Building water pipe fault diagnosis method and system based on data analysis

Abstract

The invention relates to the technical field of fault detection, in particular to a building water pipe fault diagnosis method and system based on data analysis, and the method comprises the following steps: segmenting a signal according to a time sequence based on a water pipe vibration signal collected in a building environment, and calculating the amplitude change rate and the power spectral density peak value of each segment; and extracting a paragraph whose power spectral density peak value exceeds a threshold value. According to the method, the frequency range is analyzed point by point, the power spectrum proportion and the amplitude change standard deviation are calculated, accurate capture of frequency component change is enhanced, the decomposition result is optimized by adjusting parameters, the signal segment with the drift rate exceeding the reference value is extracted, the amplitude and time window correlation analysis is combined, dynamic characteristic signals are comprehensively captured, and the accuracy of the dynamic characteristic signals is improved. Complete expression of signal features is enhanced, through weighted integration of the frequency drift rate and instantaneous energy change, key features are extracted, fusion vectors are generated, classification accuracy is improved, classification boundaries are dynamically determined through the distance between the features and the weight ratio, and accurate classification of faults is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to a method and system for diagnosing building water pipe faults based on data analysis. Background Art

[0002] The field of fault detection technology mainly studies how to detect and diagnose abnormal conditions, faults or potential problems in real time by monitoring the status and performance of system operation. This technical field usually covers a variety of means such as signal acquisition, data analysis, pattern recognition and intelligent algorithms to improve the reliability, stability and safety of the system. Fault detection technology is widely used in industrial equipment, transportation, power systems, building facilities and other fields to help achieve efficient maintenance and management, reduce the risk of sudden accidents, and extend the service life of equipment.

[0003] Among them, the building water pipe fault diagnosis method refers to the method of detecting and diagnosing the blockage, leakage, aging and other faults of the water pipes by technical means according to the operating status of the water pipe system in the building facilities. Its purpose is to timely discover and locate water pipe faults, prevent water resource waste, property loss and reduced operating efficiency, and provide efficient and reliable solutions for the maintenance and management of building facilities.

[0004] The existing technology lacks pertinence in extracting features from vibration signals, and is prone to introducing noise interference, resulting in distorted analysis results. The ability to capture changes in frequency components is insufficient, and the processing sensitivity for dynamic signal features is low, making it impossible to accurately identify key features. Key features such as frequency drift rate and instantaneous energy changes are not fully expressed, and the feature fusion effect is poor, resulting in low signal classification accuracy. The method for determining the classification boundary is single, which makes it difficult to cope with the dynamic changes of complex signal features. The reliability of the classification results is low, which increases the misjudgment rate of fault diagnosis and maintenance costs. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a building water pipe fault diagnosis method and system based on data analysis.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a building water pipe fault diagnosis method based on data analysis, comprising the following steps: S1: Based on the water pipe vibration signal collected in the building environment, the signal is segmented according to the time series, the amplitude change rate and power spectrum density peak of each segment are calculated, and the segments with the power spectrum density peak exceeding the threshold are extracted, and the matching signal segments are matched with the reference frequency band to generate the filtered signal segment data; S2: Based on the filtered signal segment data, the signal segment frequency range is analyzed point by point, the power spectrum proportion and the amplitude change standard deviation are calculated, the validity of the decomposition result is judged, the frequency component is filtered according to the change fluctuation value, the decomposition parameter is adjusted, and the frequency component energy proportion data is obtained; S3: Based on the frequency component energy proportion data, the short-time spectrum amplitude and frequency drift rate of each signal are calculated in segments, the segments whose drift rate exceeds the reference value are extracted, the correlation between the amplitude and the time window is analyzed, the characteristic segments that meet the conditions are identified, and the energy distribution characteristic data is generated; S4: Based on the energy distribution characteristic data, perform weighted integration on the frequency drift rate and instantaneous energy change of the characteristic segment, calculate the frequency offset amplitude and drift standard deviation, extract the associated characteristic frequency band, convert the energy change into a signal strength feature, and generate a signal feature fusion vector; S5: Based on the signal feature fusion vector, the frequency drift rate and signal strength change characteristics of the feature vector are normalized, the Euclidean distance and weight ratio between the features are calculated, the classification boundary is determined by the ratio and distance, the classification judgment is completed, and the water pipe fault classification result is generated.

[0007] The screened signal segment data includes segments where the power spectrum density peak exceeds the threshold, signal segments matching the reference frequency band, and signal segment data generated after screening; the frequency component energy proportion data includes significantly changed frequency components, adjusted decomposition parameters, and energy proportion information of frequency components; the energy distribution feature data includes segments where the frequency drift rate exceeds the reference value, short-time spectrum amplitude, and the correlation between the amplitude and the time window; the signal feature fusion vector includes frequency offset amplitude, drift standard deviation, and signal strength characteristics; the water pipe fault classification result includes classification boundaries, Euclidean distance between features, and weight proportion.

[0008] As a further solution of the present invention, the step of obtaining the screening signal segment data is specifically as follows: S111: calculating the water pipe vibration signal in segments according to the time series, obtaining the amplitude change rate for each signal segment, and comparing it with the set amplitude change rate threshold, filtering the signal segment whose amplitude change rate exceeds the threshold, and obtaining the signal segment after the amplitude change rate filtering; S112: performing power spectrum analysis on the signal segments after the amplitude change rate screening, extracting the power spectrum density curve of the signal segments, obtaining the power spectrum density peak value of each signal segment, and establishing a power spectrum density peak value set; S113: Screen each peak in the power spectrum density peak set and match it with the reference frequency band, and compare whether the power spectrum density peak falls within the range of the reference frequency band, using the formula: ; Calculate the weight of the matching signal in the reference frequency band, filter the signal segment that highly matches the reference frequency band, and generate filtered signal segment data; in, represents the weight of the matching signal segment, is the power spectral density value of each signal segment, is the peak frequency of the power spectrum of the signal segment, is the reference frequency center value, is the reference frequency bandwidth, is the total number of signal segments to be filtered.

[0009] As a further solution of the present invention, the step of acquiring the frequency component energy proportion data is specifically as follows: S211: analyzing the frequency range of the signal segments point by point from the filtered signal segment data, calculating the power spectrum proportions of multiple signal segments, extracting data within the frequency range through frequency analysis, and obtaining frequency range analysis data; S212: Analyzing the data within the frequency range, calculating the amplitude variation standard deviation of each signal segment, evaluating the stability and variability of the signal segment, screening signal segments with high stability, and obtaining amplitude variation standard deviation screening results; S213: Apply the change fluctuation value screening to the amplitude change standard deviation screening result, iteratively adjust the decomposition parameters, and screen the signal segments with low volatility and obvious frequency characteristics, using the formula: ; Calculate and generate frequency component energy ratio data; in, Represents the energy proportion of the frequency component, is the standard deviation of the amplitude variation of the signal segment, is the actual frequency of the signal segment, is the reference frequency, is the frequency sensitivity coefficient, is the number of signal segments.

[0010] As a further solution of the present invention, the step of acquiring the energy distribution characteristic data is specifically as follows: S311: using the frequency component energy proportion data, calculating the short-time spectrum amplitude of each signal in segments, verifying the spectrum distribution of the signal in the short-time window, and obtaining short-time spectrum amplitude data; S312: Calculate the frequency drift rate of the short-time spectrum amplitude data, compare the frequency drift rates of multiple signal segments with a preset reference value, select the signal segment whose drift rate exceeds the reference value, and obtain the frequency drift signal data exceeding the reference value; S313: Analyze the correlation between the super-reference value frequency drift signal data and the time window, identify the signal segment that meets the target time window change pattern, and use the formula: ; Calculate the energy distribution of each characteristic segment that meets the conditions and generate energy distribution characteristic data; in, Represents the energy distribution characteristics, is the short-time spectrum amplitude, is the frequency drift rate, is the reference drift rate, Adjust the parameter for drift rate sensitivity, is the time window length, is the number of feature segments.

[0011] As a further solution of the present invention, the step of acquiring the signal feature fusion vector is specifically as follows: S411: Based on the energy distribution characteristic data, perform weighted integral calculation of instantaneous energy and frequency drift rate for each characteristic segment, determine the frequency offset amplitude and instantaneous energy of multiple characteristic segments, and obtain frequency offset and energy change data; S412: Analyze the frequency offset and energy change data, calculate the drift standard deviation of each characteristic segment, evaluate the stability of the characteristic segment, iteratively screen the characteristic segments with high stability, and obtain the drift standard deviation screening result; S413: extracting a signal segment matching the associated characteristic frequency band from the drift standard deviation screening result, and converting the energy change of the signal segment into a signal strength feature, using the formula: ; Calculate the feature vectors of multiple signal segments and synthesize them into a signal feature fusion vector; in, represents the signal feature fusion vector, is the instantaneous energy of the multi-signal segment, is the signal strength, is the mean value of the drift standard deviation, is the number of signal segments.

[0012] As a further solution of the present invention, the steps of obtaining the water pipe fault classification result are specifically as follows: S511: Based on the signal feature fusion vector, normalize the frequency drift rate and signal strength change characteristics, calculate the Euclidean distance between multiple vectors using the normalized eigenvalue of each vector, and generate a Euclidean distance matrix between normalized eigenvectors; S512: Combine the Euclidean distance matrix to calculate the weight ratio of multiple vectors, adjust the weight according to the normalized eigenvalue and distance distribution, select key eigenvectors, and use the formula: ; Calculate the weighted value of the feature vector and determine the classification boundary to generate weight proportion data; in, is the weighted value of the vector, For vector With vector The Euclidean distance of For vector The feature weights of For vector The weight distribution of is the total number of vectors; S513: Based on the weighted proportion data, combined with the Euclidean distance and the weighted value, the classification determination is completed by determining the correlation between the feature vector distribution and the classification boundary, and a water pipe fault classification result is generated.

[0013] A building water pipe fault diagnosis system based on data analysis, the building water pipe fault diagnosis system based on data analysis is used to execute the above-mentioned building water pipe fault diagnosis method based on data analysis, the system comprises: The vibration signal segmentation module collects water pipe vibration signals in the building environment, segments the signals according to time series, calculates the amplitude change rate and power spectrum density peak, filters the signal segments that match the reference frequency band, and generates filtered signal segment data; The frequency component screening module analyzes the frequency range of the signal segment based on the screening signal segment data, calculates the power spectrum proportion and the standard deviation of the amplitude change, screens the significantly changed frequency components, adjusts the decomposition parameters, extracts the component energy proportion, and generates the frequency component energy proportion data; The feature segment extraction module calculates the signal short-time spectrum amplitude and frequency drift rate based on the frequency component energy proportion data, extracts the drift rate exceeding the reference value segment, analyzes the correlation between the amplitude and the time window, identifies the qualified frequency segment, and generates energy distribution feature data; The feature fusion processing module is based on the energy distribution feature data, weighted integral frequency drift rate and instantaneous energy change, calculates the frequency offset amplitude and drift standard deviation, statistically correlates the energy change of the feature frequency band, converts the signal strength feature, and generates a signal feature fusion vector; The fault classification and determination module normalizes the frequency drift rate and signal strength change characteristics based on the signal feature fusion vector, calculates the Euclidean distance and weight ratio between the features, determines the classification boundary to complete the classification and determination, and generates the water pipe fault classification result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by segmenting the vibration signal and extracting the amplitude change rate and power spectrum density peak, the interference of invalid signals is reduced and the pertinence of signal feature extraction is significantly improved. The frequency range is analyzed point by point and the power spectrum ratio and amplitude change standard deviation are calculated to enhance the accurate capture of frequency component changes, and the decomposition results are optimized by adjusting parameters. The signal segments with drift rates exceeding the reference value are extracted, and the dynamic characteristic signals are fully captured by combining the amplitude and time window correlation analysis, thereby enhancing the complete expression of signal characteristics. Through the weighted integration of the frequency drift rate and instantaneous energy change, key features are extracted and a fusion vector is generated to improve classification accuracy. The classification boundary is dynamically determined by the distance and weight ratio between features to achieve accurate classification of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 A flow chart of the steps for obtaining the screening signal segment data of the present invention; Figure 3 A flow chart of the steps for obtaining frequency component energy ratio data of the present invention; Figure 4 A flow chart of the steps for obtaining energy distribution characteristic data of the present invention; Figure 5 Flow chart of the steps of obtaining the signal feature fusion vector of the present invention; Figure 6 This is a flow chart of the steps for obtaining the water pipe fault classification results of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a building water pipe fault diagnosis method based on data analysis, comprising the following steps: S1: Based on the water pipe vibration signal collected in the building environment, the signal is segmented according to the time series, the amplitude change rate and power spectrum density peak of each segment are calculated, and the segments with the power spectrum density peak exceeding the threshold are extracted, and the matching signal segments are matched with the reference frequency band to generate the filtered signal segment data; S2: Based on the filtered signal segment data, analyze the signal segment frequency range point by point, calculate the power spectrum proportion and amplitude change standard deviation, judge the validity of the decomposition result, filter the frequency components according to the change fluctuation value, adjust the decomposition parameters, and obtain the frequency component energy proportion data; S3: Based on the frequency component energy proportion data, the short-time spectrum amplitude and frequency drift rate of each signal are calculated segment by segment, the segments with drift rates exceeding the reference value are extracted, the correlation between the amplitude and the time window is analyzed, the characteristic segments that meet the conditions are identified, and the energy distribution characteristic data is generated; S4: Based on the energy distribution feature data, perform weighted integration on the frequency drift rate and instantaneous energy change of the feature segment, calculate the frequency offset amplitude and drift standard deviation, extract the associated feature frequency band, convert the energy change into signal strength features, and generate a signal feature fusion vector; S5: Based on the signal feature fusion vector, the frequency drift rate and signal strength change characteristics of the feature vector are normalized, the Euclidean distance and weight ratio between the features are calculated, the classification boundary is determined by the ratio and distance, the classification judgment is completed, and the water pipe fault classification result is generated.

[0019] The filtered signal segment data include segments where the power spectral density peak exceeds the threshold, signal segments matching the reference frequency band, and signal segment data generated after filtering. The frequency component energy proportion data include significantly changed frequency components, adjusted decomposition parameters, and frequency component energy proportion information. The energy distribution feature data include segments where the frequency drift rate exceeds the reference value, short-time spectrum amplitude, and the correlation between the amplitude and time window. The signal feature fusion vector includes frequency offset amplitude, drift standard deviation, and signal strength characteristics. The water pipe fault classification results include classification boundaries, Euclidean distance between features, and weight proportion.

[0020] See also Figure 2 , the specific steps for obtaining the filtering signal segment data are: S111: calculating the water pipe vibration signal in segments according to the time series, obtaining the amplitude change rate for each signal segment, and comparing it with the set amplitude change rate threshold, filtering the signal segment whose amplitude change rate exceeds the threshold, and obtaining the signal segment after the amplitude change rate filtering; After the water pipe vibration signal is segmented, the amplitude change rate of each signal segment is calculated and compared with the preset threshold to screen out the signal segments that exceed the threshold. This process relies on the monitoring of the signal amplitude and the accurate recording of the data. The amplitude data of each signal segment is acquired in real time by the sensor, and the amplitude change rate is calculated through digital signal processing technology. Through comparative analysis, those signal segments with a change amplitude greater than the preset threshold are screened out. These data will be used as the input of the next power spectrum analysis for further signal screening and analysis.

[0021] S112: performing power spectrum analysis on the signal segment after the amplitude change rate is filtered, extracting the power spectrum density curve of the signal segment, obtaining the power spectrum density peak value of each signal segment, and establishing a power spectrum density peak value set; This step involves complex mathematical operations, including Fourier transform to obtain the signal's spectrum. The power spectral density of each signal segment is calculated and recorded. Based on this, the maximum power spectral density peak of each signal segment is extracted. These peaks constitute a power spectral density peak set. This data set is one of the key bases for screening signal segments and provides necessary data support for subsequent frequency band matching screening.

[0022] S113: Screen each peak in the power spectrum density peak set and match it with the reference frequency band, and compare whether the power spectrum density peak falls within the range of the reference frequency band, using the formula: ; Calculate the weight of the matching signal in the reference frequency band, filter the signal segment that highly matches the reference frequency band, and generate filtered signal segment data; in, represents the weight of the matching signal segment, is the power spectral density value of each signal segment, is the peak frequency of the power spectrum of the signal segment, is the reference frequency center value, is the reference frequency bandwidth, is the total number of signal segments to be filtered.

[0023] formula: ; The formula is useful in that it takes into account the difference between the power spectral density of the signal and the center of the reference band, and by adjusting the coefficients To control the influence of bandwidth, the sensitivity and accuracy of signal screening are enhanced to adapt to different application scenarios and needs.

[0024] Detailed explanation of the formula and the process of formula calculation and derivation: Assume there are 5 signal segments, among which (Power spectral density value) is , (Power spectrum peak frequency) is MHz, reference frequency center For 60MHz, adjust the coefficient For 5: ; ; ; The result shows that the calculated average matching signal weight is 21.2, which represents a quantitative expression of the matching degree between the screened signal segment and the reference frequency band, helping decision makers understand the degree of consistency between the signal segment and the set frequency band. Further deduction can be used to derive the signal segment that meets the reference frequency band.

[0025] See also Figure 3 , the specific steps for obtaining the frequency component energy ratio data are: S211: analyzing the frequency range of the signal segments point by point from the filtered signal segment data, calculating the power spectrum proportions of multiple signal segments, extracting data within the frequency range through frequency analysis, and obtaining frequency range analysis data; First, a point-by-point frequency analysis is performed on each signal segment from the screened signal segment data. The characteristic frequency distribution of the signal is obtained by calculating the power spectrum ratio of each signal segment. This frequency distribution provides insights into the signal energy distribution in different frequency ranges. The power spectrum ratio is calculated by Fourier transform of the signal, and the frequency points reflect the main energy components of the signal. By analyzing these frequency points, the basic structure and possible physical sources of the signal can be understood. These frequency distribution data serve as the basic input for subsequent steps for further amplitude change analysis and stability assessment. This process is achieved through advanced signal processing technology and does not rely on specific algorithms or software. Instead, it uses basic signal processing theory to interpret and process data.

[0026] S212: Analyzing the data in the frequency range, calculating the amplitude change standard deviation of each signal segment, evaluating the stability and variability of the signal segment, screening the signal segment with high stability, and obtaining the amplitude change standard deviation screening result; The stability and consistency of the signal are evaluated by measuring the fluctuation of the signal amplitude. Calculating the standard deviation is a common method in statistics to describe the degree of discreteness of data distribution. It is applied here to the signal amplitude in time series data. The calculation process includes obtaining the average amplitude of each signal segment, and then calculating the square of the difference between each data point and the average value. Finally, these square differences are summed and divided by the total number of data points to obtain the variance. The square root of the variance is the standard deviation. This data is crucial for screening stable signal segments, because highly stable signal segments are more likely to reflect the essential characteristics of the physical process in the subsequent frequency component analysis, rather than noise or other interference.

[0027] S213: Apply the change fluctuation value screening to the amplitude change standard deviation screening results, iteratively adjust the decomposition parameters, and screen the signal segments with low volatility and obvious frequency characteristics, using the formula: ; Calculate and generate frequency component energy ratio data; in, Represents the energy proportion of the frequency component, is the standard deviation of the amplitude variation of the signal segment, is the actual frequency of the signal segment, is the reference frequency, is the frequency sensitivity coefficient, is the number of signal segments.

[0028] formula: ; The benefit of the formula is that it can weight signal segments according to the standard deviation of amplitude variation and frequency deviation, thereby optimizing the energy calculation of frequency components and improving the accuracy and reliability of frequency analysis.

[0029] Detailed explanation of the formula and the process of formula calculation and derivation: Determine the number of signal segments , assuming it is 5; set the reference frequency is 60Hz, the frequency sensitivity coefficient Set to 10; the actual frequency of each signal segment They are 55Hz, 60Hz, 65Hz, 70Hz, and 75Hz respectively; the standard deviation of the amplitude change of each signal segment Set to 2, 2.5, 1.5, 1.8, 1.3. Calculate the energy proportion of each frequency component and calculate the energy proportion of each item. , and then calculate the overall weighted average: ; ; ; ; ; The overall weighted average is then calculated as: ; The result shows that the calculated frequency component energy ratio reflects the energy concentration of the signal near the reference frequency, which is a comprehensive quantitative expression after weight adjustment and effectively describes the main energy distribution of the signal.

[0030] See also Figure 4, the specific steps for obtaining energy distribution characteristic data are: S311: using the frequency component energy proportion data, calculating the short-time spectrum amplitude of each signal in segments, verifying the spectrum distribution of the signal in the short-time window, and obtaining the short-time spectrum amplitude data; First, the input signal segment data needs to be preprocessed, including denoising and standardization, to ensure the accuracy of the analysis. Then, the short-time spectrum amplitude of each signal segment is calculated. This step requires converting the time domain representation of the signal to the frequency domain through fast Fourier transform (FFT) to extract the spectrum amplitude. This process also includes the application of window function, usually Hamming window or Hanning window is selected to reduce spectrum leakage. After FFT transformation, the amplitude of the obtained spectrum data is calculated to ensure that the amplitude of each spectrum component can correctly reflect the energy distribution of the signal. The obtained short-time spectrum amplitude data will be used for the next step of frequency drift rate calculation.

[0031] S312: Calculate the frequency drift rate of the short-time spectrum amplitude data, compare the frequency drift rates of multiple signal segments with a preset reference value, select the signal segment whose drift rate exceeds the reference value, and obtain the frequency drift signal data exceeding the reference value; This step involves measuring the frequency change between two consecutive frames and determining the frequency drift rate by calculating the slope of the frequency change. The specific calculation formula is the frequency change divided by the time interval. This calculation requires that the frequency data must be aligned on the same time base to ensure the consistency of each calculation. The frequency drift rate of each signal segment is compared with the preset reference value, and the signal segments that exceed the reference value are screened out through the threshold judgment method. These signal segments are marked as high-risk signal segments and require further analysis and processing. This screening method helps to quickly locate those signal segments with abnormal frequency changes.

[0032] S313: Analyze the correlation between the frequency drift signal data of the super-reference value and the time window, identify the signal segment that meets the change pattern of the target time window, and use the formula: ; formula: ; The benefit of the formula is that, through the weighted combination of amplitude and frequency drift rate, it can effectively distinguish the key features in the signal and associate these features with the changes in the time window, thereby enhancing the reliability of the signal's time-frequency feature analysis, which is particularly important for monitoring and fault diagnosis of dynamic systems.

[0033] Detailed explanation of the formula and the process of formula calculation and derivation: Assume there are three signal segments, among which They are 50, 60, and 55 respectively. (frequency drift rate) are 0.5, 0.6, 0.4 respectively, (reference frequency drift rate) is 0.5, sensitivity adjustment parameter is 0.1, time window is 1, and we can substitute it into the formula to get: ; ; ; ; ; The results show that by considering the correlation between the amplitude and the time window and the frequency drift rate, the characteristics of the energy distribution in the signal segment can be quantified, providing a method to evaluate the sensitivity of the signal to time changes, which is beneficial for applications in the field of signal processing such as fault detection and health monitoring.

[0034] See also Figure 5 , the steps for obtaining the signal feature fusion vector are as follows: S411: Based on the energy distribution characteristic data, a weighted integral calculation of the instantaneous energy and the frequency drift rate is performed on each characteristic segment to determine the frequency offset amplitude and the instantaneous energy of the multiple characteristic segments, and obtain the frequency offset and energy change data; First, the characteristic segments are weighted integrated to quantify the instantaneous energy and frequency drift rate of each signal segment. This step is achieved through data acquisition and analysis. The data set includes real-time records of the frequency offset amplitude and the calculated value of the instantaneous energy. The weighted integration process takes into account the energy density of the signal and the sensitivity to frequency changes. The energy and frequency drift of the signal are combined through an integration algorithm to calculate the comprehensive performance of each signal segment in a specific time window. This method extracts the key characteristics of each signal segment through in-depth analysis of real-time monitoring data, combined with the comprehensive performance of frequency and energy, thereby obtaining frequency offset and energy change data.

[0035] S412: Analyze the frequency offset and energy change data, calculate the drift standard deviation of each characteristic segment, evaluate the stability of the characteristic segment, iteratively screen the characteristic segment with high stability, and obtain the drift standard deviation screening result; The stability of each signal segment is evaluated by calculating the drift standard deviation. The calculation of the drift standard deviation is based on the statistical analysis of the frequency drift rate and involves the calculation of the standard deviation of the frequency drift data to quantify the frequency stability of the signal during the observation period. This process analyzes the statistical characteristics of the frequency in detail and associates it with the energy data. The calculation results can effectively indicate those signal segments with little frequency change and uniform energy distribution. These signal segments show higher stability. Screening these signal segments provides accurate basic data for subsequent analysis, so as to further extract signal segments that match the associated characteristic frequency bands.

[0036] S413: extract the signal segment matching the associated characteristic frequency band from the drift standard deviation screening result, and convert the energy change of the signal segment into a signal strength feature using the formula: ; Calculate the feature vectors of multiple signal segments and synthesize them into a signal feature fusion vector; in, represents the signal feature fusion vector, is the instantaneous energy of the multi-signal segment, is the signal strength, is the mean value of the drift standard deviation, is the number of signal segments.

[0037] formula: ; The formula is beneficial in that it combines instantaneous energy and the adjusted signal strength , and the mean of the drift standard deviation , which can effectively convert the energy changes of each signal segment into a unified signal strength feature, and then form a comprehensive signal feature fusion vector , this fused vector helps further signal processing and feature analysis.

[0038] Detailed explanation of the formula and the process of formula calculation and derivation: Assume there are three signal segments, and the instantaneous energies are The signal strengths are , drift standard deviation average ,but: ; ; ; The results show that through the weighted integration method, the various characteristic segments of the signal can be effectively fused into a vector, indicating the contribution and importance of different characteristic segments in the total signal, so that subsequent analysis can process these key signal segments more specifically.

[0039] See also Figure 6 ,The specific steps for obtaining the water pipe fault classification results are: S511: based on the signal feature fusion vector, normalize the frequency drift rate and the signal strength change characteristics, calculate the Euclidean distance between the multiple vectors using the normalized eigenvalue of each vector, and generate the Euclidean distance matrix between the normalized eigenvectors; Based on the signal feature fusion vector, normalization is a key step. The purpose of this step is to standardize the frequency drift rate and signal strength change characteristics to ensure comparability between different signals. First, the eigenvalues ​​of each vector are normalized and the Euclidean distance of the normalized eigenvector is calculated. Through the normalization process, each eigenvalue is scaled to the same scale. The normalization process includes the steps of subtracting the mean and dividing by the standard deviation. The Euclidean distance between normalized eigenvectors is calculated based on the formula carried out, among which and represents two different normalized eigenvectors, is the number of features in the vector.

[0040] S512: Combine the Euclidean distance matrix to calculate the weight ratio of multiple vectors, adjust the weight according to the normalized eigenvalue and distance distribution, and select the key eigenvectors using the formula: ; Calculate the weighted value of the feature vector and determine the classification boundary to generate weight proportion data; in, is the weighted value of the vector, For vector With vector The Euclidean distance of For vector The feature weights of For vector The weight distribution of is the total number of vectors; formula: ; The benefit of the formula is that it integrates the relative distance between each feature vector and the importance of each vector through the weighted proportion method, so as to effectively identify the representative feature vectors and then determine the classification boundary.

[0041] Detailed explanation of the formula and the process of formula calculation and derivation: Given the Euclidean distance , weight and eigenvector importance In this case, the calculation steps are as follows: ; Assume that the total number of vectors , and all vectors ,but: ; ; The result shows that the weighted influence of this feature vector in classification judgment is 0.2205, indicating that its importance in the overall classification is relatively low, so it can be used to assist in determining the boundary and optimize the classification effect.

[0042] S513: Based on the weighted proportion data, combined with the Euclidean distance and the weighted value, the classification determination is completed by determining the correlation between the feature vector distribution and the classification boundary, and a water pipe fault classification result is generated.

[0043] The key link in applying theory to practice is to first sort the feature vectors according to the weighted value and Euclidean distance, and then use the sorting results to determine the classification boundaries. This process involves threshold setting techniques in statistical analysis and machine learning algorithms to ensure the optimal classification effect. The threshold used in the classification judgment process is usually adjusted according to the actual data distribution to ensure the accuracy and reliability of the classification results. The generated water pipe fault classification results can be directly applied to the water pipe monitoring system to provide the system with real-time fault detection functions, thereby quickly responding to potential fault problems.

[0044] A building water pipe fault diagnosis system based on data analysis, the building water pipe fault diagnosis system based on data analysis is used to execute the above-mentioned building water pipe fault diagnosis method based on data analysis, the system comprises: The vibration signal segmentation module collects water pipe vibration signals in the building environment, segments the signals according to time series, calculates the amplitude change rate and power spectrum density peak, filters the signal segments that match the reference frequency band, and generates filtered signal segment data; The frequency component screening module is based on screening signal segment data, analyzing the frequency range of the signal segment, calculating the power spectrum proportion and the standard deviation of the amplitude change, screening the significantly changing frequency components, adjusting the decomposition parameters, extracting the component energy proportion, and generating the frequency component energy proportion data; The feature segment extraction module calculates the signal short-time spectrum amplitude and frequency drift rate based on the frequency component energy ratio data, extracts the drift rate exceeding the reference value segment, analyzes the correlation between the amplitude and the time window, identifies the qualified frequency segment, and generates energy distribution feature data; The feature fusion processing module is based on the energy distribution feature data, weighted integral frequency drift rate and instantaneous energy change, calculates the frequency offset amplitude and drift standard deviation, statistically correlates the energy change of the feature frequency band, converts the signal strength feature, and generates a signal feature fusion vector; The fault classification module is based on the signal feature fusion vector, normalized frequency drift rate and signal strength change characteristics, calculated the Euclidean distance and weight ratio between features, determined the classification boundary to complete the classification judgment, and generated the water pipe fault classification result.

[0045] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A building water pipe fault diagnosis method based on data analysis, characterized in that: The following steps are involved: S1: Based on the water pipe vibration signal collected in the building environment, the signal is segmented according to the time series, the amplitude change rate and power spectrum density peak of each segment are calculated, and the segments with the power spectrum density peak exceeding the threshold are extracted, and the matching signal segments are matched with the reference frequency band to generate the filtered signal segment data; S2: Based on the filtered signal segment data, the signal segment frequency range is analyzed point by point, the power spectrum proportion and the amplitude change standard deviation are calculated, the validity of the decomposition result is judged, the frequency component is filtered according to the change fluctuation value, the decomposition parameter is adjusted, and the frequency component energy proportion data is obtained; S3: Based on the frequency component energy proportion data, the short-time spectrum amplitude and frequency drift rate of each signal are calculated in segments, the segments whose drift rate exceeds the reference value are extracted, the correlation between the amplitude and the time window is analyzed, the characteristic segments that meet the conditions are identified, and the energy distribution characteristic data is generated; S4: Based on the energy distribution characteristic data, perform weighted integration on the frequency drift rate and instantaneous energy change of the characteristic segment, calculate the frequency offset amplitude and drift standard deviation, extract the associated characteristic frequency band, convert the energy change into a signal strength feature, and generate a signal feature fusion vector; S5: Based on the signal feature fusion vector, the frequency drift rate and signal strength change characteristics of the feature vector are normalized, the Euclidean distance and weight ratio between the features are calculated, the classification boundary is determined by the ratio and distance, the classification judgment is completed, and the water pipe fault classification result is generated.

2. The building water pipe fault diagnosis method based on data analysis according to claim 1 is characterized in that: The screened signal segment data includes segments where the power spectrum density peak exceeds the threshold, signal segments matching the reference frequency band, and signal segment data generated after screening; the frequency component energy proportion data includes significantly changed frequency components, adjusted decomposition parameters, and energy proportion information of frequency components; the energy distribution feature data includes segments where the frequency drift rate exceeds the reference value, short-time spectrum amplitude, and the correlation between the amplitude and the time window; the signal feature fusion vector includes frequency offset amplitude, drift standard deviation, and signal strength characteristics; the water pipe fault classification result includes classification boundaries, Euclidean distance between features, and weight proportion.

3. The building water pipe fault diagnosis method based on data analysis according to claim 2 is characterized in that: The steps of obtaining the screening signal segment data are specifically as follows: S111: calculating the water pipe vibration signal in segments according to the time series, obtaining the amplitude change rate for each signal segment, and comparing it with the set amplitude change rate threshold, filtering the signal segment whose amplitude change rate exceeds the threshold, and obtaining the signal segment after the amplitude change rate filtering; S112: performing power spectrum analysis on the signal segments after the amplitude change rate screening, extracting the power spectrum density curve of the signal segments, obtaining the power spectrum density peak value of each signal segment, and establishing a power spectrum density peak value set; S113: Screen each peak in the power spectrum density peak set and match it with the reference frequency band, and compare whether the power spectrum density peak falls within the range of the reference frequency band, using the formula: ； Calculate the weight of the matching signal in the reference frequency band, filter the signal segment that highly matches the reference frequency band, and generate filtered signal segment data; in, represents the weight of the matching signal segment, is the power spectral density value of each signal segment, is the peak frequency of the power spectrum of the signal segment, is the reference frequency center value, is the reference frequency bandwidth, is the total number of signal segments to be filtered.

4. The building water pipe fault diagnosis method based on data analysis according to claim 3 is characterized in that: The steps for obtaining the frequency component energy proportion data are specifically as follows: S211: analyzing the frequency range of the signal segments point by point from the filtered signal segment data, calculating the power spectrum proportions of multiple signal segments, extracting data within the frequency range through frequency analysis, and obtaining frequency range analysis data; S212: Analyzing the data within the frequency range, calculating the amplitude variation standard deviation of each signal segment, evaluating the stability and variability of the signal segment, screening signal segments with high stability, and obtaining amplitude variation standard deviation screening results; S213: Apply the change fluctuation value screening to the amplitude change standard deviation screening result, iteratively adjust the decomposition parameters, and screen the signal segments with low volatility and obvious frequency characteristics, using the formula: ； Calculate and generate frequency component energy ratio data; in, Represents the energy proportion of the frequency component, is the standard deviation of the amplitude variation of the signal segment, is the actual frequency of the signal segment, is the reference frequency, is the frequency sensitivity coefficient, is the number of signal segments.

5. The building water pipe fault diagnosis method based on data analysis according to claim 4 is characterized in that: The steps for obtaining the energy distribution characteristic data are specifically as follows: S311: using the frequency component energy proportion data, calculating the short-time spectrum amplitude of each signal in segments, verifying the spectrum distribution of the signal in the short-time window, and obtaining short-time spectrum amplitude data; S312: Calculate the frequency drift rate of the short-time spectrum amplitude data, compare the frequency drift rates of multiple signal segments with a preset reference value, select the signal segment whose drift rate exceeds the reference value, and obtain the frequency drift signal data exceeding the reference value; S313: Analyze the correlation between the super-reference value frequency drift signal data and the time window, identify the signal segment that meets the target time window change pattern, and use the formula: ； Calculate the energy distribution of each characteristic segment that meets the conditions and generate energy distribution characteristic data; in, Represents the energy distribution characteristics, is the short-time spectrum amplitude, is the frequency drift rate, is the reference drift rate, Adjust the parameter for drift rate sensitivity, is the time window length, is the number of feature segments.

6. The building water pipe fault diagnosis method based on data analysis according to claim 5 is characterized in that: The steps of obtaining the signal feature fusion vector are specifically as follows: S411: Based on the energy distribution characteristic data, perform weighted integral calculation of instantaneous energy and frequency drift rate for each characteristic segment, determine the frequency offset amplitude and instantaneous energy of multiple characteristic segments, and obtain frequency offset and energy change data; S412: Analyze the frequency offset and energy change data, calculate the drift standard deviation of each characteristic segment, evaluate the stability of the characteristic segment, iteratively screen the characteristic segments with high stability, and obtain the drift standard deviation screening result; S413: extracting a signal segment matching the associated characteristic frequency band from the drift standard deviation screening result, and converting the energy change of the signal segment into a signal strength feature, using the formula: ； Calculate the feature vectors of multiple signal segments and synthesize them into a signal feature fusion vector; in, represents the signal feature fusion vector, is the instantaneous energy of the multi-signal segment, is the signal strength, is the mean value of the drift standard deviation, is the number of signal segments.

7. The building water pipe fault diagnosis method based on data analysis according to claim 6 is characterized in that: The steps for obtaining the water pipe fault classification result are specifically as follows: S511: Based on the signal feature fusion vector, normalize the frequency drift rate and signal strength change characteristics, calculate the Euclidean distance between multiple vectors using the normalized eigenvalue of each vector, and generate a Euclidean distance matrix between normalized eigenvectors; S512: Combine the Euclidean distance matrix to calculate the weight ratio of multiple vectors, adjust the weight according to the normalized eigenvalue and distance distribution, select key eigenvectors, and use the formula: ； Calculate the weighted value of the feature vector and determine the classification boundary to generate weight proportion data; in, is the weighted value of the vector, For vector With vector The Euclidean distance of For vector The feature weights of For vector The weight distribution of is the total number of vectors; S513: Based on the weighted proportion data, combined with the Euclidean distance and the weighted value, the classification determination is completed by determining the correlation between the feature vector distribution and the classification boundary, and a water pipe fault classification result is generated.

8. Building water pipe fault diagnosis system based on data analysis, characterized in that: According to the method for diagnosing building water pipe faults based on data analysis according to any one of claims 1 to 7, the system comprises: The vibration signal segmentation module collects water pipe vibration signals in the building environment, segments the signals according to time series, calculates the amplitude change rate and power spectrum density peak, filters the signal segments that match the reference frequency band, and generates filtered signal segment data; The frequency component screening module analyzes the frequency range of the signal segment based on the screening signal segment data, calculates the power spectrum proportion and the standard deviation of the amplitude change, screens the significantly changed frequency components, adjusts the decomposition parameters, extracts the component energy proportion, and generates the frequency component energy proportion data; The feature segment extraction module calculates the signal short-time spectrum amplitude and frequency drift rate based on the frequency component energy proportion data, extracts the drift rate exceeding the reference value segment, analyzes the correlation between the amplitude and the time window, identifies the qualified frequency segment, and generates energy distribution feature data; The feature fusion processing module is based on the energy distribution feature data, weighted integral frequency drift rate and instantaneous energy change, calculates the frequency offset amplitude and drift standard deviation, statistically correlates the energy change of the feature frequency band, converts the signal strength feature, and generates a signal feature fusion vector; The fault classification and determination module normalizes the frequency drift rate and signal strength change characteristics based on the signal feature fusion vector, calculates the Euclidean distance and weight ratio between the features, determines the classification boundary to complete the classification and determination, and generates the water pipe fault classification result.

Images