluctuation pattern is identified by analyzing the distribution characteristics of the eigenvalues.

[0264] 3. Multi-dimensional feature fusion principle: The present invention designs a set of multi-dimensional feature extraction and fusion mechanisms. At the physical layer, an evaluation matrix is ​​constructed by calculating indicators such as signal strength, signal-to-noise ratio, and bit error rate; at the protocol layer, an abnormal feature matrix is ​​constructed based on the data frame structure and verification mechanism. These features are fused through an embedded mathematical model, which contains five core equations: signal quality evaluation equation, protocol anomaly detection equation, carrier stability analysis equation, environmental impact assessment equation, and feature fusion equation. Each equation models a specific fault feature and ultimately generates a unified feature vector.

[0265] 4. Principle of deep learning diagnosis: A five-layer fully connected neural network is used to realize intelligent fault diagnosis. The input layer of the network receives a 216-dimensional fused feature vector and extracts deep feature patterns through multi-layer nonlinear transformations. The hidden layer uses the rectified linear unit (ReLU) activation function to effectively avoid the gradient vanishing problem. The output layer uses the Softmax function to map the results into the probability distribution of 10 fault types. The network optimizes parameters through the back-propagation algorithm and uses the Adam optimizer to adaptively adjust the learning rate to ensure that the model converges to the optimal solution.

[0266] The technical solution of the present invention has strict logic in both theory and practice. First, the dual-band carrier technology provides channel redundancy and enhances the anti-interference ability of the system; second, the multi-dimensional feature fusion mechanism comprehensively captures various manifestations of faults; finally, the adaptive learning ability of the deep learning model ensures the accuracy and real-time nature of fault diagnosis. These technical innovations support each other and are organically unified, together forming a complete fault diagnosis solution.

[0267] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: This embodiment takes the power consumption information collection system in a certain city's intelligent power distribution network as an application scenario, and describes in detail the specific implementation process of the present invention. The power distribution network covers an area of ​​about 12 square kilometers, including 8 distribution substations, managing a total of 2,000 residential users and 150 industrial and commercial users. The system uses HPLC communication technology for data collection, and daily collected data includes information such as power consumption, voltage quality, and load curves. The specific implementation process and experimental results of the method of the present invention are described in detail below.

[0268] 1. The hardware configuration used in this embodiment is shown in Table 1 below:

[0269] Table 1 Hardware Configuration Table

[0270]

[0271] 2. Specific implementation steps

[0272] Data acquisition configuration. The system sampling frequency is set to 1000 Hz, the sampling time is 1 second, and each data contains 1000 sampling points. The collected data items include: HPLC low-frequency carrier signal (1 MHz to 10 MHz), HPLC high-frequency carrier signal (10 MHz to 30 MHz), working voltage (rated 220 volts), working current (range 0 to 30 amps), ambient temperature (range 0 to 60 degrees Celsius), electromagnetic interference intensity (range minus 100 dBm to 0 dBm). The collected data is synchronously stored according to a unified timestamp, and the data storage format is binary, with each sampling point occupying 4 bytes.

[0273] Carrier signal decomposition processing. The acquired dual-band carrier signal is processed using the empirical mode decomposition method, with the maximum number of iterations set to 1000 and the residual threshold set to 0.0001. For the low-frequency carrier signal, the first IMF component contains 85% of the high-frequency noise component, the second IMF component reflects the main change characteristics of the signal, and the remaining part is used as a stable component. For the high-frequency carrier signal, the same decomposition method is used, but the threshold is set to 0.00015 to adapt to the higher frequency characteristics.

[0274] Environmental impact analysis configuration. The system uses a sliding window method to perform multiple linear regression analysis. The window length is set to 100 milliseconds and the sliding step is 10 milliseconds. In the regression analysis, the electromagnetic interference intensity weight coefficient is set to 0.4, the ambient temperature weight coefficient is 0.3, and the operating voltage weight coefficient is 0.3. The regression coefficient is calculated by the least squares method and used to construct a 3x3 dimensional coefficient of variation matrix.

[0275] Evaluation matrix construction. The system constructs three 6x6-dimensional evaluation matrices based on the following parameters: Signal strength evaluation uses negative 85 dBm as the reference value, and the matrix elements are formed by calculating the ratio of the signal to the reference value; Signal-to-noise ratio evaluation uses 15 dB as the threshold, and calculates the logarithmic ratio of signal power to noise power; Bit error rate evaluation is based on the number of frame errors that occur per minute, and the threshold is set to one thousandth. The update period of each matrix is ​​100 milliseconds.

[0276] Neural network model configuration. A five-layer fully connected neural network structure is used, with the number of nodes in each layer being: 216 nodes in the input layer (corresponding to the feature vector dimension), 512 nodes in the first hidden layer, 256 nodes in the second hidden layer, 128 nodes in the third hidden layer, and 10 nodes in the output layer (corresponding to 10 types of faults). The activation function uses ReLU, the initial value of the learning rate is set to 0.001, and the Adam optimizer is used for parameter update. The batch size is set to 64, and the number of training rounds is 500.

[0277] 3. Experimental verification results: In the actual operation process, the system conducted a three-month fault detection experiment on 2,000 residential users and 150 industrial and commercial users. The experimental results are shown in Table 2 below:

[0278] Table 2 Fault detection experiment results

[0279] Fault type Occurrence Detection accuracy Average response time Low frequency carrier signal lost 156 times 96.8％ 15 ms High frequency carrier signal is lost 143 times 95.7％ 16 ms Dual carrier synchronization failed 89 times 94.2％ 18 ms Low frequency channel quality deterioration 267 times 93.5％ 17 ms High frequency channel quality degradation 245 times 92.8％ 17 ms Communication protocol parsing error 178 times 94.6％ 16 ms Data frame check failed 312 times 95.9％ 15 ms Carrier signal interference is serious 198 times 93.7％ 18 ms Device response timeout 167 times 94.8％ 19 ms Multipath Fading Failure 134 times 92.4％ 18 ms

[0280] The system operation status monitoring results are shown in Table 3 below:

[0281] Table 3 System operation monitoring table

[0282] Operational indicators average value Maximum Minimum Standard Deviation CPU usage 45.6％ 78.2％ 22.4％ 8.7％ Memory usage 62.3％ 85.6％ 38.9％ 7.2％ Storage space usage 58.7％ 76.4％ 42.1％ 6.8％ Communication bandwidth utilization 42.8％ 68.9％ 25.6％ 9.4％ System response time 16.8ms 19.9ms 14.2ms 1.2ms

[0283] The environmental parameter monitoring results are shown in Table 4 below:

[0284] Table 4 Environmental parameter monitoring table

[0285] Environmental parameters average value Maximum Minimum Coefficient of variation Ambient temperature 28.5℃ 45.2℃ 15.6℃ 0.156 Electromagnetic interference intensity -75dBm -45dBm -95dBm 0.183 Operating voltage 220.5V 235.8V 208.4V 0.042 Working current 12.6A 25.8A 3.2A 0.276

[0286] 4. Experimental results analysis: After three months of actual operation test, this embodiment has achieved the following specific effects:

[0287] Fault detection performance. The system has an average detection accuracy of 94.4% for 10 predefined fault types, with the highest detection accuracy of 96.8% (low-frequency carrier signal loss) and the lowest detection accuracy of 92.4% (multipath fading fault). The system has an average response time of 16.8 milliseconds, which fully meets the 20 millisecond response time requirement. The system shows good stability under different working conditions, and the standard deviation of the detection accuracy is only 1.5%.

[0288] 5. Stability evaluation: The system will calculate the stability index every 1 minute and set the weight coefficients for the 10 predefined fault types as shown in Table 5 below:

[0289] Table 5 Weight coefficient table

[0290] Fault type <![CDATA[Weight coefficient (ζ k )]]> Low frequency carrier signal lost 2.0 High frequency carrier signal is lost 1.8 Dual carrier synchronization failed 1.6 Low frequency channel quality deterioration 1.4 High frequency channel quality degradation 1.2 Communication protocol parsing error 1.0 Data frame check failed 0.8 Carrier signal interference is serious 0.7 Device response timeout 0.6 Multipath Fading Failure 0.5

[0291] The statistical results of the stability index during system operation are shown in Table 6 below:

[0292] Table 6 Stability calculation results

[0293] Statistical indicators Numeric Average stability index 0.865 Maximum stability index 0.978 Minimum stability index 0.682 Standard Deviation 0.047

[0294] It can be seen that the stability index remains above 0.8 for more than 90% of the operating time; when the stability index is lower than 0.7, a system warning will be triggered.

[0295] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A HPLC dual-mode communication stability method, characterized in that: The following steps are involved: S10, real-time data collection, all collected data are stored in the form of one-dimensional arrays, the sampling frequency is 1000 Hz, each one-dimensional array contains 1000 sampling point data; specifically collect the following data: HPLC low-frequency carrier signal data, HPLC high-frequency carrier signal data, working voltage data, working current data, ambient temperature data, electromagnetic interference data; S20, using an empirical mode decomposition method to decompose the collected HPLC low-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC low-frequency carrier stable component data array, an HPLC low-frequency carrier variable component data array, and an HPLC low-frequency carrier noise component data array; S30, using an empirical mode decomposition method to decompose the collected HPLC high-frequency carrier signal data to obtain three one-dimensional arrays: an HPLC high-frequency carrier stable component data array, an HPLC high-frequency carrier variable component data array, and an HPLC high-frequency carrier noise component data array; S40, establishing the relationship between the HPLC low-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data by a multivariate linear regression method, and obtaining a 3-row and 3-column HPLC low-frequency carrier variation coefficient matrix, wherein each element in the HPLC low-frequency carrier variation coefficient matrix represents the influence weight of the corresponding factor; Based on the HPLC low-frequency carrier variation coefficient matrix, a 3-row and 3-column HPLC low-frequency carrier stability coefficient matrix is ​​calculated using a singular value decomposition method; S50, using a multivariate linear regression method to establish a relationship between the HPLC high-frequency carrier variation component data array and the electromagnetic interference data array, the ambient temperature data, and the operating voltage data, to obtain a 3-row 3-column HPLC high-frequency carrier variation coefficient matrix, wherein each element in the HPLC high-frequency carrier variation coefficient matrix represents an influence weight of a corresponding factor; Based on the HPLC high-frequency carrier variation coefficient matrix, a 3-row and 3-column HPLC high-frequency carrier stability coefficient matrix is ​​calculated using a singular value decomposition method; S60, based on the HPLC low-frequency carrier stability coefficient matrix and the HPLC high-frequency carrier stability coefficient matrix, respectively calculate three 6-row and 6-column evaluation matrices: a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, and a bit error rate evaluation matrix, wherein the signal strength evaluation matrix is ​​obtained by calculating the root mean square value of the stable component data array, the signal-to-noise ratio evaluation matrix is ​​obtained by calculating the power ratio of the stable component data array to the noise component data array, and the bit error rate evaluation matrix is ​​calculated by analyzing the number of fluctuations exceeding a preset threshold group in the variable component data array; S70, based on the data frame structure, check code type, and timing parameter requirements defined in the communication protocol specification, construct a 12-row and 12-column dual-carrier communication protocol abnormal feature matrix, wherein the data frame structure includes an 8-byte frame header, a data segment length, and a 4-byte frame tail, and the check code type is a 32-bit cyclic redundancy check code; S80, adopting an embedded five-layer fully connected neural network structure fault diagnosis model, inputting a signal strength evaluation matrix, a signal-to-noise ratio evaluation matrix, a bit error rate evaluation matrix and a dual-carrier communication protocol abnormal feature matrix into the fault diagnosis model, the number of input layer nodes of the fault diagnosis model is all elements of the evaluation matrix and the abnormal feature matrix, the hidden layer of the fault diagnosis model uses a rectified linear unit activation function, and the output layer of the fault diagnosis model uses a normalized exponential function to output probability distribution vectors of 10 predefined fault types; S90, input each predefined fault type and its probability distribution vector into a predefined stability detection equation, obtain the stability index of HPLC dual-mode communication and output it.

2. The HPLC dual-mode communication fault detection system according to claim 2, characterized in that: The 10 predefined fault types include low-frequency carrier signal loss, high-frequency carrier signal loss, dual-carrier synchronization failure, low-frequency channel quality degradation, high-frequency channel quality degradation, communication protocol parsing error, data frame check failure, severe carrier signal interference, device response timeout, and multipath fading failure.

3. The HPLC dual-mode communication fault detection system according to claim 2, characterized in that: The preset threshold group specifically includes: signal strength threshold -85dBm, signal-to-noise ratio threshold 15dB, bit error rate threshold 1×10 -3 , carrier signal fluctuation threshold ±10%, protocol analysis error threshold 3 times / minute.

4. The HPLC dual-mode communication fault detection system according to claim 2, characterized in that: The frequency range of the HPLC low-frequency carrier signal is 1 MHz to 10 MHz, and the frequency range of the HPLC high-frequency carrier signal is 10 MHz to 30 MHz.

5. The HPLC dual-mode communication fault detection system according to claim 2, characterized in that: The timing parameter requirements include a response timeout threshold of 20 milliseconds and a synchronization code recognition window of 10 milliseconds.

6. The HPLC dual-mode communication fault detection system according to claim 2, characterized in that: Each row of the dual-carrier communication protocol abnormal feature matrix corresponds to frame header loss, frame tail error, frame length abnormality, cyclic redundancy check failure, check bit error, response timeout, and synchronization failure.

7. The methods for acquiring the HPLC low-frequency carrier signal data, HPLC high-frequency carrier signal data, working voltage data, working current data, ambient temperature data, and electromagnetic interference data are respectively to acquire the HPLC low-frequency carrier signal data by real-time sampling of the carrier signal in the frequency band of 1-10MHz by the ADC sampler in the radio frequency front-end module, to acquire the HPLC high-frequency carrier signal data by real-time sampling of the carrier signal in the frequency band of 10-30MHz by the ADC sampler in the radio frequency front-end module, to acquire the working voltage data by real-time acquisition of the system working voltage by the voltage sampling chip AD7606, to acquire the working current data by real-time acquisition of the system working current by the Hall current sensor ACS712, to acquire the ambient temperature data by real-time acquisition of the ambient temperature by the DS18B20 digital temperature sensor, and to acquire the electromagnetic interference data by real-time acquisition of the ambient electromagnetic interference intensity by the electromagnetic field strength detection module EM100.

8. The structure of the fault diagnosis model specifically includes a data preprocessing layer, a feature extraction layer, an embedded mathematical model, a deep learning network layer and a decision output layer, wherein the data preprocessing layer is responsible for standardizing and denoising the input data, the feature extraction layer extracts key features from the original data, the embedded mathematical model serves as the core mathematical equation group processing layer, the deep learning network layer adopts a five-layer fully connected neural network, and the decision output layer generates the final fault diagnosis result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the HPLC dual-mode communication stability method according to any one of claims 1 to 8.

10. It is characterized in that it comprises the computer-readable storage medium described in claim 9.

Images