Water turbine oil bubble monitoring method, system, equipment and medium

Abstract

The invention relates to the technical field of hydroelectric power generation equipment monitoring, in particular to a water turbine oil bubble monitoring method, system, equipment and medium, and the method comprises the steps: collecting a target scattered light signal generated by bubbles in oil flowing through a monitoring area through an optical imaging unit, and collecting a target parameter signal of the oil through a sensor group; converting the target scattered light signal into a digital image, and extracting primary feature information of the bubble based on the digital image; performing fusion analysis on the primary feature information and the target parameter signal uploaded to the cloud to generate a bubble state analysis result; and generating a control instruction based on the bubble state analysis result and feeding back the control instruction to the optical imaging unit so as to dynamically adjust acquisition working parameters of the optical imaging unit. The method has the beneficial effect that the sensing precision and the early warning timeliness of the oil bubble state of the water turbine are improved.

Description

Technical Field

[0001] This invention relates to the field of hydropower equipment monitoring technology, and in particular to a method, system, equipment and medium for monitoring oil bubbles in a water turbine. Background Technology

[0002] The turbine is the core equipment of a hydroelectric power station, and the condition of its hydraulic system directly affects the safe operation and power generation efficiency of the equipment. With the continuous increase in installed hydropower capacity, especially in high-head hydroelectric power stations, the problem of air bubbles in the turbine's hydraulic system is becoming increasingly prominent. Air bubbles in the hydraulic fluid can lead to oil film damage, reduced transmission efficiency, and in severe cases, equipment failure, affecting the safe and stable operation of the power generation equipment.

[0003] However, current turbine oil monitoring mainly suffers from the following problems: First, existing monitoring technologies can only monitor basic physicochemical indicators of the oil (such as moisture and viscosity), and cannot monitor and quantitatively analyze the state of air bubbles in the oil in real time; Second, under special operating conditions, the equipment operates under high pressure and high speed for a long time, and the oil system is prone to cavitation and cavitation phenomena, which traditional monitoring methods cannot cope with; Third, existing monitoring systems are limited to single data acquisition and lack the ability to qualitatively and quantitatively analyze oil bubbles, thus failing to provide effective decision support for equipment maintenance. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] In a first aspect, the present invention provides a method for monitoring air bubbles in hydraulic turbine oil, comprising acquiring target scattered light signals generated by air bubbles in oil flowing through a monitoring area through an optical imaging unit, and acquiring target parameter signals of the oil through a sensor group;

[0006] The target scattered light signal is converted into a digital image, and primary feature information of the bubble is extracted based on the digital image;

[0007] The primary feature information and the target parameter signal uploaded to the cloud are fused and analyzed to generate bubble state analysis results;

[0008] Based on the bubble state analysis results, control commands are generated and fed back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

[0009] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the step of extracting primary feature information of bubbles based on the digital image includes:

[0010] The digital image is preprocessed to enhance contrast and suppress noise;

[0011] An adaptive threshold edge detection algorithm is used to extract bubble contours from the preprocessed image;

[0012] Morphological analysis was performed on the extracted bubble contours to calculate the target value of a single bubble, and the number and size distribution of bubbles in the entire image were statistically analyzed.

[0013] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the step of fusing and analyzing the primary feature information and the target parameter signal uploaded to the cloud includes:

[0014] The bubble image features in the primary feature information are spatiotemporally aligned and correlated with the temperature, pressure, and flow rates in the parameter signals.

[0015] The target model is used to process the associated data, and the output results are used to classify the bubble types and predict the trend of bubble parameter changes in the future.

[0016] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the generated bubble state analysis results include:

[0017] The classification results and prediction results are judged based on a preset early warning rule base;

[0018] If the judgment result meets the triggering condition of any early warning rule, an early warning message of the corresponding level will be generated.

[0019] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the method includes, before acquiring the target scattered light signal generated by bubbles in the oil flowing through the monitoring area via the optical imaging unit, the following steps are taken:

[0020] The incident angle of the light source of the optical imaging unit is dynamically adjusted according to the current operating condition of the oil, wherein the current operating condition is the state of the oil determined based on historical data or the target parameter signal collected in real time.

[0021] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the acquisition of the target parameter signal of the oil is achieved by time-division multiplexing, including,

[0022] Within a preset sampling period, the channels are switched in a time-sharing manner by controlling the analog switch to sample the temperature signal, pressure signal and flow signal in sequence;

[0023] The target parameter signal is output by averaging multiple sampling points of each parameter signal.

[0024] As a preferred embodiment of the turbine oil bubble monitoring method of the present invention, the dynamic adjustment of the acquisition operating parameters of the optical imaging unit includes dynamically adjusting the light source intensity, exposure time, and imaging focal length of the optical imaging unit.

[0025] In a second aspect, the present invention provides a turbine oil bubble monitoring system, comprising: a data acquisition module, used to acquire target scattered light signals generated by bubbles in oil flowing through the monitoring area via an optical imaging unit, and to acquire target parameter signals of the oil via a sensor group;

[0026] An extraction module is used to convert the target scattered light signal into a digital image and extract the primary feature information of the bubble based on the digital image;

[0027] The analysis module is used to perform fusion analysis on the primary feature information and the target parameter signal uploaded to the cloud to generate bubble state analysis results;

[0028] The adjustment module is used to generate control commands based on the bubble state analysis results and feed them back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

[0029] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0030] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0031] Compared with existing technologies, the beneficial effects of this invention are as follows: by collecting data through optical imaging and multi-parameter sensing, bubble image feature extraction is completed locally to improve real-time performance. Then, multi-dimensional data such as temperature, pressure, and flow rate are integrated in the cloud for intelligent analysis and trend prediction, ultimately forming a closed-loop system from monitoring, analysis to feedback control. This significantly improves the perception accuracy and early warning timeliness of the turbine oil bubble state. Furthermore, by adaptively adjusting optical parameters, it effectively suppresses oil environment interference and ensures the stability and monitoring accuracy of the system under different operating conditions. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the method for monitoring air bubbles in hydraulic turbine oil. Detailed Implementation

[0034] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0035] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for monitoring air bubbles in hydraulic turbine oil, including:

[0036] S100: Acquires the target scattered light signal generated by bubbles in the oil flowing through the monitoring area through the optical imaging unit, and acquires the target parameter signal of the oil through the sensor group;

[0037] S200: Convert the target scattered light signal into a digital image, and extract the primary feature information of the bubble based on the digital image;

[0038] S300: The primary feature information and the target parameter signal uploaded to the cloud are fused and analyzed to generate bubble state analysis results;

[0039] S400: Based on the bubble state analysis results, generate control commands and feed them back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

[0040] It should be noted that the turbine oil system operates under high pressure, high speed, and variable temperature conditions for a long time. The generation, aggregation, and collapse of bubbles in the oil are instantaneous and random. Traditional manual sampling or offline testing cannot capture the dynamic behavior of bubbles, while fixed sensors are difficult to maintain signal stability under complex operating conditions, resulting in delayed bubble state recognition and a high misjudgment rate, which in turn affects the lubrication safety and operating efficiency of the unit.

[0041] Therefore, to address the aforementioned problems of difficulty in real-time bubble identification, poor adaptability to operating conditions, and lack of closed-loop control, the following steps (S100–S400) are executed cyclically: High signal-to-noise ratio bubble scattering signals are acquired at the source using adjustable-angle optical imaging and multi-parameter synchronous acquisition; the scattered light is converted into digital images in real time and primary features are extracted to achieve instantaneous quantification of bubble size and quantity; the primary features are correlated and fused with temperature, pressure, and flow rate in the cloud, and bubble state and trend predictions are output using a target model; S400 then converts the prediction results into control commands, adjusting the light source angle, light intensity, and sensor gain in reverse, forming a closed-loop "acquisition-analysis-optimization" chain. This ensures the accuracy, real-time performance, and robustness of bubble monitoring under different operating conditions, enabling online health assessment and early warning of the turbine oil bubble state.

[0042] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a method for monitoring air bubbles in hydraulic turbine oil is provided.

[0043] In this embodiment of the application, step S100 involves acquiring the target scattered light signal generated by bubbles in the oil flowing through the monitoring area using an optical imaging unit, and acquiring the target parameter signal of the oil using a sensor group, including the following steps A1-A2:

[0044] It should be noted that the optical imaging unit includes a light source and an optical sensor array. The light source generates a light beam of a specific wavelength and is mounted on the optical path of the light source via a flange connection to collect the scattered light signal from bubbles in the oil. The sensor array includes a temperature sensor, a pressure sensor, and a flow sensor, used to collect the temperature, pressure, and flow signals of the oil, respectively. Furthermore, the optical imaging unit and sensor array are mounted outside the thrust bearing, specifically below the thrust oil tank frame, using a flange designed at the bottom of the oil tank to sample oil and return it to the oil pipe. It is understood that the oil sample information from the bottom of the thrust bearing oil tank is more representative, and the oil pipe is designed with a filter to remove impurities that interfere with optical detection. Furthermore, this oil pipe section is made of 304# stainless steel, with the same inner diameter as the return oil pipe. Furthermore, the optical imaging unit also includes an optical path adjustment circuit and a light-emitting circuit. The optical path adjustment circuit adjusts the imaging focal length of the optical sensor array, while the light-emitting circuit, positioned around the optical sensor array, provides auxiliary illumination. Specifically, the optical path adjustment circuit adjusts the imaging focal length of the optical sensor array to ensure that the focal plane always coincides with the detection plane where the bubble is located, maintaining the clarity of the bubble image even when oil pressure fluctuations cause changes in refractive index. The light-emitting circuit evenly arranges 16 auxiliary light sources around the optical sensor array. These light sources operate synchronously with the main light source, providing supplementary illumination when the number of bubbles is high, ensuring sufficient scattered light signal is obtained even when the bubble volume fraction reaches 10%. Preferably, by adjusting the luminous intensity and timing of these auxiliary light sources, the spatial distribution of scattered light can be optimized, improving the accuracy of bubble detection.

[0045] It should be further noted that the light source uses a near-infrared light source with a wavelength range of 800nm ​​to 900nm, preferably 850nm. It can be understood that the light source uses light in this wavelength range because, firstly, the absorption coefficient in commonly used turbine oils such as turbine oil is low, and the transmittance can reach more than 80%, which has good penetrability; secondly, the light in this wavelength range is not sensitive to the scattering characteristics of common impurities in oil (such as small metal particles, water, etc.), which can effectively reduce background interference. In addition, the light in this wavelength range is not easily affected by changes in the color of the oil, and can still maintain a stable detection effect even when the oil undergoes slight oxidation and discoloration. Furthermore, in practical applications, the use of near-infrared light sources may face the problem of light intensity attenuation. Specifically, when the light beam propagates in oil, in addition to scattering caused by bubbles, the absorption and scattering of the oil itself will also cause the light intensity to decrease exponentially with the propagation distance. To solve this problem, this method sets up a light intensity feedback control loop in the driving circuit of the light source. The light intensity feedback control loop monitors the light intensity received by one of the photodiodes in the photodiode array in real time and dynamically adjusts the driving current of the light source so that the incident light intensity is always kept within the optimal detection range. When the detected light intensity is lower than a preset threshold (e.g., 50% of the initial light intensity), the light intensity feedback control loop automatically increases the driving current; when the detected light intensity exceeds the preset threshold (e.g., 150% of the initial light intensity), the light intensity feedback control loop automatically decreases the driving current. Furthermore, the optical sensor array employs a CMOS image sensor and is configured as a ring structure consisting of 16 photodiodes. These photodiodes are uniformly distributed in a ring around the oil flow pipe. Specifically, an annular groove is formed on the outer wall of the oil flow pipe, and the 16 photodiodes are installed at 22.5-degree intervals along the inner wall of this groove to collect light signals at different scattering angles. Each photodiode has an effective photosensitive area of ​​1mm × 1mm, a response time of less than 1μs, and a sensitivity of not less than 0.5A / W.

[0046] Understandably, the light source is positioned outside the oil flow pipe, with its emission direction perpendicular to the radial direction of the pipe. The light source is fixed 5mm from the outer wall of the pipe via an angle-adjustable mounting base, which allows adjustment of the incident angle from 0 to 45 degrees. The light source is located within the same plane of the ring-shaped photodiode array, directly opposite one of the photodiodes. (This mounting method has two advantages: firstly, because the emission direction is perpendicular to the radial direction of the oil flow pipe, and the incident beam is perpendicular to the oil flow direction, the velocity field generated by the oil flow does not change the beam's propagation path, thus avoiding light intensity fluctuations caused by the oil flow. Secondly, when the beam is incident in this manner, the scattering of light by the bubble mainly occurs in a plane perpendicular to the incident direction, i.e., the same as the detection plane of the photodiode array, allowing the scattered light to be fully received by the array.) Sixteen photodiodes are evenly distributed circumferentially along the oil flow pipe within an imaginary circular plane, perpendicular to the pipe's axial direction. The photosensitive surface of each photodiode faces the center of this circular plane. This circular plane serves as the detection plane for the photodiode array. The light source is located within this plane, and its angle is adjustable via an angle-adjustable mounting bracket. This ensures that all photodiodes receive the scattered light signal from the bubble within the same plane, facilitating accurate acquisition of the bubble's two-dimensional scattering characteristics. Furthermore, the incident angle of the light source is adjusted using a stepper motor-driven angle adjustment assembly. Specifically, this assembly includes a stepper motor, a reducer, and an angle sensor: the stepper motor's shaft is connected to the light source's mounting bracket via the reducer; the minimum step angle of the stepper motor is 1.8 degrees, and a 1:10 reduction ratio achieves a minimum adjustment accuracy of 0.18 degrees for the light source's incident angle. The angle sensor detects the actual incident angle of the light source in real time and feeds the detection signal back to the control circuit, forming a closed-loop control.

[0047] Preferably, the light source is positioned on one side of the oil flow pipe, and the emitted light beam is collimated and incident perpendicularly into the oil. When the beam encounters a bubble, it scatters in all directions. Since 16 photodiodes are evenly distributed along the circumference within the annular groove, and the photosensitive surface of each photodiode faces the central axis of the oil flow pipe, scattered light signals from different angles can be received simultaneously. This annular distribution structure ensures that there is a detection point every 22.5 degrees within the range of 0 to 360 degrees, achieving omnidirectional acquisition of bubble scattered light.

[0048] Furthermore, it should be noted that in this method, the scattered light signal, temperature signal, pressure signal, and flow signal are amplified and filtered by a signal conditioning circuit. The signal conditioning circuit includes an independent microcontroller, a programmable gain amplifier, and a digital filter (to suppress power frequency interference and high-frequency noise) to achieve adaptive gain control and real-time adjustment of filter parameters.

[0049] A1: Based on the current operating condition of the oil, dynamically adjust the incident angle of the light source of the optical imaging unit, wherein the current operating condition is the oil state determined based on historical data or the target parameter signal collected in real time.

[0050] It should be noted that the operating conditions mainly include the following situations: first, the content of air bubbles in the oil varies, specifically the volume fraction of air bubbles varies between 0.1% and 10%; second, the flow state of the oil varies, specifically the flow velocity varies between 0.1 m / s and 5 m / s; and third, the pressure state of the oil varies, specifically the pressure varies between 0 MPa and 20 MPa. For the aforementioned three operating conditions, the specific dynamic adjustment of the incident angle of the light source of the optical imaging unit is as follows:

[0051] (1) Regarding the operating conditions where the bubble volume fraction varies between 0.1% and 10%: Specifically, ① when the bubble volume fraction in the oil is less than 1%, the incident angle is adjusted to 0 degrees (i.e., perpendicular incident). The reason is that with perpendicular incident, the cross-sectional area of ​​the incident beam interacting with the bubbles is the largest, and the intensity of the scattered light is most uniformly distributed in the plane perpendicular to the incident direction. This results in the maximum intensity of the scattered light reaching the photodiode array, which is beneficial for improving the signal-to-noise ratio; ② when the bubble volume fraction in the oil is between 1% and 5%, the incident angle is adjusted to 5 degrees to 25 degrees. The specific value of the adjusted angle is directly proportional to the bubble volume fraction. The reason is that as the bubble volume fraction increases, the probability of multiple scattering during the propagation of the beam gradually increases. By gradually increasing the incident angle, the actual propagation path of the beam can be gradually reduced, thereby offsetting the influence of multiple scattering against the increase in the number of bubbles. For example, when the bubble volume fraction is 2%, the signal distortion caused by multiple scattering is about 15%. Adjusting the incident angle to 10 degrees at this time can reduce the propagation path by 15%, thereby compensating for this distortion. When the bubble volume fraction is 4%, the signal distortion caused by multiple scattering is approximately 30%. Adjusting the incident angle to 20 degrees at this point can reduce the propagation path by 30%. This progressive angle adjustment scheme can gradually reduce the impact of multiple scattering while maintaining a high intensity of scattered light. ③ When the bubble volume fraction in the oil exceeds 5%, the incident angle is adjusted to 30 to 45 degrees. The reason is that when there are many bubbles, the beam is prone to multiple scattering during propagation, i.e., the scattered light is scattered again by other bubbles. This multiple scattering leads to a chaotic distribution of scattered light intensity, affecting measurement accuracy. By obliquely incidenting the beam, the propagation path of the beam in the oil can be reduced, thereby reducing the probability of multiple scattering. Experiments have verified that when the incident angle is 45 degrees, the measurement error caused by multiple scattering can be reduced by more than 50%.

[0052] (2) Regarding the working condition where the oil flow rate varies between 0.1 m / s and 5 m / s, specifically: ① When the oil flow rate exceeds 2 m / s, the incident angle is adjusted to 15 to 25 degrees. The reason is that high-speed flowing oil will generate a density gradient in the flow direction, which will cause the light beam to be slightly deflected. By slightly tilting the incident beam, the direction of the deflected beam can be better aligned with the detection plane of the photodiode array. ② Further, when the flow rate is 3 m / s, setting the incident angle to 20 degrees can control the angle between the actual propagation direction of the deflected beam and the detection plane of the photodiode array within ±2 degrees. This ensures that the scattered light signal can be effectively received by the photodiode array, and also keeps the bubble detection accuracy of the system above 95% under high flow rate conditions, with the measurement error controlled within ±3%.

[0053] (3) Regarding the operating condition where the oil pressure varies between 0 MPa and 20 MPa, the increased pressure leads to an increase in oil density, which in turn changes the refractive index of light. To compensate for this pressure-induced change in refractive index, the incident angle needs to be adjusted according to the pressure value. For example, when the pressure is 10 MPa, the originally set incident angle needs to be increased by 2 degrees. When the pressure is 20 MPa, the originally set incident angle needs to be increased by 5 degrees. This pressure-compensated angle adjustment ensures that the optical path always maintains optimal coupling with the detection plane of the photodiode array.

[0054] In an optional implementation, the dynamic adjustment of the incident angle of the light source in the optical imaging unit in step A1 can also be achieved through MEMS micromirror electronic scanning. Specifically, a 2mm × 2mm MEMS mirror (±20° dual-axis adjustable, response time <1ms) is placed in front of the light source outlet. Based on real-time operating conditions (bubble concentration, flow rate, pressure), the target angle is calculated using a PID algorithm, driving the MEMS micromirror to deflect rapidly, achieving continuous variation of the incident beam from 0° to 45° while the light source itself remains fixed.

[0055] In another optional implementation, the dynamic adjustment of the incident angle of the light source of the optical imaging unit in step A1 can also be achieved by a wedge-shaped optical wedge turntable. This involves fabricating a set of miniature quartz optical wedges with different wedge angles (0°, 5°, 10°, 15°…45°) into a toothed disc and installing them within the light source channel. A stepper motor drives the toothed disc to rotate, thus cutting the corresponding wedge angle into the optical path and causing a fixed deflection of the light beam.

[0056] A2: The target parameter signals of the oil are acquired using time-division multiplexing, including...

[0057] Within a preset sampling period, the channels are switched in a time-sharing manner by controlling the analog switch to sample the temperature signal, pressure signal and flow signal in sequence;

[0058] Understandably, the temperature signal, pressure signal, and flow signal are sampled sequentially by using an analog multiplexer and control circuit. Specifically, the analog multiplexer uses an 8-channel analog switch chip, whose input is connected to the signal output of the temperature sensor, pressure sensor, and flow sensor, respectively, and whose output is connected to the signal conditioning circuit. The control circuit controls the channel selection of the analog switch chip through a 3-bit binary address code and switches different sensor channels using a predetermined sampling frequency (e.g., 50Hz).

[0059] The target parameter signal is output by averaging multiple sampling points of each parameter signal.

[0060] Specifically, within each 20ms sampling period, 2ms is first used for channel switching and signal stabilization, followed by 6ms for temperature signal sampling, during which 100 data points are collected. Next, 6ms is allocated for pressure signal sampling, also collecting 100 data points. Finally, 6ms is allocated for flow signal sampling, collecting 100 data points. A 16-bit analog-to-digital converter (ADC) is used during sampling at a sampling rate of 16.67kHz. The target parameter signal is output after averaging the 100 data points for each parameter (effectively suppressing sampling noise and improving measurement accuracy). It should be noted that a 2ms processing time is reserved between two sampling periods for data calculation and channel switching preparation; this timing arrangement ensures the sampling quality of each parameter signal.

[0061] In an optional implementation, the target scattered light signal generated by bubbles in the oil flowing through the monitoring area in step S100 can also be acquired by an integrated fiber bundle-scattering array probe. Specifically, a Φ10mm optical window is opened on the side wall of the thrust bearing return oil pipeline, and sapphire pressure-resistant glass is welded to the inside of the window. Seven 200μm multimode optical fibers are bundled into a ring "transmitting bundle", and a central optical fiber of the same specification serves as a "receiving bundle". The two bundles are inserted into the oil 3mm through the same window. An 850nm VCSEL laser is coupled to the far end of the transmitting bundle, and a miniature spectrometer or CMOS linear array is connected to the far end of the receiving bundle to directly obtain the 180° backscattered light intensity distribution. The sapphire window and the end face of the optical fiber are coated with an 850nm antireflection film, with a pressure resistance of 25MPa and an operating temperature of -40℃ to 200℃, which completely covers the operating range of the turbine.

[0062] In another optional implementation, the target scattered light signal generated by bubbles in the oil flowing through the monitoring area in step S100 can also be acquired by ultrasonic modulation-optical scattering synchronous detection. That is, while keeping the original 850nm ring photoelectric array unchanged, a pair of 40kHz ultrasonic transducers are symmetrically arranged on the outer wall of the same cross section of the pipeline. The ultrasonic pulse generates local density modulation in the oil, so that the bubble interface forms repeatable and controllable microscale vibration, thereby enhancing the scattering signal by 3-6dB. The photoelectric array synchronously samples at the ultrasonic pulse repetition frequency (1kHz), and the "ultrasonic-optical" related scattering component is extracted by lock-in amplification technology, which significantly suppresses background noise. By adjusting the ultrasonic power, the dynamic range of the scattering signal can be kept constant within the range of 0.1%-10% bubble volume fraction, thus solving the signal saturation problem caused by high-concentration multiple scattering.

[0063] In this embodiment of the application, step S200 converts the target scattered light signal into a digital image and extracts primary feature information of the bubble based on the digital image, including the following steps B1-B3:

[0064] It is understandable that the signal conditioned by the signal conditioning circuit is digitized by the image processing circuit. The image processing circuit includes an independent microcontroller and an embedded processor, which are used to perform image preprocessing, feature extraction and data compression, etc., to achieve image processing.

[0065] It should be noted that the image processing circuit includes an analog-to-digital converter, a bubble edge extractor, and a bubble feature calculator. The analog-to-digital converter is used to convert the scattered light signal into a digital image, the bubble edge extractor is used to extract the bubble outline from the digital image, and the bubble feature calculator is used to calculate the size and number distribution of the bubbles. In this method, a 16-bit analog-to-digital converter (ADC) is used for digitization.

[0066] B1: The digital image is preprocessed to enhance contrast and suppress noise;

[0067] It should be noted that the bubble edge extractor includes an image preprocessing unit, which employs a two-stage filtering structure. The first-stage filtering structure uses a Gaussian filter to remove image noise, with the kernel size adjustable between 3×3 and 7×7, and the standard deviation adjustable between 0.5 and 2. The second-stage filtering structure uses an adaptive histogram equalization algorithm to enhance image contrast, dividing the image into 16×16 sub-blocks for local enhancement, and using bilinear interpolation between adjacent sub-blocks to achieve a smooth transition.

[0068] In an optional implementation, the preprocessing of the digital image in step B1 to enhance contrast and suppress noise can also be performed by a combination of spatial-frequency domain denoising and adaptive histogram stretching. Specifically, an FFT is performed on the original 256×256 bubble image, and a Butterworth band-stop filter is used in the frequency domain to suppress 50kHz high-frequency noise (harmonics from the switching of common electro-hydraulic servo valves in water turbines). After inverse transformation back to the spatial domain, a modified two-dimensional wavelet soft thresholding (BayesShrink) is used to further remove low-frequency random noise. Finally, CLAHE (contrast-limited adaptive histogram equalization, tile = 8×8, clip = 2.0) is used to enhance the grayscale difference between the bubble edges and the background oil.

[0069] In another alternative implementation, the preprocessing of the digital image in step B1 to enhance contrast and suppress noise can also be performed based on a deep learning self-supervised denoising-enhancement integrated network. That is, a target number of unlabeled original bubble images are acquired, and a lightweight U-Net (with less than 1M parameters) is trained using a Blind-Spot self-supervised network to learn the noise distribution. During the inference phase, the network outputs the "denoised + contrast-enhanced" image in one step without manual parameter tuning. Under the constraint of edge preservation loss (GradientLoss), the bubble contour region obtains an additional 0.8dBPSNR improvement.

[0070] B2: An adaptive threshold edge detection algorithm is used to extract bubble contours from the preprocessed image;

[0071] It should be noted that the bubble edge extractor includes an edge detection unit, which enables accurate extraction of bubble contours from the image. Specifically, the input image is first subjected to directional gradient calculation. In both the horizontal and vertical directions, the gradient values ​​are obtained by calculating the gray-level difference between adjacent pixels, thereby obtaining a gradient image that reflects the local change characteristics of the image. Then, based on the optical scattering characteristics of bubbles, an adaptive threshold is set for edge detection: First, the average gradient of the entire image is calculated and used as a baseline. Then, two thresholds are set: a low threshold of 20% of the baseline value is used to detect possible edge points, and a high threshold of 60% of the baseline value is used to confirm real edge points, thus obtaining preliminary edge detection results. Further, after obtaining the preliminary edge detection results, for each pixel identified as an edge point, other edge points within its 8-neighborhood are checked. These edge points are organized into edge chains according to connectivity. During the formation of the edge chains, curvature constraints are introduced: the angle formed by three adjacent points on the edge chain is calculated. If the curvature change corresponding to this angle exceeds a preset value (e.g., 45 degrees / pixel), the edge point is considered to be a noise point caused by oil disturbance and requires further processing. For edge points that do not meet the curvature constraints, other possible edge points are first searched within the neighborhood of that point. These candidate points need to simultaneously satisfy both the gradient threshold and the curvature constraints. Then, the connection cost between these candidate points and the existing edge chains is calculated. The cost function considers three factors: gradient value, directional continuity, and curvature change. Finally, the candidate point with the lowest cost is selected as the new edge point, and the edge chain is updated. If no suitable candidate point is found in the neighborhood, the edge point is marked as a noise point and removed from the edge chain.

[0072] Preferably, this multi-level edge detection and optimization method can effectively extract the true edge of the bubble while suppressing interference from factors such as oil disturbance and light intensity changes. This results in good continuity and accuracy of the final edge detection results, which is beneficial for extracting more accurate bubble features in subsequent steps.

[0073] It should be further noted that the bubble edge extractor also includes a feature extraction unit. The feature extraction unit first performs a closure check on the edge detection results. For unclosed edges, the shortest path method is used to complete them. Then, the eight-neighbor labeling algorithm is used to label the closed regions to obtain each possible bubble region. Finally, based on the morphological characteristics of the bubble (such as a roundness of not less than 0.8 and an area between 100 and 10,000 pixels), the real bubble regions are selected.

[0074] B3: Perform morphological analysis on the extracted bubble contours, calculate the target value of a single bubble, and statistically analyze the number and size distribution of bubbles in the entire image. The target values ​​include basic morphological parameters such as area, perimeter, roundness, and eccentricity.

[0075] It should be noted that a three-dimensional shape model of the bubble is established based on the relevant parameters of the calculated target value. Specifically, the bubble is assumed to be a rotating ellipsoid, and its actual volume is estimated by projected area and contour curvature. Then, statistical analysis is performed on all bubbles in the same image to calculate the bubble quantity distribution, size distribution, and spatial distribution characteristics. For example, the bubbles are divided into 10 grades according to their equivalent diameter (from 10μm to 500μm), and the number of bubbles in each grade is counted. The spatial density distribution of the bubbles, i.e., the number of bubbles per unit volume, is calculated. The degree of bubble aggregation is analyzed, and the nearest neighbor distance method is used to determine whether the bubbles have clustered. It should be noted that these calculation results, after data format conversion, are transmitted to the cloud processing module through a high-speed serial interface (such as LVDS, with a transmission rate of 1Gbps) for subsequent analysis.

[0076] In an alternative implementation, the target value of a single bubble in step B3 can also be calculated by ellipse fitting based on curvature scale space (CSS). That is, CSS corner point detection is performed on the bubble outline, the curvature maxima are retained, and the remaining outline points are fitted into an ellipse using least squares to obtain the major axis a and the minor axis b. Finally, the target value of a single bubble is calculated by formula.

[0077] In another alternative implementation, the target value of a single bubble in step B3 can also be calculated based on the fast cumulative moment method of the contour chain code. That is, the contour is represented by an 8-direction Freeman chain code, and the zero-order, first-order, and second-order spatial moments are calculated in real time while traversing. The area, perimeter, circularity, and eccentricity are calculated at once by Hu moment formula, avoiding pixel-by-pixel traversal.

[0078] In this embodiment of the application, step S300 involves fusing and analyzing the primary feature information and the target parameter signal uploaded to the cloud to generate bubble state analysis results, including the following steps C1-C2:

[0079] C1: Spatiotemporally align and correlate the bubble image features in the primary feature information with the temperature, pressure, and flow rates in the parameter signals;

[0080] It should be noted that this method employs a hierarchical storage structure, including a time-series database and a relational database. The time-series database layer stores the raw sampling data, using a columnar storage scheme to organize parameters such as temperature, pressure, and flow rate by timestamps. To improve storage efficiency, the time-series database uses a differential encoding algorithm to compress consecutive sampling points, reducing the data storage volume to one-tenth of the original data volume. The relational database layer stores the processed feature data, including bubble statistical characteristics (such as quantity distribution and size distribution), equipment operating parameters, and analysis results. Furthermore, to achieve rapid retrieval, the relational database also includes a bubble feature table, an equipment parameter table, and an analysis results table, with corresponding index structures established in each of these three tables.

[0081] The target model is used to process the associated data, and the output results are used to classify the bubble types and predict the trend of bubble parameter changes in the future.

[0082] It should be noted that the target model uses a convolutional neural network to classify and predict bubble features. Specifically, it comprises two parts: a feature extraction network and a predictive analysis network. The feature extraction network contains five convolutional layers and three fully connected layers. The convolutional layers extract the morphological features of the bubbles, including edge features, texture features, and spatial distribution features. The fully connected layers are used for feature fusion and classification prediction. The input layer has two branches: an image branch receives 256×256 pixel bubble image features, where the grayscale value of each pixel represents the intensity of scattered light received at that point; and a parameter branch receives real-time acquired temperature values ​​(-40℃ to 200℃), pressure values ​​(0MPa to 20MPa), and flow rate values ​​(0.1m / s to 5m / s). The output layer includes two parts: a classification output and a prediction output. The classification output shows the bubble type (including four types: single spherical bubble, ellipsoidal bubble, bubble cluster, and bubble chain), with each type corresponding to a probability value between 0 and 1. The prediction output provides the trend of bubble volume fraction changes over the next hour, outputting 12 predicted values ​​at 5-minute intervals.

[0083] C2: Judge the classification results and the prediction results based on the preset early warning rule base;

[0084] If the judgment result meets the triggering condition of any early warning rule, an early warning message of the corresponding level will be generated.

[0085] It should be noted that this invention establishes a three-level early warning mechanism (bubble recognition): The first-level early warning mechanism is based on a single parameter threshold, exemplified by conditions such as a bubble volume fraction exceeding 5%, a bubble number growth rate exceeding 200% / hour, and a maximum bubble size exceeding 500μm; the second-level early warning is based on multi-parameter combination rules, exemplified by triggering an early warning when the bubble volume fraction exceeds 3% and the flow rate is below 0.5m / s for 5 minutes, and when the probability of identifying a "bubble cluster" type exceeds 80% and its total integral exceeds 3%; the third-level early warning is based on trend analysis, predicting the changing trend of bubble parameters through a Kalman filter algorithm, and issuing an early warning when the predicted value may exceed the warning value within 15 minutes. The early warning rules are stored in an XML configuration file, containing three parts: trigger conditions, level determination, and processing actions. The rule parameters can be dynamically modified through a Web interface during system operation. When an early warning is triggered, the system automatically generates an early warning report containing the early warning level, trigger reason, on-site data, and processing suggestions, pushes it to the remote monitoring unit via an encrypted WebSocket channel, and notifies relevant maintenance personnel via telephone, SMS, and email.

[0086] In this embodiment of the application, step S400 generates control commands based on the bubble state analysis results and feeds them back to the optical imaging unit to dynamically adjust the acquisition operating parameters of the optical imaging unit, including the following step D1:

[0087] It is understood that the remote monitoring unit includes a display unit and a control unit, wherein the display unit is used to display the bubble feature analysis results and the control unit is used to adjust the acquisition operating parameters of the optical imaging unit.

[0088] D1: Dynamically adjust the light source intensity, exposure time, and imaging focal length of the optical imaging unit.

[0089] It should be noted that when the peak value of the scattered light signal is detected to be less than 20% of the full scale, the control unit automatically increases the light source intensity by 20%; when the scattered light signal is detected to be saturated, the control unit automatically reduces the exposure time by 50%; when the edge sharpness of the bubble image is detected to be less than the threshold, the control unit adjusts the position of the lens group through the stepper motor and adjusts the focal length in 0.1mm steps until the edge sharpness meets the requirements.

[0090] In summary, this invention acquires data through the collaborative acquisition of optical imaging and multi-parameter sensing, extracts bubble image features locally to improve real-time performance, and then integrates multi-dimensional data such as temperature, pressure, and flow rate in the cloud for intelligent analysis and trend prediction. Ultimately, it forms a closed-loop system from monitoring and analysis to feedback control, which significantly improves the perception accuracy and early warning timeliness of the turbine oil bubble state. Furthermore, it effectively suppresses oil environment interference by adaptively adjusting optical parameters, and can also ensure the stability and monitoring accuracy of the system under different operating conditions.

[0091] Example 3 illustrates a schematic scheme for a method of monitoring oil bubbles in a water turbine. It should be noted that the technical solution of this water turbine oil bubble monitoring system belongs to the same concept as the technical solution of the water turbine oil bubble monitoring method described above. Details not described in detail in this embodiment can be found in the description of the technical solution of the water turbine oil bubble monitoring method described above.

[0092] This embodiment also provides a turbine oil bubble monitoring system, including:

[0093] The acquisition module is used to acquire the target scattered light signal generated by bubbles in the oil flowing through the monitoring area through the optical imaging unit, and to acquire the target parameter signal of the oil through the sensor group.

[0094] An extraction module is used to convert the target scattered light signal into a digital image and extract the primary feature information of the bubble based on the digital image;

[0095] The analysis module is used to perform fusion analysis on the primary feature information and the target parameter signal uploaded to the cloud to generate bubble state analysis results;

[0096] The adjustment module is used to generate control commands based on the bubble state analysis results and feed them back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

[0097] This embodiment also provides an electronic device suitable for monitoring oil bubbles in water turbines, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for monitoring oil bubbles in water turbines as proposed in the above embodiment.

[0098] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for monitoring oil bubbles in a water turbine as proposed in the above embodiments.

[0099] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for monitoring oil bubbles in water turbines proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0100] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring air bubbles in hydraulic turbine oil, characterized in that: include, The target scattered light signal generated by bubbles in the oil flowing through the monitoring area is acquired by an optical imaging unit, and the target parameter signal of the oil is acquired by a sensor group. The target scattered light signal is converted into a digital image, and primary feature information of the bubble is extracted based on the digital image; The primary feature information and the target parameter signal uploaded to the cloud are fused and analyzed to generate bubble state analysis results; Based on the bubble state analysis results, control commands are generated and fed back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

2. The method for monitoring air bubbles in hydraulic turbine oil as described in claim 1, characterized in that: The primary feature information of the bubble extracted based on the digital image includes, The digital image is preprocessed to enhance contrast and suppress noise; An adaptive threshold edge detection algorithm is used to extract bubble contours from the preprocessed image; Morphological analysis was performed on the extracted bubble contours to calculate the target value of a single bubble, and the number and size distribution of bubbles in the entire image were statistically analyzed.

3. The method for monitoring air bubbles in hydraulic turbine oil as described in claim 2, characterized in that: The process of fusing and analyzing the primary feature information and the target parameter signal uploaded to the cloud includes, The bubble image features in the primary feature information are spatiotemporally aligned and correlated with the temperature, pressure, and flow rates in the parameter signals. The target model is used to process the associated data, and the output results are used to classify the bubble types and predict the trend of bubble parameter changes in the future.

4. The method for monitoring air bubbles in hydraulic turbine oil as described in claim 3, characterized in that: The generated bubble state analysis results include, The classification results and prediction results are judged based on a preset early warning rule base; If the judgment result meets the triggering condition of any early warning rule, an early warning message of the corresponding level will be generated.

5. The method for monitoring air bubbles in hydraulic turbine oil as described in claim 4, characterized in that: Before acquiring the target scattered light signal generated by bubbles in the oil flowing through the monitoring area via the optical imaging unit, the process includes: The incident angle of the light source of the optical imaging unit is dynamically adjusted according to the current operating condition of the oil, wherein the current operating condition is the state of the oil determined based on historical data or the target parameter signal collected in real time.

6. The method for monitoring air bubbles in hydraulic turbine oil as described in claim 5, characterized in that: The acquisition of the target parameter signals of the oil is achieved using a time-division multiplexing method, including: Within a preset sampling period, the channels are switched in a time-sharing manner by controlling the analog switch to sample the temperature signal, pressure signal and flow signal in sequence; The target parameter signal is output by averaging multiple sampling points of each parameter signal.

7. A method for monitoring air bubbles in hydraulic turbine oil as described in any one of claims 1-6, characterized in that: The dynamic adjustment of the acquisition parameters of the optical imaging unit includes dynamically adjusting the light source intensity, exposure time, and imaging focal length of the optical imaging unit.

8. A turbine oil bubble monitoring system, using the method described in any one of claims 1-7, characterized in that, include: The acquisition module is used to acquire the target scattered light signal generated by bubbles in the oil flowing through the monitoring area through the optical imaging unit, and to acquire the target parameter signal of the oil through the sensor group. An extraction module is used to convert the target scattered light signal into a digital image and extract the primary feature information of the bubble based on the digital image; The analysis module is used to perform fusion analysis on the primary feature information and the target parameter signal uploaded to the cloud to generate bubble state analysis results; The adjustment module is used to generate control commands based on the bubble state analysis results and feed them back to the optical imaging unit to dynamically adjust the acquisition parameters of the optical imaging unit.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images