Non-destructive testing method and system for water content in insulating oil based on multi-frequency ultrasound

Abstract

The present application relates to the technical field of insulating oil detection, and discloses a method and system for nondestructive detection of the micro-water content of insulating oil based on multi-frequency ultrasound. Multi-frequency acoustic wave signals are emitted to an insulating oil sample, and data is collected to generate a frequency domain phase spectrum containing phase information. Target frequency band phase trajectories are obtained through sequence extraction and trend analysis, and data is generated after noise filtering and identification of features to preliminarily delineate the boundaries of water accumulation areas. The reliability of the boundaries is verified, and an initial contour is determined. The amplitude is calculated based on the contour center, and a fluctuation intensity distribution map is generated in combination with the gradient. After smoothing, iterative analysis of the data is performed to draw an extended boundary. Finally, the boundary closure detection is completed in combination with multiple features, a distribution mapping diagram is generated, and the detection result is obtained. The present application realizes nondestructive detection of the micro-water content of insulating oil based on layered analysis and verification of multi-frequency ultrasound phase characteristics, accurately identifies the position and range of water accumulation areas, improves the detection accuracy and reliability, and meets the needs of on-site detection of the micro-water content of insulating oil.

Description

Technical Field

[0001] This invention relates to the field of insulating oil testing technology, and in particular to a non-destructive testing method and system for the trace water content of insulating oil based on multi-frequency ultrasound. Background Technology

[0002] Insulating oil is the core insulating medium for power equipment, and its performance directly determines the safety and reliability of the equipment's operation. Under high-voltage conditions, trace amounts of moisture in the oil can significantly degrade its insulating properties, becoming a major cause of insulation failures in power equipment. Accurate detection of the moisture content in insulating oil is a core means of ensuring the safe and stable operation of power systems, and related research has significant practical value for extending equipment lifespan, preventing safety accidents, and reducing operation and maintenance risks.

[0003] Current insulation oil moisture detection technology has obvious limitations in engineering applications: traditional detection methods require complex pretreatment of oil samples, the detection process is cumbersome and time-consuming, and it is impossible to achieve rapid on-site detection; existing detection methods are easily affected by external environmental interference, have insufficient sensitivity to detect trace moisture, and the stability and reliability of the detection results are difficult to guarantee, which cannot meet the core needs of refined operation and maintenance of power systems.

[0004] Ultrasonic testing technology has the natural advantages of being non-destructive, fast, and easy to deploy on-site. However, its application in the field of detecting trace moisture in insulating oil is not yet mature. Existing technologies have a single detection dimension and insufficient analysis of the response characteristics of trace moisture in oil, making it difficult to achieve accurate quantitative detection of trace moisture content. This has become a key issue restricting the large-scale application of this technology in the field of power insulation testing. Summary of the Invention

[0005] This invention provides a non-destructive testing method and system for the trace water content of insulating oil based on multi-frequency ultrasound, so as to realize the non-destructive testing of the trace water content of insulating oil, accurately identify the location and range of moisture accumulation areas, improve the accuracy and reliability of testing, and adapt to the on-site testing needs of trace water in insulating oil.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a non-destructive testing method for the trace water content of insulating oil based on multi-frequency ultrasound, comprising: Multi-band acoustic signals are emitted to an insulating oil sample, and the original acoustic data after penetrating the insulating oil sample is collected by a receiving end to generate a frequency domain phase spectrum containing phase information. The phase shift value sequence is extracted from the frequency domain phase spectrum, and the trend of the phase shift value sequence is analyzed to separate the phase change trajectory of the target frequency band. The phase change trajectory is filtered for noise interference, the corresponding curve abrupt change points and fluctuation amplitudes are identified, and the preliminary boundaries of the water accumulation area are initially delineated and preliminary boundary data is generated. The reliability of the preliminary boundary data is verified based on adjacent frequency band data, the preliminary boundary is adjusted, and the initial outline of the water accumulation area is determined. Based on the center position of the initial contour, the phase fluctuation amplitude of the data around the corresponding center coordinates is calculated, and the main direction of the water accumulation area is identified by combining the fluctuation intensity gradient to generate a fluctuation intensity distribution map. The fluctuation intensity distribution map is smoothed, the continuity of fluctuation is confirmed by iterative analysis of adjacent frequency band data, and the boundary outline of the water accumulation area is drawn. By combining the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map, boundary closure detection is completed, generating a distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtaining the final detection result.

[0007] In one optional embodiment, the step of transmitting multi-band acoustic signals to the insulating oil sample and acquiring the original acoustic data after penetrating the insulating oil sample through a receiving end to generate a frequency domain phase spectrum containing phase information includes: A multi-band acoustic wave signal covering different frequency ranges is emitted to an insulating oil sample, so that the multi-band acoustic wave signal penetrates the insulating oil sample and is transmitted to the receiving end. The receiver captures and records the original acoustic wave data after penetrating the insulating oil sample, and removes abnormal interference signals from the original acoustic wave data. Frequency domain transformation technology is used to perform frequency domain transformation on the processed original acoustic wave data, extract the phase information in the original acoustic wave data, and generate a frequency domain phase spectrum containing phase information after integration processing.

[0008] In one optional implementation, the step of extracting a phase shift value sequence based on the frequency domain phase spectrum and performing trend analysis on the phase shift value sequence to separate the phase change trajectory of the target frequency band includes: Phase offset data corresponding to each frequency is extracted one by one from the frequency domain phase spectrum, and the extracted phase offset data is arranged in frequency order to form a continuous phase offset value sequence. The phase offset value sequence is analyzed for change trends. The change pattern of phase offset data in the phase offset value sequence is analyzed to identify fluctuation characteristics and phase offset differences between frequency bands. Then, the target frequency band selection criteria that include fluctuation characteristic patterns and the corresponding threshold of phase offset differences between frequency bands are determined. Based on the target frequency band selection criteria and combined with the frequency band distribution characteristics of the frequency domain phase spectrum, the phase offset data corresponding to the target frequency band is selected from the phase offset value sequence and integrated to form the phase change trajectory of the target frequency band.

[0009] In one optional implementation, the step of filtering noise interference from the phase change trajectory, identifying corresponding curve abrupt change points and fluctuation amplitudes, and initially delineating the preliminary boundaries of the water accumulation area and generating preliminary boundary data includes: The phase change trajectory is filtered for noise interference. A preset noise reduction algorithm is used to suppress high-frequency random noise and interference signals generated by equipment operation in the phase change trajectory, so as to obtain the effective feature signal in the phase change trajectory. The effective feature signals are subjected to curve feature recognition, the corresponding curve abrupt change points are captured and the phase parameters corresponding to each curve abrupt change point are recorded, and the fluctuation amplitude in different intervals is statistically analyzed and the distribution characteristics of the fluctuation amplitude are clarified. By combining the effective feature signals and the identified curve abrupt change points and fluctuation amplitudes, the boundaries of the water accumulation area are initially delineated to obtain preliminary boundaries; The preliminary boundaries are organized to form preliminary boundary data reflecting the boundaries of the water accumulation area.

[0010] In one optional implementation, the step of verifying the reliability of the preliminary boundary data based on adjacent frequency band data, adjusting the preliminary boundary, and determining the initial outline of the water accumulation region includes: Retrieve data from adjacent frequency bands, wherein the adjacent frequency band data includes the phase change trajectory, boundary feature parameters, and correlation patterns between frequency bands of the corresponding frequency bands; Perform consistency verification between the feature parameters corresponding to the preliminary boundary data and the adjacent frequency band data, calibrate the fit between the preliminary boundary data and the correlation pattern, and output the calibration result; Based on the calibration results, the preliminary boundary is calibrated to correct the boundary deviation for the parts where the preliminary boundary data does not match the adjacent frequency band data. After integrating and calibrating the preliminary boundaries, the boundary range of the water accumulation area is clarified, and an initial contour reflecting the characteristics of water accumulation is output.

[0011] In one optional implementation, the step of calculating the phase fluctuation amplitude of the surrounding data corresponding to the center coordinates of the initial contour, and identifying the main direction of the water accumulation area by combining the fluctuation intensity gradient, and generating a fluctuation intensity distribution map, includes: Determine the center position of the initial contour, extract the corresponding coordinate parameters, and determine the corresponding center coordinates; Based on the center coordinates, phase detection data within a preset range is retrieved, and the amplitude of the phase detection data is calculated to obtain the phase fluctuation amplitude corresponding to the data around the center coordinates. Gradient analysis is performed on the phase fluctuation amplitude to obtain the fluctuation intensity gradient in different directions, and the main direction of the water accumulation area is identified based on the fluctuation intensity gradient. The phase fluctuation amplitude and fluctuation intensity gradient are integrated and the data is visualized according to a preset mapping rule to generate a fluctuation intensity distribution map that includes fluctuation intensity decay points.

[0012] In one optional implementation, the step of smoothing the fluctuation intensity distribution map, confirming the continuity of fluctuations through iterative analysis of adjacent frequency band data, and drawing the boundary contour of the water accumulation area includes: The fluctuation intensity distribution map is subjected to data smoothing processing, and a preset smoothing algorithm is used to correct data fluctuations and eliminate high-frequency interference. Based on the smoothed wave intensity distribution map, acoustic wave detection data and wave intensity related parameters of adjacent frequency bands are retrieved and matched to establish the data correspondence between frequency bands. Based on the data correspondence between the frequency bands, the adjacent frequency band data are repeatedly verified by iterative analysis. The fluctuation intensity change pattern under different iterations is compared to confirm the fluctuation continuity between adjacent frequency bands and eliminate false fluctuation signals in non-clustered areas. By combining the continuity of the fluctuations and the intensity gradient changes of the fluctuation intensity distribution map, the boundary lines of the water accumulation area are delineated, and the boundary lines are integrated to form the boundary outline of the water accumulation area.

[0013] In one optional implementation, the boundary closure detection is completed by combining the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map, generating a distribution mapping map of the moisture accumulation region in the insulating oil sample, and obtaining the final detection result, including: Retrieve the boundary contour, center coordinates, and wave intensity attenuation points from the wave intensity distribution map, and establish data association relationships; Based on the data association, boundary closure detection is performed. Combined with the positional characteristics of the wave intensity attenuation point, the unclosed parts of the boundary contour are supplemented to achieve complete closure of the boundary contour. The boundary contour, center coordinates and wave intensity distribution after integration and closure are reconstructed by data visualization according to a preset mapping standard to generate a distribution mapping map of the moisture accumulation area in the insulating oil sample. The distribution mapping is validated, and the final detection results of the trace moisture content detection of the insulating oil sample are output in a preset format.

[0014] In one alternative implementation, after outputting the final detection result, the method further includes: Regional density assessment is performed on the distribution map. If the density distribution is uneven, the phase shift statistical characteristics of high-density sub-regions are extracted to generate a water accumulation risk level report.

[0015] Secondly, the present invention also provides a non-destructive testing system for the trace water content of insulating oil based on multi-frequency ultrasound, comprising: Ultrasonic spectrum acquisition module: emits multi-band acoustic signals to the insulating oil sample, and acquires the original acoustic data after penetrating the insulating oil sample through the receiving end, generating a frequency domain phase spectrum containing phase information. Track analysis and frequency band selection module: Extracts the phase shift value sequence based on the frequency domain phase spectrum, analyzes the changing trend of the phase shift value sequence, and separates the phase change trajectory of the target frequency band; Noise filtering and initial boundary determination module: filters noise interference from the phase change trajectory, identifies the corresponding curve abrupt change points and fluctuation amplitudes, and initially delineates the initial boundary of the water accumulation area and generates initial boundary data; Verification and delineation module: Verifies the reliability of the preliminary boundary data based on adjacent frequency band data, adjusts the preliminary boundary, and determines the initial outline of the water accumulation area; Amplitude calculation and mapping module: Based on the center position of the initial contour, calculate the phase fluctuation amplitude of the data around the corresponding center coordinates, and combine the fluctuation intensity gradient to identify the main direction of the water accumulation area, and generate a fluctuation intensity distribution map; Smoothing and Boundary Extension Module: Performs data smoothing on the fluctuation intensity distribution map, confirms the continuity of fluctuations through iterative analysis of adjacent frequency band data, and draws the boundary outline of the water accumulation area; The result closure module combines the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map to complete the boundary closure detection, generate a distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtain the final detection result.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) Based on multi-band acoustic wave signals, a frequency domain phase spectrum containing phase information is generated, the phase offset value sequence is extracted and the phase change trajectory of the target frequency band is separated, and the influence law of water on sound wave propagation is captured from the phase feature level, providing an accurate signal basis for water distribution identification. Compared with the traditional detection method that only relies on the amplitude of sound waves, the detection dimensions are richer and the signal identification is higher.

[0017] (2) Construct a hierarchical and progressive process for identifying the boundaries of water accumulation areas. After filtering and identifying features, the boundaries are initially divided and data is generated. The boundaries are verified and calibrated by adjacent frequency band data. The expanded boundaries are drawn by combining fluctuation intensity gradient analysis and iterative verification. The boundaries of the accumulation areas are gradually accurately located, effectively distinguishing between uniform water dissolution and local accumulation, and clearly identifying the location, range and other core features of the accumulation areas.

[0018] (3) During the detection process, cross-validation of adjacent frequency band data is introduced multiple times to complete boundary reliability verification and fluctuation continuity iterative analysis. Interference signals are eliminated by means of data smoothing and gradient analysis to avoid signal misjudgment caused by external interference, improve the anti-interference and reliability of the detection results, and ensure the accuracy of the judgment of the cluster area characteristics.

[0019] (4) The entire process is based on multi-frequency ultrasound to complete non-destructive testing. No pretreatment of insulating oil samples is required. The testing process is simple and has no secondary impact, which is suitable for the on-site testing needs of power equipment. The test results are presented in a distribution mapping diagram, which intuitively reflects the distribution of moisture in the oil. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a non-destructive testing system for the trace water content of insulating oil based on multi-frequency ultrasound, provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 This invention provides a non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound, comprising the following steps: S11, a multi-band acoustic wave signal is emitted to the insulating oil sample, and the original acoustic wave data after penetrating the insulating oil sample is collected by the receiving end to generate a frequency domain phase spectrum containing phase information. S12, extract the phase shift value sequence based on the frequency domain phase spectrum, and perform trend analysis on the phase shift value sequence to separate the phase change trajectory of the target frequency band; S13, filter noise interference from the phase change trajectory, identify the corresponding curve abrupt change points and fluctuation amplitudes, and preliminarily delineate the preliminary boundaries of the water accumulation area and generate preliminary boundary data. S14, verify the reliability of the preliminary boundary data based on adjacent frequency band data, adjust the preliminary boundary and determine the initial outline of the water accumulation area; S15, Based on the center position of the initial contour, calculate the phase fluctuation amplitude of the data surrounding the center coordinates, and combine the fluctuation intensity gradient to identify the main direction of the water accumulation area, and generate a fluctuation intensity distribution map; S16, perform data smoothing on the fluctuation intensity distribution map, confirm the continuity of fluctuation by iteratively analyzing adjacent frequency band data, and draw the boundary outline of the water accumulation area; S17, combine the boundary contour, center coordinates and the fluctuation intensity attenuation point in the fluctuation intensity distribution map to complete the boundary closure detection, generate the distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtain the final detection result.

[0023] In step S11, a multi-band acoustic signal is emitted to the insulating oil sample, and the original acoustic data after penetrating the insulating oil sample is collected by the receiving end to generate a frequency domain phase spectrum containing phase information.

[0024] In one embodiment, this step involves transmitting multi-band acoustic signals to an insulating oil sample, acquiring the raw acoustic data after penetration, and generating a frequency domain phase spectrum containing phase information. The object of the test is insulating oil used in power equipment, and the test is performed using a multi-frequency ultrasonic testing system composed of an intelligent sensor module, an ultrasonic transmitting module, an ultrasonic receiving module, and a data processing unit. Before testing, the insulating oil sample to be tested is placed in a customized oil-resistant sealed ultrasonic testing cell. Inside the testing cell, the intelligent sensor and the ultrasonic transmitting transducer and receiving transducer are fixedly arranged facing each other, with a transducer spacing of 8 cm to adapt to the acoustic wave propagation characteristics of the insulating oil. The ultrasonic transmitting module can transmit tunable acoustic signals from 200 kHz to 5 MHz, and the intelligent sensor and ultrasonic receiving module have a sampling rate of 20 MSps, enabling accurate capture and real-time recording of the acoustic signals.

[0025] During testing, the ultrasonic transmitting module emits multi-band acoustic signals covering 500kHz, 1MHz, 2MHz, 3MHz, and 5MHz into the insulating oil sample in the testing cell. Each band is emitted continuously with equal power and timing. The duration of a single band transmission is 50ms, the switching interval between bands is 10ms, and the emitted sound intensity is controlled at 0.5W / cm². 2This ensures that multi-band acoustic signals can completely penetrate the insulating oil sample and be stably transmitted to the intelligent sensor and ultrasonic transducer at the receiving end. The ultrasonic receiving module simultaneously captures and records the raw acoustic data after penetrating the insulating oil sample, converting the acquired analog acoustic signals into 16-bit digital signals to form a raw data sequence of time-voltage values.

[0026] The data processing unit employs the 3σ criterion to remove interference from the raw acoustic data. First, it calculates the mean and standard deviation of the raw data sequence. Abnormal signals deviating from the mean by more than three times the standard deviation are identified as interference signals generated by equipment electromagnetic interference or vibration of the detection pool and are removed. Simultaneously, linear interpolation is performed on the discontinuous data after removing abnormal signals to obtain interference-free processed raw acoustic data. Subsequently, the data processing unit uses Fast Fourier Transform (FFT) as the frequency domain transformation technique to perform frequency domain transformation on the processed raw acoustic data. The number of Fourier transform points is set to 2048, and the frequency resolution is 9.77Hz. The raw acoustic data in the time domain is converted into frequency-amplitude-phase three-dimensional data in the frequency domain. Phase information corresponding to each frequency point is extracted from the transformation result, with the phase angle accuracy controlled within ±0.1°. This process provides basic frequency domain data support for subsequent signal processing.

[0027] Detailed explanation of the preset rules for generating frequency domain phase spectra: The detection frequency (500kHz~5MHz) is used as the horizontal axis, and the phase angle (0~360°) is used as the vertical axis. The horizontal axis scale interval is 100kHz, and the vertical axis scale interval is 30°. The phase angle data corresponding to each frequency point are mapped one by one according to the coordinates, and adjacent data points are connected by a smooth curve. At the same time, the phase angle values ​​corresponding to each key frequency (500kHz, 1MHz, 2MHz, 3MHz, 5MHz) are marked. The spectrum background is white, the curve is black, and the key markings are red. The generated spectrum format is PNG, and the resolution is set to 1920×1080 to ensure that the spectrum is clear and intuitive and facilitates data extraction and analysis in subsequent signal processing.

[0028] The extracted phase information is correlated and integrated with its corresponding frequency to form a two-dimensional data set of frequency-phase angle. Finally, based on this data set, a frequency domain phase spectrum containing phase information is generated according to preset rules. The horizontal axis of the spectrum represents the detection frequency from 500kHz to 5MHz, and the vertical axis represents the phase angle from 0 to 360°. This allows for a direct reflection of the phase change characteristics of acoustic signals in each frequency band after passing through the insulating oil sample, thus completing the generation and storage of the frequency domain phase spectrum and providing core input data for subsequent signal processing.

[0029] In step S12, a phase shift value sequence is extracted based on the frequency domain phase spectrum, and the trend of the phase shift value sequence is analyzed to separate the phase change trajectory of the target frequency band.

[0030] In one implementation, this step, based on the generated frequency domain phase spectrum containing phase information, extracts the phase shift value sequence and analyzes the changing trend through systematic signal processing, ultimately separating the phase change trajectory of the target frequency band. The data processing unit directly retrieves the stored frequency domain phase spectrum data, extracts the phase shift data corresponding to each frequency within the full detection frequency band of 500kHz to 5MHz at preset frequency extraction intervals (10kHz), and collects and preprocesses this data using intelligent sensors to ensure the continuity and integrity of the extracted data. Subsequently, all extracted phase shift data are organized in ascending order of frequency to form a continuous phase shift value sequence covering the entire detection frequency band, completing the basic data organization step of signal processing.

[0031] Based on the formed phase shift value sequence, the data processing unit performs trend analysis on it using trend analysis algorithms in signal processing. Specifically, it analyzes the overall change pattern of the phase shift data in the sequence using the sliding window method. The sliding window size is set to 20 sets of data, and the step size is 5 sets of data. This accurately captures different characteristics of the phase shift in the sequence, such as linear changes and fluctuations. At the same time, it identifies significant fluctuation characteristics in the sequence through variance analysis (variance ≥ 0.5° is considered significant fluctuation). Then, it calculates the quantitative analysis of the phase shift difference between frequency bands by using the phase shift data difference between adjacent frequency points. This phase shift difference is calculated according to formula (1): In equation (1), Δφ is the phase shift difference between adjacent frequency points (unit: °), φ n+1 φ is the phase angle (in degrees) at the (n+1)th frequency point. n The phase angle (in degrees) at the nth frequency point is given. The selection criteria for target frequency bands are comprehensively determined by considering the influence of moisture in insulating oil on the phase propagation of sound waves. These criteria include the fluctuation characteristics of the phase shift value sequence and the threshold corresponding to the phase shift difference between frequency bands. The difference threshold is set to 5° to ensure that the selected target frequency bands can effectively reflect the distribution characteristics of moisture in the insulating oil.

[0032] Based on the established target frequency band selection criteria and considering the overall frequency band distribution characteristics of the frequency domain phase spectrum, the data processing unit accurately selects the phase shift data corresponding to the target frequency band that meets the selection requirements from the phase shift value sequence. In this embodiment, the selected target frequency band is 1.2MHz~3.8MHz. The phase shift data within this frequency band can effectively reflect the influence of moisture on the acoustic wave phase. The selected target frequency band phase shift data is integrated, and linear interpolation is used to complete and optimize the data. The interpolation interval is set to 5kHz. Through data completion and optimization in signal processing, a continuous and smooth target frequency band phase change trajectory is finally formed. This trajectory can intuitively reflect the specific characteristics of acoustic wave phase change with frequency within the target frequency band, providing accurate phase feature data for subsequent identification of moisture accumulation areas.

[0033] In step S13, noise interference is filtered out from the phase change trajectory, corresponding curve abrupt change points and fluctuation amplitudes are identified, and preliminary boundaries of the water accumulation area are initially delineated and preliminary boundary data is generated.

[0034] In one implementation, this step, based on the obtained target frequency band phase change trajectory, performs noise filtering and curve feature recognition through multi-dimensional signal processing to complete the preliminary boundary delineation of the water accumulation area and generate corresponding data. The data processing unit retrieves the complete data of the target frequency band phase change trajectory, which is collected and initially denoised by an intelligent sensor. A preset denoising algorithm (wavelet threshold denoising algorithm) is used to filter noise interference from the trajectory. A db4 wavelet basis is selected, the decomposition level is set to 5, and an adaptive soft thresholding method (threshold calculation method is...) is employed. (where σ is the noise standard deviation and N is the data length), accurately suppressing high-frequency random noise generated by signal transmission and low-frequency interference signals caused by equipment operation in the trajectory. While preserving the original trend of trajectory change, invalid fluctuation interference is eliminated to obtain an effective characteristic signal that can truly reflect the phase characteristics of insulating oil.

[0035] Based on the extracted effective feature signals, the data processing unit performs curve feature recognition using signal processing feature recognition algorithms. It calculates the feature signal curve point-by-point using the first-order difference method, setting an absolute value of the first-order difference ≥2° as the criterion for identifying abrupt change points. This accurately captures all abrupt change points in the curve and records complete phase parameters such as frequency, phase angle, and abrupt change amplitude for each point. Subsequently, the effective feature signals are divided into equidistant intervals based on frequency, with an interval interval of 100kHz. The fluctuation amplitude of the phase offset data within each interval is statistically analyzed (fluctuation amplitude = maximum phase angle value - minimum phase angle value within the interval). Simultaneously, the distribution characteristics of the fluctuation amplitude across the entire target frequency band are analyzed to clarify the range of locally high-amplitude fluctuations, providing a feature basis for subsequent boundary delineation.

[0036] Combining the overall trend of effective feature signals with the curve abrupt change points and fluctuation amplitude distribution characteristics identified through signal processing, the local high-amplitude fluctuation range is designated as the suspected core area of ​​water accumulation. The abrupt change points on both sides of this area are used as boundary nodes. These boundary nodes are connected sequentially according to frequency to form a closed curve profile, thus completing the preliminary boundary delineation of the water accumulation area. The data processing unit organizes the preliminary boundary data, extracting the frequency and phase angle coordinates of all boundary nodes. It also supplements the corresponding phase deviation, fluctuation amplitude, and other related feature information for each node. All data is integrated and stored in a preset structured format (JSON format). The structured format fields include: boundary node number, frequency (kHz), phase angle (°), phase deviation (°), fluctuation amplitude (°), and node type (core node / transition node). Through the signal processing data structuring and storage process, data storage is ensured to be standardized and easy to retrieve and verify later, forming preliminary boundary data that fully reflects the boundary characteristics of the water accumulation area.

[0037] In step S14, the reliability of the preliminary boundary data is verified based on adjacent frequency band data, the preliminary boundary is adjusted, and the initial outline of the water accumulation area is determined.

[0038] In one implementation, this step verifies the reliability of preliminary boundary data based on adjacent frequency band data, completes the calibration and adjustment of the preliminary boundary, and determines the initial outline of the water accumulation area. The data processing unit retrieves adjacent frequency band data on both sides of the target frequency band (1.2MHz~3.8MHz) from the previously stored full-band phase data, specifically the phase change trajectories of the 0.8MHz~1.2MHz and 3.8MHz~4.2MHz frequency bands. At the same time, it extracts the boundary feature parameters (including the location of abrupt change points and fluctuation amplitude range within the frequency band) and the correlation rules between frequency bands corresponding to these adjacent frequency bands. The correlation rules between frequency bands are calculated using the Pearson correlation coefficient, with a correlation ≥0.8 indicating a strong correlation. This clarifies the correlation rules of phase changes between different frequency bands, providing a reference benchmark for the verification stage of signal processing.

[0039] The core feature parameters (frequency, phase angle, and fluctuation amplitude of the boundary nodes) in the preliminary boundary data are checked point by point with the corresponding parameters in the adjacent frequency band data. The deviation quantization algorithm of signal processing is used to quantify and evaluate the fit between the preliminary boundary data and the correlation pattern between the frequency bands. The deviation rate is calculated according to formula (2): In equation (2), η is the deviation rate of the initial boundary parameters (unit: %), P is the characteristic parameter of the initial boundary node (including phase angle, fluctuation amplitude, etc.), and P0 is the reference characteristic parameter of the corresponding position of the adjacent frequency band. The deviation rate is set to ≤10% as a fit and >10% as a misfit. The calibration result containing the fit status and deviation rate value of each boundary node is output according to the verification result.

[0040] Based on the calibration results, for the mismatched parts in the initial boundary where the deviation rate exceeds the standard, the initial boundary position is linearly calibrated based on the correlation law of adjacent frequency bands. The correction formula is: Corrected phase angle = original phase angle - (original phase angle - adjacent frequency band reference phase angle) × 0.8 (correction coefficient). In this way, the phase deviation of all boundary nodes is corrected to ensure that the deviation rate of each node is ≤10% after calibration.

[0041] After completing the calibration of all mismatched areas, the data processing unit integrates all the calibrated preliminary boundary nodes and uses cubic spline interpolation to smoothly fit the boundary curves, eliminating abrupt fluctuations at the node connections and clarifying the complete boundary range of the water accumulation area. The final output is an initial contour that clearly reflects the core area and boundary transition area of ​​water accumulation. This contour is stored in the form of a coordinate point set, which is arranged in ascending order of frequency. Each coordinate point contains frequency (accurate to 1kHz) and phase angle (accurate to 0.01°) parameters to ensure that the contour data is accurate and traceable.

[0042] In step S15, based on the center position of the initial contour, the phase fluctuation amplitude of the surrounding data corresponding to the center coordinates is calculated, and the main direction of the water accumulation area is identified by combining the fluctuation intensity gradient to generate a fluctuation intensity distribution map.

[0043] In one implementation, this step, based on the determined initial contour of the water accumulation area, uses signal processing amplitude calculation and gradient analysis to determine the center coordinates and calculate the amplitude of surrounding phase fluctuations. Combined with fluctuation intensity gradient identification to extend the main direction, a fluctuation intensity distribution map including fluctuation intensity attenuation points is finally generated. The data processing unit retrieves the coordinate point set data of the initial contour, calculates the center position of the initial contour using the geometric center method, and obtains the arithmetic mean of the frequency and phase angle coordinates of all boundary nodes. This average value is used as the center coordinates of the initial contour, and the corresponding frequency and phase angle parameters are accurately extracted as the reference origin for subsequent surrounding data calculation and analysis.

[0044] Using the center coordinates as a reference, the data processing unit retrieves all phase detection data within a preset range. This data is collected and transmitted in real time by the intelligent sensor. The preset range is set as a two-dimensional area with the center coordinates as the center, a frequency direction of ±500kHz, and a phase angle direction of ±10°, ensuring coverage of the core and surrounding transition parts of the water accumulation area. The retrieved phase detection data is processed using the root mean square method to calculate the phase fluctuation amplitude, which is calculated according to formula (3): In equation (3), A is the amplitude of phase fluctuation (unit: °), x i For a single phase detection data point (unit: °). The average value of the phase detection data within the preset range (unit: °) is given by the formula, where N is the amount of phase detection data within the preset range. The phase fluctuation amplitude corresponding to each point around the center coordinate is calculated using this formula, which quantitatively reflects the degree of phase fluctuation at different locations.

[0045] Based on all calculated phase fluctuation amplitude data, the data processing unit uses the finite difference method to conduct gradient analysis, with a difference step size of 10kHz (frequency direction) and 1° (phase angle direction). The fluctuation intensity gradients are calculated in eight principal directions, including the positive and negative directions of the frequency axis and the positive and negative directions of the phase angle axis. The gradient values ​​visually reflect the rate of change of phase fluctuation amplitude in each direction; a larger gradient value indicates a more significant change in fluctuation amplitude in that direction. Based on the magnitude of the gradient values ​​in the eight principal directions, the top three directions with the highest gradient values ​​are identified as the main directions for the expansion of the water accumulation region. The gradient values ​​and amplitude change trends in each principal direction are also labeled.

[0046] Detailed explanation of the preset mapping rules for generating the fluctuation intensity distribution map: It is presented in polar coordinates, with the polar angle (0~360°) corresponding to the direction of water accumulation and expansion, and the polar radius (0~20°) corresponding to the amplitude of phase fluctuations. The polar angle scale interval is 15°, and the polar radius scale interval is 2°. The fluctuation intensity gradient is represented by color depth (gradient value 0~5° / kHz corresponds to a color gradient from light blue to dark red). Fluctuation intensity attenuation points are marked at locations where the gradient value drops sharply by ≥50% (the marking style is a red circle with a diameter of 5 pixels, and the coordinates of the attenuation point and the gradient change value are attached next to the marking). The distribution map background is black, the scale lines are white, and the main expansion direction is marked with a yellow arrow. The generated distribution map format is PNG, with a resolution of 1920×1080, and supports zooming to view details.

[0047] Finally, the phase fluctuation amplitude data around the central coordinates, the fluctuation intensity gradient information in each direction, and the identified main expansion direction are integrated. The data is then processed using a signal processing visualization mapping function according to a preset visualization mapping rule. This rule constructs a two-dimensional polar coordinate system with the central coordinates as the origin. The polar angle corresponds to the expansion direction, and the polar radius corresponds to the phase fluctuation amplitude. The gradient value is represented by color intensity, and fluctuation intensity decay points are marked at locations where the gradient value drops sharply by ≥50%. Based on this rule, data visualization processing is completed, generating a fluctuation intensity distribution map. The map clearly shows the core fluctuation characteristics, expansion trend, and intensity decay locations of the water accumulation area, providing an intuitive and accurate basis for subsequent drawing of the expansion boundary contour.

[0048] In step S16, the fluctuation intensity distribution map is smoothed, the continuity of fluctuation is confirmed by iterative analysis of adjacent frequency band data, and the boundary outline of the water accumulation area is drawn.

[0049] In one implementation, this step performs data smoothing based on the generated fluctuation intensity distribution map. Iterative analysis of adjacent frequency band data confirms the continuity of fluctuations, ultimately drawing the extended boundary outline of the water accumulation area. The data processing unit retrieves the full data of the fluctuation intensity distribution map and uses a preset smoothing algorithm (Gaussian smoothing algorithm) to smooth the data. The Gaussian kernel size is set to 5×5 and the standard deviation to 1.0. While preserving the core fluctuation characteristics and gradient change trends of the distribution map, it accurately corrects local abnormal fluctuations in the data, eliminates high-frequency interference caused by signal transmission and equipment operation, and ensures that the relative error of the fluctuation intensity and gradient value in each region is controlled within 5% after smoothing, resulting in an optimized fluctuation intensity distribution map.

[0050] Based on the smoothed wave intensity distribution map, the data processing unit retrieves acoustic wave detection data from adjacent frequency bands (0.8MHz to 4.2MHz) from previously stored detection data. This data is synchronously acquired by intelligent sensors. Simultaneously, the unit extracts wave intensity-related parameters for each frequency band, including wave amplitude, gradient value, and attenuation point location at each location. The adjacent frequency band data and the data in the wave intensity distribution map are matched one-to-one according to frequency-phase angle coordinates. A signal processing correlation fitting algorithm using the least squares method is employed to fit the correspondence between the data, establishing a complete data correspondence between frequency bands and clarifying the correlation and consistency of data from different frequency bands.

[0051] Based on the established data correspondence between frequency bands, the data processing unit performs iterative analysis to repeatedly verify the data of adjacent frequency bands. The number of iterations is set to 10. Each iteration optimizes the data matching accuracy based on the previous verification result. By comparing the fluctuation intensity change pattern under different iterations and combining the phase characteristic pattern of moisture accumulation in insulating oil, the continuity of fluctuations between adjacent frequency bands is determined. Signals without continuous fluctuation characteristics and inconsistent with the phase pattern of moisture accumulation are identified as false fluctuation signals in non-accumulation areas and are removed. Valid data that can truly reflect the expansion characteristics of moisture accumulation are retained.

[0052] Combining the fluctuation continuity characteristics obtained after verification and the intensity gradient change trend of the smoothed fluctuation intensity distribution map, continuous boundary lines are gradually delineated along the main direction of water accumulation region expansion, based on the critical position of gradient value change (gradient value ≤ 0.5° / kHz). Simultaneously, the boundary lines are completed and optimized by incorporating the location characteristics of fluctuation intensity attenuation points to ensure that the boundary lines accurately match the actual expansion range of the water accumulation region. Finally, all delineated boundary lines are integrated, and curve fitting is used to eliminate discontinuities and abrupt changes at the line junctions, forming a complete and smooth boundary outline of the water accumulation region expansion, clearly presenting the actual distribution and expansion pattern of the water accumulation region.

[0053] In step S17, the boundary closure detection is completed by combining the boundary contour, center coordinates and the fluctuation intensity attenuation point in the fluctuation intensity distribution map, generating a distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtaining the final detection result.

[0054] In one implementation, this step combines the boundary contour, center coordinates, and wave intensity attenuation points of the expanded water accumulation region. Through signal processing's closure detection and visualization reconstruction functions, boundary closure detection is completed, and the final detection result is generated. The data processing unit simultaneously retrieves the boundary contour coordinate point set of the expanded water accumulation region, the center coordinate parameters of the initial contour, and the location data of all wave intensity attenuation points marked on the wave intensity distribution map from the previously stored detection data. These three types of data are uniformly calibrated according to a two-dimensional coordinate system of frequency-phase angle, establishing a multi-dimensional data correlation between the boundary contour, center coordinates, and wave intensity attenuation points. The data calibration function of signal processing ensures accurate matching of all types of data under the same coordinate reference.

[0055] Based on the established multidimensional data association, boundary closure detection is carried out. The line connection status of the extended boundary contour is checked one by one. For the unclosed parts such as discontinuities and openings in the contour, the position characteristics of the wave intensity attenuation point are combined to accurately fill in the gaps. The attenuation point is used as the closure node. The connection curve is fitted according to the trend of the surrounding wave intensity gradient. The unclosed parts are smoothly connected by the boundary completion algorithm of signal processing. At the same time, the integrity and rationality of the closed contour are verified to ensure that the closed boundary contour can completely and accurately delineate the actual area of ​​water accumulation, thus achieving complete closure of the boundary contour.

[0056] Detailed explanation of the preset mapping standard for generating the distribution map: The horizontal axis is frequency (500kHz~5MHz), and the vertical axis is phase angle (0~360°). The horizontal axis scale interval is 50kHz, and the vertical axis scale interval is 15°. The closed boundary outline is drawn with a solid red line with a line width of 2 pixels. The center coordinates are marked with a yellow pentagram (side length 8 pixels), and the fluctuation intensity attenuation point is marked with a blue triangle (side length 4 pixels). The fluctuation intensity distribution is presented as a color heatmap, divided into 5 levels according to the fluctuation intensity value (0~10°), with corresponding colors ranging from light green to dark blue. Each level is labeled with the corresponding fluctuation intensity range. The map background is white, the scale lines are black, and the caption includes coordinate descriptions, color level descriptions, and label symbol descriptions. The generated map format is PNG with a resolution of 2048×1080, and it supports zooming in to view details.

[0057] After completing the boundary closure detection, the boundary contour, center coordinates, and fluctuation intensity distribution data of the entire region are integrated. The data is then visualized and reconstructed using a signal processing visualization reconstruction algorithm according to the above-mentioned preset mapping standard. This generates a distribution mapping map of the moisture accumulation area in the insulating oil sample, which intuitively presents the location, range, core area, and intensity distribution characteristics of moisture accumulation in the oil.

[0058] The data processing unit performs comprehensive data verification on the generated distribution map. It uses signal processing verification algorithms to check key indicators such as coordinate matching accuracy, boundary closure, and the accuracy of fluctuation intensity data labeling, ensuring the map is free of data deviations and labeling errors. After successful verification, the final test results for the trace moisture content detection of insulating oil samples are output in a preset format (PDF). This preset format includes four main modules: basic test information (test time, sample number, ambient temperature / humidity), core parameters (test frequency band, target frequency band, key signal processing algorithm parameters), characteristics of moisture accumulation areas (core coordinates, boundary range, peak fluctuation intensity, number and location of attenuation points), and a distribution map appendix. This ensures the test results are standardized and complete, providing a comprehensive and accurate testing basis for the assessment of trace moisture content in insulating oil and the operation and maintenance of power equipment.

[0059] It is worth noting that after step S17, there is also step S18: performing regional density assessment on the distribution map; if the density distribution is uneven, extracting the phase shift statistical features of high-density sub-regions to generate a water accumulation risk level report.

[0060] After the final results of the moisture content detection of the insulating oil sample are output, this step performs a regional density assessment on the distribution map of the moisture accumulation area using density analysis and feature extraction functions of signal processing, and generates a moisture accumulation risk level report. The data processing unit retrieves the generated distribution map of the moisture accumulation area, and based on the frequency-phase angle two-dimensional coordinate system in the map, it uses a grid division method to uniformly divide the entire map into grids. The grid cell is set to 20kHz×1°. The density statistical algorithm of signal processing calculates the mean fluctuation intensity and the density of phase offset data points in each grid cell. Combined with the boundary range of the moisture accumulation area, the overall regional density assessment of the map is completed, the density distribution status of each grid cell and the overall area is quantitatively determined, and the density uniformity characteristics are clarified.

[0061] If the assessment determines that the density distribution of water accumulation areas in the distribution map is uneven, the data processing unit further extracts the phase shift statistical features of all high-density sub-regions using a feature extraction algorithm for signal processing. The criteria for determining high-density sub-regions are that the density of data points within the grid cell exceeds the global average by 1.5 times or more. The extracted statistical features include the core coordinates, boundary range, mean and variance of phase shift, peak fluctuation intensity, number and distribution characteristics of phase abrupt change points, etc., of the high-density sub-regions. At the same time, the relationship between the density gradient changes of each high-density sub-region and the surrounding areas is analyzed to clarify the core dense areas of water accumulation and the diffusion trend.

[0062] Based on the extracted phase shift statistical characteristics of high-density sub-regions, and combined with industry standards for the operation of insulating oil in power equipment and the micro-water accumulation risk assessment specifications, a multi-dimensional risk assessment index system is established. The indicators cover the area proportion of high-density sub-regions, the degree of phase shift fluctuation, the peak value of the fluctuation intensity, and the correspondence with key insulation parts of the equipment. According to the weight and threshold of each assessment indicator, the moisture accumulation risk is divided into three levels: low, medium, and high. For each risk level, the corresponding impact on insulation performance, potential fault risks, and operation and maintenance recommendations are clearly defined.

[0063] Detailed instructions for the preset format of the moisture accumulation risk level report: The report uses Word format, A4 paper, portrait orientation, with 2.5cm top and bottom margins and 3cm left and right margins. The report includes eight chapters: cover, table of contents, testing overview, density assessment results, characteristics of high-density sub-regions, risk level determination, operation and maintenance suggestions, and appendices. The cover indicates the report name, sample number, testing unit, and report date. The table of contents includes the titles of each chapter and their corresponding page numbers. The appendices include a distribution mapping map, characteristic curves of high-density sub-regions, and a risk assessment indicator weight table, ensuring that the report is clearly structured, data is complete, and easy to view and archive.

[0064] Finally, by integrating the regional density assessment results, the phase shift statistical characteristics of high-density sub-regions, and the multi-dimensional risk assessment results, a standardized moisture accumulation risk level report is generated according to the above-mentioned preset format. This provides comprehensive and accurate technical support for the insulation status assessment of power equipment, early warning of micro-water faults, and subsequent operation and maintenance decisions, realizing integrated analysis from insulation oil micro-water detection to risk assessment.

[0065] In summary, by utilizing the modulation effect of moisture content and distribution on the phase of sound waves when multi-band ultrasound propagates in insulating oil, and combining this with a systematic processing logic of intelligent sensor acquisition, frequency domain analysis, noise suppression, boundary calibration, and data iterative optimization, this invention can achieve non-destructive testing of trace moisture content in insulating oil. It accurately captures the spatial distribution characteristics of moisture accumulation areas, effectively overcoming the limitations of traditional testing methods that can only quantify but not locate moisture. The testing process relies on intelligent sensors for accurate signal acquisition and preprocessing, eliminating the need for sample pretreatment, making operation convenient and without secondary damage. The test results are both accurate and reliable, and the generated distribution mapping map and risk assessment report can directly provide accurate basis for power equipment operation and maintenance, fully meeting the core needs of refined power system operation and maintenance for trace moisture detection in insulating oil.

[0066] refer to Figure 2 The second embodiment of the invention provides a non-destructive testing system for the trace water content of insulating oil based on multi-frequency ultrasound, comprising: Ultrasonic spectrum acquisition module: emits multi-band acoustic signals to the insulating oil sample, and acquires the original acoustic data after penetrating the insulating oil sample through the receiving end, generating a frequency domain phase spectrum containing phase information. Track analysis and frequency band selection module: Extracts the phase shift value sequence based on the frequency domain phase spectrum, analyzes the changing trend of the phase shift value sequence, and separates the phase change trajectory of the target frequency band; Noise filtering and initial boundary determination module: filters noise interference from the phase change trajectory, identifies the corresponding curve abrupt change points and fluctuation amplitudes, and initially delineates the initial boundary of the water accumulation area and generates initial boundary data; Verification and delineation module: Verifies the reliability of the preliminary boundary data based on adjacent frequency band data, adjusts the preliminary boundary, and determines the initial outline of the water accumulation area; Amplitude calculation and mapping module: Based on the center position of the initial contour, calculate the phase fluctuation amplitude of the data around the corresponding center coordinates, and combine the fluctuation intensity gradient to identify the main direction of the water accumulation area, and generate a fluctuation intensity distribution map; Smoothing and Boundary Extension Module: Performs data smoothing on the fluctuation intensity distribution map, confirms the continuity of fluctuations through iterative analysis of adjacent frequency band data, and draws the boundary outline of the water accumulation area; The result closure module combines the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map to complete the boundary closure detection, generate a distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtain the final detection result.

[0067] It should be noted that the non-destructive testing system for the moisture content of insulating oil based on multi-frequency ultrasound provided in this embodiment of the invention is used to execute all the process steps of the non-destructive testing method for the moisture content of insulating oil based on multi-frequency ultrasound in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0068] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound, characterized in that, include: Multi-band acoustic signals are emitted to an insulating oil sample, and the original acoustic data after penetrating the insulating oil sample is collected by a receiving end to generate a frequency domain phase spectrum containing phase information. The phase shift value sequence is extracted from the frequency domain phase spectrum, and the trend of the phase shift value sequence is analyzed to separate the phase change trajectory of the target frequency band. The phase change trajectory is filtered for noise interference, the corresponding curve abrupt change points and fluctuation amplitudes are identified, and the preliminary boundaries of the water accumulation area are initially delineated and preliminary boundary data is generated. The reliability of the preliminary boundary data is verified based on adjacent frequency band data, the preliminary boundary is adjusted, and the initial outline of the water accumulation area is determined. Based on the center position of the initial contour, the phase fluctuation amplitude of the data around the corresponding center coordinates is calculated, and the main direction of the water accumulation area is identified by combining the fluctuation intensity gradient to generate a fluctuation intensity distribution map. The fluctuation intensity distribution map is smoothed, the continuity of fluctuation is confirmed by iterative analysis of adjacent frequency band data, and the boundary outline of the water accumulation area is drawn. By combining the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map, the boundary closure detection is completed, and a distribution mapping map of the moisture accumulation area in the insulating oil sample is generated to obtain the final detection result.

2. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The process of transmitting multi-band acoustic signals to an insulating oil sample and acquiring the original acoustic data after penetrating the insulating oil sample through a receiving end to generate a frequency domain phase spectrum containing phase information includes: A multi-band acoustic wave signal covering different frequency ranges is emitted to an insulating oil sample, so that the multi-band acoustic wave signal penetrates the insulating oil sample and is transmitted to the receiving end. The receiver captures and records the original acoustic wave data after penetrating the insulating oil sample, and removes abnormal interference signals from the original acoustic wave data. Frequency domain transformation technology is used to perform frequency domain transformation on the processed original acoustic wave data, extract the phase information in the original acoustic wave data, and generate a frequency domain phase spectrum containing phase information after integration processing.

3. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The step of extracting the phase shift value sequence based on the frequency domain phase spectrum and performing trend analysis on the phase shift value sequence to separate the phase change trajectory of the target frequency band includes: Phase offset data corresponding to each frequency is extracted one by one from the frequency domain phase spectrum, and the extracted phase offset data is arranged in frequency order to form a continuous phase offset value sequence. The phase offset value sequence is analyzed for change trends. The change pattern of phase offset data in the phase offset value sequence is analyzed to identify fluctuation characteristics and phase offset differences between frequency bands. Then, the target frequency band selection criteria that include fluctuation characteristic patterns and the corresponding threshold of phase offset differences between frequency bands are determined. Based on the target frequency band selection criteria and combined with the frequency band distribution characteristics of the frequency domain phase spectrum, the phase offset data corresponding to the target frequency band is selected from the phase offset value sequence and integrated to form the phase change trajectory of the target frequency band.

4. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The process of filtering noise interference from the phase change trajectory, identifying corresponding curve abrupt change points and fluctuation amplitudes, and initially delineating the preliminary boundaries of the water accumulation area and generating preliminary boundary data includes: The phase change trajectory is filtered for noise interference. A preset noise reduction algorithm is used to suppress high-frequency random noise and interference signals generated by equipment operation in the phase change trajectory, so as to obtain the effective feature signal in the phase change trajectory. The effective feature signals are subjected to curve feature recognition, the corresponding curve abrupt change points are captured and the phase parameters corresponding to each curve abrupt change point are recorded, and the fluctuation amplitude in different intervals is statistically analyzed and the distribution characteristics of the fluctuation amplitude are clarified. By combining the effective feature signals and the identified curve abrupt change points and fluctuation amplitudes, the boundaries of the water accumulation area are initially delineated to obtain preliminary boundaries; The preliminary boundaries are organized to form preliminary boundary data reflecting the boundaries of the water accumulation area.

5. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The step of verifying the reliability of the preliminary boundary data based on adjacent frequency band data, adjusting the preliminary boundary, and determining the initial outline of the water accumulation region includes: Retrieve data from adjacent frequency bands, wherein the adjacent frequency band data includes the phase change trajectory, boundary feature parameters, and correlation patterns between frequency bands of the corresponding frequency bands; Perform consistency verification between the feature parameters corresponding to the preliminary boundary data and the adjacent frequency band data, calibrate the fit between the preliminary boundary data and the correlation pattern, and output the calibration result; Based on the calibration results, the preliminary boundary is calibrated to correct the boundary deviation for the parts where the preliminary boundary data does not match the adjacent frequency band data. After integrating and calibrating the preliminary boundaries, the boundary range of the water accumulation area is clarified, and an initial contour reflecting the characteristics of water accumulation is output.

6. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The step of calculating the phase fluctuation amplitude of the surrounding data corresponding to the center coordinates of the initial contour, and identifying the main direction of the water accumulation area by combining the fluctuation intensity gradient, and generating a fluctuation intensity distribution map includes: Determine the center position of the initial contour, extract the corresponding coordinate parameters, and determine the corresponding center coordinates; Based on the center coordinates, phase detection data within a preset range is retrieved, and the amplitude of the phase detection data is calculated to obtain the phase fluctuation amplitude corresponding to the data around the center coordinates. Gradient analysis is performed on the phase fluctuation amplitude to obtain the fluctuation intensity gradient in different directions, and the main direction of the water accumulation area is identified based on the fluctuation intensity gradient. The phase fluctuation amplitude and fluctuation intensity gradient are integrated and the data is visualized according to a preset mapping rule to generate a fluctuation intensity distribution map that includes fluctuation intensity decay points.

7. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The process of smoothing the fluctuation intensity distribution map, confirming the continuity of fluctuations through iterative analysis of adjacent frequency band data, and drawing the boundary contour of the water accumulation area includes: The fluctuation intensity distribution map is smoothed by using a preset smoothing algorithm to correct data fluctuations and eliminate high-frequency interference. Based on the smoothed wave intensity distribution map, acoustic wave detection data and wave intensity related parameters of adjacent frequency bands are retrieved and matched to establish the data correspondence between frequency bands. Based on the data correspondence between the frequency bands, the adjacent frequency band data are repeatedly verified by iterative analysis. The fluctuation intensity change pattern under different iterations is compared to confirm the fluctuation continuity between adjacent frequency bands and eliminate false fluctuation signals in non-clustered areas. By combining the continuity of the fluctuations and the intensity gradient changes of the fluctuation intensity distribution map, the boundary lines of the water accumulation area are delineated, and the boundary lines are integrated to form the boundary outline of the water accumulation area.

8. The non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, The boundary closure detection is completed by combining the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map, generating a distribution mapping map of the moisture accumulation region in the insulating oil sample, and obtaining the final detection result, including: Retrieve the boundary contour, center coordinates, and wave intensity attenuation points from the wave intensity distribution map, and establish data association relationships; Based on the data association, boundary closure detection is performed. Combined with the positional characteristics of the wave intensity attenuation point, the unclosed parts of the boundary contour are supplemented to achieve complete closure of the boundary contour. The boundary contour, center coordinates and wave intensity distribution after integration and closure are reconstructed by data visualization according to a preset mapping standard to generate a distribution mapping map of the moisture accumulation area in the insulating oil sample. The distribution mapping is validated, and the final detection results of the trace moisture content detection of the insulating oil sample are output in a preset format.

9. A non-destructive testing method for trace water content in insulating oil based on multi-frequency ultrasound according to claim 1, characterized in that, After outputting the final detection results, the following is also included: A regional density assessment is performed on the distribution map. If the density assessment result shows that the density distribution is uneven, the phase shift statistical characteristics of the high-density sub-regions are extracted to generate a water accumulation risk level report.

10. A non-destructive testing system for trace water content in insulating oil based on multi-frequency ultrasound, characterized in that, include: Ultrasonic spectrum acquisition module: emits multi-band acoustic signals to the insulating oil sample, and acquires the original acoustic data after penetrating the insulating oil sample through the receiving end, generating a frequency domain phase spectrum containing phase information. Track analysis and frequency band selection module: Extracts the phase shift value sequence based on the frequency domain phase spectrum, analyzes the changing trend of the phase shift value sequence, and separates the phase change trajectory of the target frequency band; Noise filtering and initial boundary determination module: filters noise interference from the phase change trajectory, identifies corresponding curve abrupt change points and fluctuation amplitudes, and initially delineates the initial boundary of the water accumulation area and generates initial boundary data; Verification and delineation module: Verifies the reliability of the preliminary boundary data based on adjacent frequency band data, adjusts the preliminary boundary, and determines the initial outline of the water accumulation area; Amplitude calculation and mapping module: Based on the center position of the initial contour, calculate the phase fluctuation amplitude of the data around the corresponding center coordinates, and combine the fluctuation intensity gradient to identify the main direction of the water accumulation area, and generate a fluctuation intensity distribution map; Smoothing and Boundary Extension Module: Performs data smoothing on the fluctuation intensity distribution map, confirms the continuity of fluctuations through iterative analysis of adjacent frequency band data, and draws the boundary outline of the water accumulation area; The result closure module combines the boundary contour, center coordinates, and fluctuation intensity attenuation points in the fluctuation intensity distribution map to complete the boundary closure detection, generate a distribution mapping map of the moisture accumulation area in the insulating oil sample, and obtain the final detection result.

Images