Microwave sensor, mineralization degree measuring system, method for determining measurement physical quantity and frequency point of microwave sensor and mineralization degree measuring method

Abstract

The invention discloses a microwave sensor, a mineralization degree measuring system, a method for determining a measurement physical quantity and a frequency point of the microwave sensor and a mineralization degree measuring method. The sensor comprises a transmitting unit, a detecting unit and two antenna units, wherein the antenna unit is an enameled wire; the mineralization degree measuring system is used in a dynamic experiment and an anti-interference experiment, and the microwave sensor is arranged on the side wall of a pipeline of the mineralization degree measuring system; the determination method is suitable for the mineralization degree measurement system, and amplitude characteristics and frequency points are obtained; according to the measurement method, the salinity is obtained by adopting the amplitude characteristics and the frequency points; the antenna unit of the system is made of an enameled wire material, and is insulated, corrosion-resistant and high in plasticity; the emission detection unit accurately regulates and controls signals, and data quality is improved; and the mounting base is integrated, sealed and leakproof. On the basis of experimental data, a radial basis kernel function vector machine is used for fitting nonlinear correlation, two-dimensional screening amplitude is used as a core index, the optimal frequency point is determined through the three indexes, and accurate detection of complex working conditions is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sensor design, in particular to a microwave sensor, a mineralization measurement system, a method for determining a physical quantity and a frequency point of a microwave sensor, and a mineralization measurement method. BACKGROUND

[0002] Mineralization is a core indicator for measuring the total amount of soluble mineral ions in water bodies, and directly reflects the salt content level of water bodies. Its accurate measurement has important application value in many fields such as aquaculture, water conservancy engineering, environmental monitoring, and petrochemical industry. In the aquaculture scene, the stability of water mineralization directly affects the survival, growth and reproduction of cultured organisms. Excessive or insufficient mineralization will cause stress, disease and even death of cultured organisms. In water conservancy engineering, dynamic monitoring of mineralization can effectively predict the trend of water salinization, providing data support for rational use of water resources and ecological protection. In industrial production, such as chemical industry and power industry, excessive mineralization in circulating water and cooling water will cause pipeline corrosion, scaling and other problems, affecting the normal operation of production equipment.

[0003] Currently, mineralization measurement methods mainly include laboratory detection method and on-site online detection method. Laboratory detection methods mainly include gravimetric method, titration method, and conductivity method. The gravimetric method, as a traditional and classic method, measures the mass of residual water after evaporation to calculate the mineralization, which has high measurement accuracy. However, it is complicated to operate, time-consuming, requires professional laboratory personnel and environment, and cannot realize real-time detection on site. It is also difficult to adapt to the monitoring needs of dynamic water mineralization fluctuations. The titration method consumes chemical reagents, has a complex operation process, is easily disturbed by other ions in the water body, and has low detection efficiency. The conductivity method measures based on the correlation between water conductivity and mineralization, which is simple to operate and has fast response speed. However, it is greatly affected by temperature, ion types and concentration ratio in the water body, and has large measurement error and weak anti-interference ability in complex water environment.

[0004] The antenna unit of the existing microwave sensor mostly uses ordinary metal wires, which are not specially designed for water environment detection scenes, and the insulation performance and corrosion resistance are insufficient. After long-term immersion in water, the metal wires are easily corroded and short-circuited by electrolytes and impurities in the water, resulting in unstable signal transmission, even damage to the sensor components, and shortening of the service life of the sensor. At the same time, ordinary metal wires have poor plasticity and cannot flexibly adjust the shape, angle and length of the antenna extending into the water according to the spatial limitations and water depth of the detection scene, and have poor adaptability. At the same time, the existing measurement system can only simulate a single static salinity scene, and cannot accurately simulate the dynamic fluctuation process of the salinity in the actual water body, nor can it simulate the common bubble interference in actual application, resulting in a large difference between the experimental conditions and the actual application scene, the experimental data lack representativeness, and the measurement method and parameters based on such data will have a large measurement deviation when applied to the actual scene. SUMMARY

[0005] To solve the above technical problems, one of the technical solutions adopted by the present application is a microwave sensor, which comprises: a detection unit 2 for converting received microwave signals into electrical signals; the sensor further comprises: two antenna units 3 for transmitting microwave signals and receiving reflected signals; each of the antenna units 3 is connected to a corresponding transmitting unit 1 and detection unit 2 at one end, and extends vertically into the water quality to be detected at the other end; wherein the antenna unit 3 is an enameled wire, and the antenna unit 3 uses an enameled wire to avoid chemical reaction and corrosion caused by contact between the end face and salt water; the frequency point of the sensor is measured through dynamic experiment and anti-interference experiment.

[0006] Another technical solution adopted by the present application is a salinity measurement system, which is used in dynamic experiment and anti-interference experiment, and comprises: a pipeline 4; at least two peristaltic pumps 5 which are respectively connected to the pipeline 4; one of the peristaltic pumps 5 draws salt water solutions with different salinities into the pipeline; the other peristaltic pump 5 draws air into the pipeline; the above-mentioned microwave sensor is installed on the side wall of the pipeline 4; the microwave sensor transmits the collected data to a pre-set detection system.

[0007] Another technical solution adopted by the present application is a method for determining the physical quantity and frequency point measured by a microwave sensor, which is applicable to the mineralization measurement system described above. The method comprises: the mineralization measurement system respectively introduces salt water solutions with different mineralization degrees and preset mineralization degree changes or air into the pipeline; when performing a dynamic experiment, a peristaltic pump in the mineralization measurement system continuously pumps the salt water solutions with different mineralization degrees into the pipeline, and the microwave sensor records the amplitude data and phase data under different mineralization degrees and preset mineralization degree changes; when performing an anti-interference experiment, one peristaltic pump in the mineralization measurement system continuously pumps the salt water solutions with different mineralization degrees into the pipeline, and another peristaltic pump continuously pumps air into the pipeline, the rotational speed of the two peristaltic pumps is controlled to control different gas-liquid ratios, and the microwave sensor records the amplitude data and phase data under different mineralization degrees and preset mineralization degree changes; the vector machine regression fitting strategy is used to respectively perform curve fitting on the amplitude data and phase data under different mineralization degrees and preset mineralization degree changes measured in the dynamic experiment and the anti-interference experiment, to obtain amplitude characteristic sweep frequency curve graphs and phase characteristic sweep frequency curve graphs of different mineralization degrees in static state and under different gas-liquid ratios, and amplitude characteristic sweep frequency curve comparison graphs and phase characteristic sweep frequency curve comparison graphs under preset mineralization degree changes; the amplitude characteristic sweep frequency curve graphs and phase characteristic sweep frequency curve graphs of different mineralization degrees in static state and under different gas-liquid ratios, and the amplitude characteristic sweep frequency curve comparison graphs and phase characteristic sweep frequency curve comparison graphs under preset mineralization degree changes are observed and analyzed, the monotonicity between the amplitude characteristic, the phase characteristic and the mineralization degree is observed, and the sensitivity of the amplitude characteristic compared with the phase characteristic under different mineralization degrees is compared, to determine that the measurement physical quantity of the sensor is the amplitude characteristic; the amplitude characteristic correlation coefficients, data ranges and standard deviations under different mineralization degrees and different gas-liquid ratios are calculated and compared, and the frequency corresponding to the amplitude correlation coefficient closest to 1, the maximum data range and the minimum standard deviation is the frequency point.

[0008] Further, the vector machine regression fitting strategy comprises: preprocessing the original data collected in the dynamic experiment and the anti-interference experiment, eliminating abnormal values in the discrete amplitude data and phase data; normalizing the effective amplitude data and phase data after eliminating abnormal values, mapping the amplitude data, phase data, mineralization parameter, gas-liquid ratio parameter and frequency parameter to the [0, 1] interval; defining the radial basis kernel function as the kernel function of the vector machine regression model, which is used to fit the effective mineralization, gas-liquid ratio and frequency with the amplitude and phase data into a nonlinear relationship, and introduces a penalty coefficient and an insensitive loss function parameter; combining the grid search method and the five-fold cross-validation to obtain the optimal penalty coefficient, kernel function parameter and insensitive loss function parameter, and constructing an optimal vector machine regression fitting model; taking the measured mineralization, preset mineralization change amount and frequency in the dynamic experiment as input variables, and taking the corresponding amplitude data and phase data as output variables, substituting them into the optimal vector machine regression fitting model, and through the nonlinear mapping capability of the model, the discrete amplitude and phase data are fitted into continuous curves to obtain the amplitude characteristic sweep frequency curve and the phase characteristic sweep frequency curve under different mineralization, and the amplitude characteristic sweep frequency curve comparison data and the phase characteristic sweep frequency curve comparison data under the preset mineralization change; taking the measured mineralization, preset mineralization change amount, gas-liquid ratio and frequency in the anti-interference experiment as input variables, and taking the corresponding amplitude data and phase data as output variables, substituting them into the same optimal vector machine regression fitting model, and continuously fitting the discrete data to obtain the amplitude characteristic sweep frequency curve and the phase characteristic sweep frequency curve under different mineralization and different gas-liquid ratio, and the amplitude characteristic sweep frequency curve comparison data and the phase characteristic sweep frequency curve comparison data under the preset mineralization change.

[0009] Further, the effective amplitude data and phase data after preprocessing are divided into training set and test set according to a preset ratio, the training set is used to construct the vector machine regression fitting model, and the test set is used to verify the fitting precision and generalization ability of the model.

[0010] Further, the divided test set data is substituted into the fitting model to calculate the fitting error and the determination coefficient of the test set data, and the fitting result is verified; when the determination coefficient is greater than a preset threshold and the fitting error is less than a preset range, the fitting model precision meets the experimental requirements, and the fitting curve is correct; if the determination coefficient is less than the preset threshold and the fitting error is greater than the preset range, the values of the penalty coefficient, the kernel function parameter and the insensitive loss function parameter are adjusted, and the fitting and verification are performed again until the determination coefficient is greater than the preset threshold and the fitting error is less than the preset range.

[0011] Further, if the fitting curve does not capture the data variation rule, the value of the penalty coefficient is increased; if the training set can be well fitted but the test set is large, the value of the penalty coefficient is reduced; if the fitting curve cannot capture the complex correlation between the mineralization, gas-liquid ratio and amplitude or phase, the value of the kernel function parameter is gradually increased; if the fitting curve fluctuates greatly and the fitting degree is poor, the value of the kernel function parameter is reduced; if the actual fitting deviation value frequently exceeds the standard, the value of the insensitive loss function parameter is increased; if some high-frequency data fitting deviation is large, the value of the insensitive loss function parameter is reduced. Further, only one of the values of the penalty coefficient parameter, the kernel function parameter and the insensitive loss function parameter is adjusted each time, and the values of the other two parameters are fixed, the optimal value of the parameter is screened by five-fold cross-validation method, and the value adjustment of the three parameters is completed in turn, then the values of the penalty coefficient parameter, the kernel function parameter and the insensitive loss function parameter are overall fine-tuned to obtain the optimal values of the penalty coefficient parameter, the kernel function parameter and the insensitive loss function parameter.

[0012] Another technical solution adopted by the present application is a mineralization measurement method, which comprises: determining the amplitude characteristic and frequency point of the microwave sensor by the method for determining the physical quantity and frequency point of the microwave sensor; passing the antenna unit 3 of the microwave sensor into the water quality to be detected; starting the microwave sensor to continuously work at the determined frequency point, generating a microwave signal in a preset range through the transmitting unit 1, emitting the microwave signal to the water quality to be detected through the antenna unit 3, synchronously receiving the microwave reflection signal through the antenna unit 3 and transmitting the microwave reflection signal to the detecting unit 2, converting the reflection signal into corresponding amplitude data by the detecting unit 2 and transmitting the amplitude data to the detecting system in real time; performing outlier rejection and normalization preprocessing on the received amplitude data by the detecting system, substituting the amplitude data into the optimal vector machine regression fitting model in the method for determining the physical quantity and frequency point of the microwave sensor, and inversely calculating the mineralization value of the water quality to be detected through the nonlinear correlation between the amplitude characteristic and the mineralization in the optimal vector machine regression fitting model.

[0013] This invention offers the following advantages: The antenna unit utilizes enameled wire, which possesses excellent insulation, corrosion resistance, and ductility. This not only isolates the antenna from water impurities, electrolyte corrosion, and short-circuit risks, ensuring interference-free signal transmission, but also allows for flexible adjustment of its shape, angle, and length within the water, adapting to detection scenarios with varying depths and spatial constraints. The transmitting unit integrates an oscillator, signal conditioning, and power control module, enabling precise control of the microwave signal's frequency and power to ensure signal stability. The detection unit, through three stages of processing—detection, amplification, and filtering—converts weak reflected signals into stable electrical signals, effectively filtering out electromagnetic interference and water clutter, thus improving the quality of the original detection data. Based on dynamic and anti-interference experimental data, a radial basis function kernel vector machine regression fitting strategy is employed, combined with a grid search method and five-fold cross-validation to optimize model parameters. This effectively eliminates outliers and dimensional differences, accurately fitting the nonlinear correlation between mineralization, gas-liquid ratio, frequency, and amplitude / phase characteristics, providing reliable data support for the selection of measured physical quantities. By screening the measured physical quantities using both monotonicity and sensitivity, amplitude characteristics were determined as the core detection indicator. Under different mineralization and gas-liquid ratio interferences, the amplitude characteristic exhibits a stable correlation with mineralization and its response sensitivity is significantly higher than that of phase characteristics, overcoming the pain point of traditional detection methods being susceptible to interference. Based on a comprehensive selection of correlation coefficient, data range, and standard deviation, the optimal frequency point was chosen. At this frequency point, the amplitude characteristic shows the strongest correlation with mineralization, the highest data discrimination, and the smallest dispersion, ensuring that the sensor can accurately capture changes in mineralization even under complex operating conditions. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the structure of the microwave sensor described in Example 1.

[0015] Figure 2 This is a frequency sweep curve of the amplitude characteristics of the microwave sensor described in Example 1.

[0016] Figure 3 This is a schematic diagram of the mineralization measurement system.

[0017] Figure 4 A flowchart illustrating the method for determining the physical quantities and frequency points measured by a microwave sensor.

[0018] Figure 5 This is a flowchart illustrating the method for measuring mineralization.

[0019] Figure 6 The sweep frequency curves show the amplitude characteristics at static conditions and different gas-liquid ratios when the mineralization is 5 g / L.

[0020] Figure 7 The amplitude sweep curves are shown for static and different gas-liquid ratios when the salinity is 5.25 g / L.

[0021] Figure 8This is a comparison of the amplitude characteristic sweep curves at mineralization levels of 5 g / L and 5.25 g / L.

[0022] Figure 9 The amplitude sweep curves are shown for static and different gas-liquid ratios when the mineralization is 10 g / L.

[0023] Figure 10 The amplitude sweep curves are shown for static and different gas-liquid ratios when the mineralization is 10.5 g / L.

[0024] Figure 11 This is a comparison of the amplitude characteristic sweep curves at mineralization levels of 10 g / L and 10.5 g / L.

[0025] Figure 12 The image shows the amplitude sweep curves for static and different gas-liquid ratios when the mineralization is 20 g / L (left figure).

[0026] Figure 13 The sweep frequency curves show the amplitude characteristics at static conditions and different gas-liquid ratios when the mineralization is 21 g / L (right figure).

[0027] Figure 14 This is a comparison of the amplitude characteristic sweep curves at mineralization levels of 20 g / L and 21 g / L.

[0028] Figure 15 The sweep frequency curves show the amplitude characteristics at static conditions and different gas-liquid ratios when the mineralization is 30 g / L.

[0029] Figure 16 The sweep frequency curves show the amplitude characteristics at static conditions and different gas-liquid ratios when the mineralization is 31.5 g / L (right figure).

[0030] Figure 17 This is a comparison of the amplitude characteristic sweep curves at mineralization levels of 30 g / L and 31.5 g / L.

[0031] Figure 18 The amplitude sweep curves are shown for static and different gas-liquid ratios when the mineralization is 40 g / L.

[0032] Figure 19 The sweep frequency curves show the amplitude characteristics at static conditions and different gas-liquid ratios when the mineralization is 42 g / L.

[0033] Figure 20 This is a comparison of the amplitude characteristic sweep curves at mineralization levels of 40 g / L and 42 g / L.

[0034] Figure 21 The image shows the phase characteristic sweep curves at static conditions and different gas-liquid ratios when the mineralization is 5 g / L.

[0035] Figure 22The image shows the phase characteristic sweep curves at static conditions and different gas-liquid ratios when the mineralization is 5.25 g / L.

[0036] Figure 23 This is a comparison of the phase characteristic sweep curves at mineralization levels of 5 g / L and 5.25 g / L.

[0037] Figure 24 The sweep frequency curves show the phase characteristics at static conditions and different gas-liquid ratios when the mineralization is 10 g / L.

[0038] Figure 25 The image shows the phase characteristic sweep curves at static conditions and different gas-liquid ratios when the mineralization is 10.5 g / L.

[0039] Figure 26 The image shows a comparison of the phase characteristic sweep frequency curves at mineralization levels of 10 g / L and 10.5 g / L.

[0040] Figure 27 The sweep frequency curves show the phase characteristics at static conditions and different gas-liquid ratios when the mineralization is 20 g / L.

[0041] Figure 28 The sweep frequency curves show the phase characteristics at static conditions and different gas-liquid ratios when the mineralization is 21 g / L.

[0042] Figure 29 This is a comparison of the phase characteristic sweep curves at mineralization levels of 20 g / L and 21 g / L.

[0043] Figure 30 The phase characteristic sweep curves are shown for static conditions and different gas-liquid ratios when the mineralization is 30 g / L.

[0044] Figure 31 The image shows the phase characteristic sweep curves at static conditions and different gas-liquid ratios when the mineralization is 31.5 g / L.

[0045] Figure 32 The image shows a comparison of the phase characteristic sweep curves at mineralization levels of 30 g / L and 31.5 g / L.

[0046] Figure 33 The sweep frequency curves show the phase characteristics at static conditions and different gas-liquid ratios when the mineralization is 40 g / L.

[0047] Figure 34 The sweep frequency curves show the phase characteristics at static conditions and different gas-liquid ratios when the mineralization is 42 g / L (right figure).

[0048] Figure 35 This is a comparison of the phase characteristic sweep curves at mineralization levels of 40 g / L and 42 g / L.

[0049] Figure 36The graph shows the regression fit of the vector machine, where x is the magnitude and y is the mineralization degree.

[0050] The components include: 1. Transmitting unit; 2. Detection unit; 3. Antenna unit; 4. Pipeline; 5. Peristaltic pump. Detailed Implementation

[0051] The technical solutions of the microwave sensor and mineralization measurement system for determining the physical quantity and frequency point measured by the microwave sensor provided by the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0052] Example 1 like Figures 1-2 As shown, a microwave sensor includes a transmitting unit 1, which generates microwave signals within a preset range.

[0053] Specifically, the transmitting unit 1, as the core of the sensor's signal transmission, functions to stably generate microwave signals within a preset range. This unit integrates a microwave oscillator, a signal conditioning module, and a power control module. It can precisely regulate the frequency and power of the microwave signal within a preset range according to actual detection needs, ensuring that the generated microwave signal possesses good stability and propagation characteristics. This allows it to effectively penetrate the water body being detected and form a recognizable reflected signal, providing a reliable signal source foundation for subsequent detection work.

[0054] The sensor also includes a detection unit 2, which converts the received microwave signal into an electrical signal. Specifically, the detection unit 2 plays a crucial role in receiving and converting the microwave reflected signal received by the antenna unit into an electrical signal that can be further processed. This unit incorporates a high-precision detection circuit, a signal amplification module, and a filtering module. Upon receiving the weak microwave reflected signal transmitted from the antenna unit, the detection circuit first demodulates the microwave signal into a low-frequency signal, then the amplification module amplifies the weak signal to a preset amplitude. Simultaneously, the filtering module filters out external electromagnetic interference, water clutter, and other irrelevant signals, ultimately outputting a stable and clear electrical signal, providing a high-quality signal input for subsequent data analysis and parameter calculation.

[0055] The sensor also includes: one end of each antenna unit 3 is connected to the corresponding transmitting unit 1 and detection unit 2, and the other end extends vertically into the water to be tested; wherein, the antenna unit 3 is enameled wire, and the use of enameled wire is to avoid chemical reaction and corrosion caused by contact between the end face and salt water. The frequency of the sensor is determined through dynamic experiments and anti-interference experiments. Specifically, the two antenna units are arranged symmetrically to cooperate in completing the transmission and reflection signal acquisition of microwave signals, forming a dual-line detection structure, which can effectively improve detection accuracy and anti-interference capability. Each antenna unit 3 adopts a bidirectional cooperative working mode, with one end precisely connected to the corresponding transmitting unit 1 and detection unit 2 respectively. The connection method uses a high-frequency sealed connector to ensure that the microwave signal has no leakage or attenuation during transmission, ensuring the integrity of signal transmission; the other end extends vertically or at a preset angle into the water to be tested. The insertion length can be flexibly adjusted according to the actual needs such as the depth of the water area to be tested and the detection range, ensuring that the microwave signal can fully contact the water body to be tested, maximizing the capture of microwave signals reflected by the water body, and improving the accuracy of detection data.

[0056] The antenna unit 3 uses enameled wire as its core material. This enameled wire possesses excellent insulation, corrosion resistance, and signal transmission performance, making it suitable for the complex conditions of water environment detection. The high-strength insulating varnish layer covering the surface of the enameled wire effectively isolates the water being detected from the wire core, preventing impurities and electrolytes in the water from corroding the wire and causing short circuits. It also avoids interference from the water on microwave signal transmission, ensuring the long-term stable operation of the antenna unit. Furthermore, the enameled wire is soft and highly malleable, allowing for flexible adjustment of its shape and angle when submerged in water according to the spatial constraints and layout requirements of the detection scenario. Its excellent conductivity enables efficient transmission of microwave signals, meeting both signal transmission and reception needs, further enhancing the sensor's detection performance and lifespan.

[0057] In this embodiment, the microwave sensor is a high-precision detection device specifically designed for water environment monitoring scenarios. It achieves accurate detection of water-related parameters through the transmission, reflection, and reception of microwave signals, and is widely used in aquaculture, water quality monitoring, water conservancy projects, and other fields requiring real-time water body monitoring. Simultaneously, it employs two symmetrical antenna units, combined with the stable microwave signal output of the transmitting unit and the anti-interference signal processing of the detection unit, effectively improving detection accuracy and anti-interference capabilities, ensuring data accuracy. The antenna units use enameled wire, which has excellent insulation and corrosion resistance, isolating water impurities and electrolytes, preventing corrosion short circuits, avoiding water interference with signal transmission, and adapting to long-term stable operation in complex water environments. The enameled wire is soft and highly malleable, allowing flexible adjustment of the antenna's shape, angle, and length in the water according to the depth and space limitations of the detection area; high-frequency sealed connectors ensure leak-free and attenuated signal transmission. Finally, the transmitting unit can precisely control microwave signal parameters, the antenna unit has excellent conductivity, and the detection unit can amplify and filter weak reflected signals, forming a complete and efficient microwave detection loop.

[0058] Example 2 like Figure 3 As shown, the mineralization measurement system is specially configured to meet the detection requirements of microwave sensors in dynamic and anti-interference experiments and to adapt to water quality parameter monitoring under different experimental conditions. As the core device for experimental condition simulation and media transportation, the mineralization measurement system works in conjunction with the microwave sensor described in Example 1 above to jointly complete the simulation of water quality changes in dynamic experiments, the simulation of interference factors in anti-interference experiments, and the real-time detection of corresponding parameters, providing a stable and controllable experimental environment for sensor performance testing.

[0059] The mineralization measurement system includes pipe 4. Specifically, pipe 4 serves as the transport channel for the experimental media (salt water solution, air) and also as the real-time detection carrier for the microwave sensor. It is made of corrosion-resistant, high-strength materials with smooth inner walls (such as polytetrafluoroethylene and quartz glass), adaptable to the corrosive characteristics of salt water solutions with different mineralization levels, preventing damage to the pipe from the media. At the same time, the smooth inner wall reduces the flow resistance of the media, prevents media residue, and ensures the smooth transport of the experimental media. The diameter and length of pipe 4 can be flexibly adjusted according to experimental needs. Its overall sealing performance is good, which can effectively prevent media leakage, ensure the cleanliness of the experimental environment and the accuracy of experimental data, and provide a reliable carrier foundation for sensor detection and media transport.

[0060] The mineralization measurement system includes at least two peristaltic pumps 5, each connected to a pipe 4. One of the peristaltic pumps 5 draws saline solutions of different mineralizations into the pipe; the other peristaltic pump 5 draws air into the pipe. The microwave sensor described in Examples 1-2 is mounted on the side wall of the pipe 4. The microwave sensor transmits the collected data to a preset detection system. Specifically, the mineralization measurement system is equipped with at least two independent peristaltic pumps 5, each connected to the pipe 4 in a sealed manner to collaboratively transport the experimental medium. The operating status of the two peristaltic pumps can be independently adjusted to adapt to different experimental conditions. One of the peristaltic pumps 5 is specifically designed to deliver brine solutions with different mineralization levels. This peristaltic pump 5 can precisely control the flow rate and delivery speed of the brine solution, and can switch between brine solutions with different mineralization levels according to the experimental settings. It can stably and uniformly pump the brine solution into the pipe 4, providing a dynamic water quality detection scenario for the microwave sensor. This meets the needs of detecting different water quality parameters in dynamic experiments and facilitates testing the sensor's detection accuracy and response speed during dynamic changes in water quality.

[0061] Another peristaltic pump 5 is used to draw air into the pipe 4. This peristaltic pump 5 can also precisely control the air delivery flow rate and frequency. By introducing air into the pipe, it simulates interference scenarios such as bubbles and aeration in natural water bodies, adapting to the needs of anti-interference experiments. It is used to test whether the microwave sensor can stably and accurately collect water quality parameters under the presence of air interference, verifying the sensor's anti-interference performance. When the two peristaltic pumps 5 work together, the delivery ratio and rhythm of the brine solution and air can be flexibly adjusted according to the experimental plan to construct diverse experimental conditions and comprehensively test the overall performance of the microwave sensor.

[0062] In this embodiment, the microwave sensor described in Embodiments 1-2 is sealed and mounted on the side wall of the pipe 4 via its mounting base 4. During installation, the mounting base 4 of the sensor is inserted into a pre-set mounting hole on the side wall of the pipe 4. A sealing structure is used to achieve a tight seal between the two, preventing leakage of the brine solution and air inside the pipe. At the same time, it ensures that the two antenna units 3 of the sensor pass through the mounting holes 5 of the mounting base 4 and extend into the brine solution inside the pipe 4, ensuring that the antenna units can normally transmit microwave signals and receive reflected signals. During operation, the microwave sensor collects relevant parameters of the brine solution inside the pipe 4 in real time, and transmits the collected detection data quickly and stably to a pre-set detection system through a pre-set signal transmission module. The detection system parses, stores, and analyzes the received data, facilitating real-time observation of the sensor's detection results by experimental personnel, comparison with experimental data, and thus completing the testing of the sensor's dynamic performance and anti-interference performance.

[0063] Example 3 like Figure 4 andFigure 36 As shown, a method for determining the physical quantity and frequency point measured by a microwave sensor is applicable to the mineralization measurement system in the above embodiment 2. The method includes: the mineralization measurement system introducing salt water solutions or air with different mineralization and preset mineralization changes into the pipeline.

[0064] Specifically, saline solutions with different mineralization levels can simulate real-world scenarios with varying salt content in natural water bodies, such as setting... , , , , Isogradient mineralization; preset mineralization changes can simulate the dynamic fluctuation process of water salinity, such as mineralization changes as... The introduction of air is used to simulate interference scenarios such as bubbles and aeration in water, adapting to the testing requirements of anti-interference experiments.

[0065] In this embodiment, brine solutions with varying mineralization and preset mineralization levels accurately simulate the static distribution and dynamic fluctuations of salinity in natural water bodies. Air is introduced to simulate bubble interference in real-world applications, achieving a high degree of replication between the experimental conditions and the actual peristaltic pump pipeline measurement scenario, thus avoiding measurement deviations caused by the disconnect between experiment and reality. Gradient mineralization settings, diverse mineralization variation modes, and multi-gradient gas-liquid ratio interference settings provide comprehensive experimental data for subsequent vector machine regression fitting, monotonicity assessment, sensitivity comparison, and frequency point selection, ensuring the universality and reliability of the analysis results in each subsequent step. The scenario settings encompass static, dynamic, and anti-interference scenarios, progressing step by step. This ensures the integrity of the basic experimental data while enhancing the specificity of the anti-interference experiments, providing a preliminary guarantee for the anti-interference performance and measurement accuracy of the subsequent mineralization measurement method. The mineralization gradient, mineralization change pattern, and air injection rate can all be adjusted according to actual experimental needs and application scenarios to adapt to the measurement requirements of different water environments. At the same time, the control requirements for the injection process are clear, which facilitates standardized operation by experimental personnel and ensures the stability and consistency of experimental conditions.

[0066] The method also includes: when conducting dynamic experiments, turning on a peristaltic pump in the mineralization measurement system to continuously pump brine solutions of different mineralization into the pipeline, and recording the amplitude and phase data under different mineralization and preset mineralization changes through a microwave sensor.

[0067] Specifically, the peristaltic pump specifically designed for delivering brine solutions in the salinity measurement system is turned on, and its rotation speed and flow rate are adjusted to preset stable values. This allows the peristaltic pump to continuously and uniformly pump brine solutions of different salinities into pipe 4 according to a preset salinity gradient or a preset salinity change pattern, ensuring that the salinity of the brine solution in pipe 4 remains stable in a preset state or changes dynamically according to a preset pattern. Simultaneously, the microwave sensor is activated to continuously collect and record amplitude and phase data in real time for different salinities and changes in salinity. For each set of salinity parameters, 30-50 sets of amplitude and phase data are collected to ensure the representativeness and reliability of the data. All collected data is transmitted to the detection system in real time for storage and subsequent retrieval.

[0068] In this embodiment, a dedicated peristaltic pump delivers brine solutions of varying mineralization at a constant speed, ensuring a continuous and stable supply. This effectively avoids issues such as mineralization fluctuations and delivery interruptions, guaranteeing the core conditions of "no gas-liquid interference and controllable mineralization" in the dynamic experiment, laying the foundation for the accuracy of subsequent fitting data. The number of samples collected for each set of mineralization parameters is clearly defined. Random errors are eliminated through repeated sampling, ensuring the representativeness of the collected amplitude and phase data. Simultaneously, the data is transmitted in real-time and stored in categories for easy retrieval and organization, improving experimental efficiency. The dynamic experimental condition settings are closely integrated with vector machine regression fitting, achieving seamless integration of condition control, data acquisition, data storage, and fitting applications. This provides sufficient and accurate raw data support for the subsequent fitting process, ensuring that the fitted curve accurately reflects the correlation between mineralization and amplitude / phase characteristics under interference-free conditions. The peristaltic pump speed, flow rate, and data acquisition quantity are all clearly defined and can be flexibly adjusted according to experimental needs, facilitating standardized operation by experimental personnel. Furthermore, the sensor and peristaltic pump work synchronously, simplifying the experimental process and reducing operational difficulty. Stable operating conditions are ensured by uniform speed delivery, and random errors are reduced by multiple data acquisitions. Errors are controlled from the source of data acquisition to avoid the accuracy of subsequent measurement of physical quantities and frequency point selection due to data deviations.

[0069] The method also includes: when conducting anti-interference experiments, one peristaltic pump in the mineralization measurement system continuously pumps salt water solutions of different mineralization into the pipeline, and another peristaltic pump continuously pumps air into the pipeline. Different gas-liquid ratios are controlled by controlling the rotation speed of the two peristaltic pumps, and amplitude and phase data under different mineralization and preset mineralization changes are recorded by a microwave sensor.

[0070] Specifically, the peristaltic pump delivering the brine solution in the salinity measurement system is kept running continuously. According to the preset parameters of the dynamic experiment, brine solutions with different salinities and preset salinity variations are continuously pumped into pipe 4 to ensure that the salinity of the brine solution in pipe 4 meets the experimental requirements. Simultaneously, another peristaltic pump specifically for delivering air is turned on. Based on the preset gas-liquid ratio gradient, the speed and flow rate of this pump are adjusted to continuously and uniformly pump air into pipe 4. By controlling the speed ratio of the two peristaltic pumps, the mixing ratio of the brine solution and air in pipe 4, i.e., the gas-liquid ratio, is precisely controlled to simulate different levels of air interference. During this process, the microwave sensor continues to operate, collecting and recording amplitude and phase data in real time under different salinities, different salinity variations, and different gas-liquid ratios. Similarly, 30-50 sets of data are collected for each set of salinity and gas-liquid ratio conditions to ensure data coverage of various interference scenarios, providing sufficient data support for subsequent analysis of the sensor's anti-interference characteristics.

[0071] In this embodiment, the coordinated control of two peristaltic pumps ensures the stability of the mineralization condition while precisely controlling the gas-liquid ratio gradient. This perfectly replicates the interference scenarios such as bubbles and aeration within the actual peristaltic pump pipeline, solving the problem that a single peristaltic pump cannot simulate gas-liquid mixing interference conditions and avoiding a disconnect between experiments and practical applications. By adjusting the gas-liquid ratio through rotational speed, precise switching between different interference levels can be achieved. Furthermore, the mineralization and gas-liquid ratio conditions are independent and do not interfere with each other, ensuring the stability and repeatability of each combination of conditions and providing standardized experimental data for subsequent anti-interference performance analysis. The system covers all combinations of conditions with different mineralization, interference levels, and mineralization trends, and each condition is repeatedly sampled to effectively eliminate random errors. This provides sufficient and reliable raw data support for subsequent anti-interference experiment fitting, sensitivity stability comparison, and frequency point selection. The system seamlessly integrates anti-interference experimental setup with vector machine regression fitting, achieving seamless integration of dual-pump collaborative control, data acquisition, classification and storage, and fitting applications. This ensures that the fitted curve accurately reflects the correlation between salinity, gas-liquid ratio, amplitude, and phase characteristics under interference conditions. The speed and flow rate control standards for the two peristaltic pumps are clear, the data acquisition quantity is well-defined, and the salinity gradient, gas-liquid ratio gradient, and acquisition quantity can be flexibly adjusted according to experimental needs, facilitating standardized operation by experimental personnel while maintaining both experimental flexibility and standardization. Sixth, errors are controllable. Dual-pump collaboration ensures stable operating conditions, and multiple acquisitions reduce random errors, controlling interference factors and data deviations from the data acquisition source to ensure the accuracy of subsequent anti-interference performance assessments and frequency point selection.

[0072] The method also includes: using a vector machine regression fitting strategy to perform curve fitting on amplitude and phase data under different mineralization degrees and preset mineralization degree changes measured in dynamic experiments and anti-interference experiments, respectively, to obtain amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under static conditions and different gas-liquid ratios with different mineralization degrees, as well as comparison diagrams of amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under preset mineralization degree changes.

[0073] Furthermore, the vector machine regression fitting strategy includes: preprocessing the raw data collected in the dynamic experiment and the anti-interference experiment to remove outliers from the discrete amplitude and phase data; normalizing the effective amplitude and phase data after removing outliers, mapping the amplitude data, phase data, mineralization parameter, gas-liquid ratio parameter, and frequency parameter to... Within the range.

[0074] Specifically, all raw amplitude and phase data collected in the dynamic and anti-interference experiments were first screened one by one, using... The principle is to remove outliers from discrete data, i.e., those values ​​that exceed the data mean. Data was first analyzed to the maximum standard deviation to avoid interference from outliers in the fitting results. Then, the effective amplitude and phase data, along with the corresponding mineralization, gas-liquid ratio, and frequency parameters, were normalized using a linear normalization method, mapping all parameters to [the appropriate normalization level]. Within the range, eliminate values ​​such as mineralization as a percentage, amplitude as a voltage value, and frequency as... By considering the dimensional differences of the various parameters, we can ensure that the weights of each parameter are balanced during the fitting process, thereby improving the accuracy and reliability of the fitting model.

[0075] In this embodiment, by The principle of precisely eliminating outlier data and avoiding fitting bias caused by sudden interference is followed by normalization to eliminate dimensional differences and balance parameter weights, thereby improving the accuracy and reliability of the fitting model from the source and ensuring that the fitted curve can truly reflect the correlation between each core parameter and amplitude and phase characteristics. The preprocessing steps simultaneously cover dynamic and anti-interference experimental data, and the fitting process takes into account both interference-free and interference-affected scenarios. The output frequency sweep curves and curve comparison charts fully cover all curve data required for subsequent physical quantity screening, achieving seamless integration of preprocessing, fitting, and application. Preprocessing is implemented in steps with clear standards, and the same optimal model is reused for curve fitting, ensuring the standardization of the experimental process and avoiding fitting bias caused by different processing methods and models, thus improving the repeatability and comparability of experimental results. The comparison data of the continuous curves output by fitting not only facilitates intuitive observation of the influence of mineralization, gas-liquid ratio, and frequency on amplitude and phase, but can also be directly used for subsequent sensitivity calculations, frequency point screening, and other steps, significantly improving experimental analysis efficiency and providing reliable fitting data support for the entire mineralization measurement process. The preprocessing method is simple, easy to understand, and highly operable. It does not require complex professional equipment, making it easy for experimenters to implement in a standardized manner. At the same time, the normalization process allows the model to adapt to experimental data with different parameter ranges, improving the flexibility and adaptability of the fitting strategy.

[0076] The vector machine regression fitting strategy further includes: defining the radial basis function kernel function as the kernel function of the vector machine regression model, which is used to fit the effective mineralization, gas-liquid ratio, frequency and amplitude, and phase data into a nonlinear relationship, and introducing a penalty coefficient and an insensitive loss function parameter.

[0077] Specifically, after data preprocessing, a vector machine regression fitting model is constructed and its core parameters are determined: the radial basis function kernel is defined as the kernel function of the vector machine regression model. This kernel function has good nonlinear mapping capabilities, and its core expression is: , in, The radial basis function value is used to measure the similarity between input sample vectors, realize the nonlinear mapping from the input space to the high-dimensional feature space, and provide core support for vector machine regression fitting. The current input sample vector consists of preprocessed effective mineralization, gas-liquid ratio, and frequency parameters, and is a combination of input variables for model fitting. The training set sample vectors; The kernel function parameter controls the radial range of the kernel function and affects the model's ability to fit nonlinear relationships. The larger the value, the narrower the range of the kernel function and the more refined the model fit. For the input sample vector With training set sample vectors The square of the Euclidean distance between the two sample vectors is used to quantify their similarity.

[0078] It can effectively fit valid data on mineralization, gas-liquid ratio, frequency, amplitude, and phase into a nonlinear relationship. Meanwhile, to avoid overfitting or underfitting issues in the model, a penalty coefficient is introduced. and insensitive loss function parameters This lays the foundation for model optimization. Penalty coefficient. To balance the accuracy and complexity of model fitting, an initial range of values ​​is selected. Insensitive loss function parameters Used to control the acceptable range of fitting error, initial value range .

[0079] In this embodiment, a radial basis function (RBF) kernel is selected, which can efficiently map the complex nonlinear relationships between mineralization, gas-liquid ratio, frequency and amplitude, and phase data to a high-dimensional feature space, accurately capturing the inherent laws between parameters, solving the nonlinearity problem that linear models cannot fit, and significantly improving fitting accuracy. By introducing a penalty coefficient and an insensitive loss function parameter, the problems of overfitting and underfitting are effectively avoided, balancing fitting accuracy and model complexity, ensuring that the model can function stably in both the training set and new experimental data, and adapting to the data fitting needs under different working conditions. Clarifying the core roles and initial value ranges of the kernel function parameters, penalty coefficient, and insensitive loss function parameters lays the foundation for subsequent model parameter optimization and facilitates flexible adjustments by experimenters based on actual experimental data, improving the model's adaptability. The kernel function fits based on preprocessed effective data, making full use of the data advantages brought by outlier removal and normalization, further enhancing the reliability of the fitting results, and providing high-quality fitting model support for subsequent frequency sweep curve generation, measurement physical quantity judgment, and frequency point selection.

[0080] The vector machine regression fitting strategy further includes: obtaining the optimal penalty coefficient, kernel function parameters, and insensitive loss function parameters through a combination of grid search and five-fold cross-validation, and constructing the optimal vector machine regression fitting model. Specifically, within a preset parameter value range, all possible parameter combinations are generated; the preprocessed valid data is randomly divided into 5 groups, with 4 groups serving as the training set and 1 group as the validation set, and these groups are validated in rotation. Each parameter combination corresponds to one five-fold cross-validation, and the fitting error of the validation set is calculated; the parameter combination with the smallest fitting error on the validation set is selected as the optimal parameter combination, and after being substituted into the vector machine regression model, the optimal vector machine regression fitting model is obtained.

[0081] Furthermore, the preprocessed effective amplitude and phase data are divided into training and test sets according to a preset ratio. The training set is used to build a vector machine regression fitting model, and the test set is used to verify the model's fitting accuracy and generalization ability.

[0082] Furthermore, the predefined test set data is substituted into the fitting model to calculate the fitting error and coefficient of determination for the test set data, thus validating the fitting results. The fitting error can be the root mean square error or the mean absolute error. When the coefficient of determination is greater than a preset threshold and the fitting error is less than a preset range, the fitting model's accuracy meets the experimental requirements, and the fitted curve is correct. If the coefficient of determination is less than the preset threshold and the fitting error is greater than the preset range, the values ​​of the penalty coefficient, kernel function parameters, and insensitive loss function parameters are adjusted, and fitting and validation are performed again until the coefficient of determination is greater than the preset threshold and the fitting error is less than the preset range.

[0083] Furthermore, if the fitted curve fails to capture the data variation pattern, the penalty coefficient is increased; if the training set fits well but the test set has a large deviation, the penalty coefficient is decreased; if the fitted curve cannot capture the complex relationship between mineralization, gas-liquid ratio, amplitude, or phase, the kernel function parameter is gradually increased; if the fitted curve fluctuates greatly and has poor fit, the kernel function parameter is decreased; if the actual fitting deviation frequently exceeds the limit, the insensitive loss function parameter is increased; if some high-frequency data shows large fitting deviation, the insensitive loss function parameter is decreased. Each time, only one of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter is adjusted, while the values ​​of the other two parameters are fixed. The optimal value of the parameter is selected using the five-fold cross-validation method. After adjusting the values ​​of the three parameters in sequence, the values ​​of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter are fine-tuned as a whole to obtain the optimal values ​​of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter.

[0084] In this embodiment, firstly, overfitting or underfitting of the model is effectively avoided. The model's generalization ability is verified through an independent test set, ensuring that the model can not only adapt to the training data but also accurately fit new working condition data not involved in the training, thus improving the model's practicality and reliability. Secondly, the fitting error is evaluated using a combination of root mean square error and mean absolute error, providing a comprehensive and objective assessment of the model's fitting accuracy. This captures overall deviation while also considering the robustness of the fit, avoiding the limitations of a single error index. Thirdly, clear verification criteria and parameter iteration adjustment logic provide a clear direction for model optimization, ensuring that the final fitted model meets the accuracy standards and accurately reflects the correlation between sensor detection characteristics and various influencing factors, providing accurate and reliable data support for the subsequent determination of physical quantities and frequency points. Finally, a standardized model verification process enhances the scientific rigor and scientific nature of the experiment, reducing the impact of human error on the experimental results and ensuring the persuasiveness and repeatability of the experimental conclusions. Fifthly, parameter iteration optimization brings the fitted model to its optimal state, maximizing the exploration of the inherent patterns in the experimental data and providing technical support for improving sensor detection accuracy. This ensures that the sensor can achieve high-precision detection under different mineralization and gas-liquid ratio conditions.

[0085] The vector machine regression fitting strategy further includes: using the mineralization degree, preset mineralization degree change, and frequency measured in the dynamic experiment as input variables, and the corresponding amplitude data and phase data as output variables, substituting them into the optimal vector machine regression fitting model, and using the nonlinear mapping capability of the model to fit the discrete amplitude and phase data into continuous curves, thereby obtaining amplitude characteristic sweep curves and phase characteristic sweep curves under different mineralization degrees, as well as comparative data of amplitude characteristic sweep curves and phase characteristic sweep curves under preset mineralization degree changes.

[0086] Specifically, dynamic experimental fitting is one of the core application scenarios of vector machine regression fitting strategy. The core purpose is to capture the intrinsic relationship between mineralization, frequency, amplitude, and phase characteristics under conditions without gas-liquid interference, providing basic data support for the subsequent selection of measured physical quantities. During the fitting process, the core data under all effective working conditions in the dynamic experiment are first extracted. The input variables are strictly selected from the measured mineralization, the preset mineralization change, and all frequency points within the experimental frequency sweep range in the dynamic experiment, ensuring that the input variables fully cover the core influencing factors of the dynamic experiment. The output variables are selected from the measured discrete amplitude data and phase data under each combination of input variables. These data need to undergo preprocessing such as outlier removal using the 3σ criterion and normalization within the [0,1] interval to maintain consistency with the training data format of the optimal vector machine regression fitting model, avoiding the influence of data deviation on the fitting effect. The preprocessed input and output variables are substituted into the optimal vector machine regression fitting model, which has passed the grid search method and five-fold cross-validation. Utilizing the model's powerful nonlinear mapping capability, the discrete amplitude and phase data are fitted to a continuous model. During the fitting process, the model automatically captures the complex nonlinear correlation between the interaction of mineralization, the preset mineralization variation, and frequency, eliminating the influence of random fluctuations in discrete data and fitting the scattered experimental data points into a smooth, continuous curve. The final output consists of two core results: first, sweep curves of amplitude and phase characteristics under different mineralization levels, each curve corresponding to a fixed mineralization level, clearly showing the variation of amplitude and phase with frequency under that mineralization level; second, comparative data of sweep curves of amplitude and phase characteristics under preset mineralization variations, which integrates the sweep curves corresponding to different mineralization levels to form a curve comparison dataset, intuitively reflecting the influence trend of mineralization changes on amplitude and phase characteristics, providing intuitive curve support for subsequent monotonicity judgment.

[0087] The vector machine regression fitting strategy further includes: using the mineralization degree, preset mineralization degree change, gas-liquid ratio, and frequency measured in the anti-interference experiment as input variables, and the corresponding amplitude data and phase data as output variables, substituting them into the same optimal vector machine regression fitting model, continuously fitting the discrete data, and obtaining amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under different mineralization degrees and different gas-liquid ratios, as well as comparative data of amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under preset mineralization degree changes.

[0088] Specifically, anti-interference experimental fitting is another key application of the vector machine regression fitting strategy. It focuses on solving data fitting problems under complex working conditions, ensuring that the fitting results adapt to the interference environment of the actual measurement scenario, and providing data support for judging the anti-interference performance of measured physical quantities. Compared with dynamic experimental fitting, anti-interference experimental fitting adds the measured gas-liquid ratio parameter from the anti-interference experiment to the input variables. The input variables still retain mineralization, preset mineralization change, and frequency. These four variables together constitute a complete input system under the anti-interference scenario, comprehensively covering the three core influencing factors: mineralization, gas-liquid ratio, and frequency. This ensures that the fitting model can capture the correlation between various parameters and amplitude and phase characteristics under interference conditions.

[0089] The output variables are also selected as discrete amplitude and phase data under the corresponding operating conditions, and they need to undergo outlier removal and normalization preprocessing consistent with the dynamic experiment to ensure data accuracy and model fit. The same optimal vector machine regression fitting model is reused during fitting, without the need to rebuild the model or adjust the parameters. This ensures the uniformity of the fitting standard, improves fitting efficiency, and avoids fitting bias caused by different models. Through nonlinear mapping of the model, the discrete amplitude and phase data in the anti-interference experiment are fitted into continuous curves, ultimately outputting three core results: First, sweep frequency curves of amplitude and phase characteristics under different mineralization and gas-liquid ratios. Each curve corresponds to a combination of mineralization and gas-liquid ratio, clearly showing the law of amplitude and phase variation with frequency under different interference conditions; Second, sweep frequency curves of amplitude and phase characteristics under different mineralization, facilitating the observation of the impact of mineralization changes; Third, comparative data of sweep frequency curves of amplitude and phase characteristics under preset mineralization changes, integrating sweep frequency curves corresponding to different mineralization and gas-liquid ratios, intuitively reflecting the impact of mineralization changes on amplitude and phase characteristics under interference environment, and at the same time, comparing the stability of the mineralization influence trend under different gas-liquid ratio interference, providing accurate data support for subsequent sensitivity comparison and anti-interference performance judgment.

[0090] After fitting is completed, the fitting results of the two types of experiments need to be preliminarily verified, the fitting error and goodness of fit are calculated, and the fitting curve can truly and accurately reflect the changing pattern of the experimental data. If the fitting deviation exceeds the standard under a certain working condition, the corresponding experimental data needs to be re-verified. After confirming that there are no abnormalities, the fitting is repeated to lay a solid data foundation for the subsequent screening of physical quantities and determination of frequency points.

[0091] In this embodiment, two major scenarios—dynamic experiments and anti-interference experiments—are adapted to accurately capture the correlation between core parameters and amplitude and phase characteristics under different operating conditions, balancing basic data support with anti-interference scenario adaptation. The same optimal vector machine regression fitting model is reused, eliminating the need for repeated model construction. This ensures consistent fitting standards, reduces fitting bias, improves fitting efficiency, and simplifies the fitting process. The continuous frequency sweep curves and comparative data output from the fitting process fully cover all the basic data required for subsequent physical quantity selection and frequency point determination, achieving seamless integration from fitting to application without the need for additional experiments or fitting. Based on prior parameter optimization and data preprocessing, the fitting curves effectively eliminate random fluctuations in discrete data, accurately reflecting the influence trends of each parameter. Both the goodness of fit and error meet experimental requirements, providing accurate and reliable data support for subsequent steps. The curves and comparative data output from the fitting process facilitate intuitive observation of the influence patterns of each parameter and can be directly used for subsequent quantitative analysis, significantly improving the efficiency of subsequent experimental analysis and parameter selection, while providing core data assurance for the implementation of mineralization measurement methods.

[0092] The method further includes: observing and analyzing the amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under static conditions and different gas-liquid ratios with different mineralization, as well as the amplitude characteristic sweep frequency curve comparison chart and phase characteristic sweep frequency curve comparison chart under preset mineralization changes. By observing the monotonicity between amplitude characteristics and phase characteristics and mineralization, and comparing the sensitivity of amplitude characteristics to phase characteristics under different mineralization, the physical quantity measured by the sensor is determined to be amplitude characteristics.

[0093] Specifically, the amplitude and phase sweep frequency curves under static conditions and different gas-liquid ratios with different mineralization degrees, as well as the amplitude and phase sweep frequency curves under preset mineralization degree variations, were observed and analyzed. The judgment was based on two criteria: first, the monotonicity between amplitude and phase characteristics and mineralization degree; and second, the sensitivity comparison of the two under different mineralization degrees. Ultimately, the physical quantity measured by the sensor was determined to be the amplitude characteristic.

[0094] Monotonicity assessment is a fundamental prerequisite for selecting effective physical quantities that sensors can measure. The logic is that the physical quantity measured by the sensor must exhibit a stable and singular trend in relation to mineralization. This ensures that mineralization can be accurately inferred from the physical quantity, avoiding excessive measurement errors due to trend irregularities. The specific assessment process involves combining frequency sweep curves from both static and anti-interference experimental scenarios, and conducting observation and judgment step-by-step: 1. Within a fixed experimental frequency range, extract amplitude and phase characteristic data corresponding to different mineralization levels, and observe their trends with mineralization. If a physical quantity (amplitude / phase) shows a stable increasing or decreasing trend with increasing (or decreasing) mineralization, without significant fluctuations or inflection points, then the physical quantity is judged to have good monotonicity with mineralization. If the physical quantity changes without a fixed pattern with mineralization, exhibiting alternating increases and decreases, large fluctuations, or no significant correlation with mineralization, then its monotonicity is judged to be poor, making it unsuitable as a measurement physical quantity. 2. Keeping the gas-liquid ratio gradient constant, observe the changing trends of amplitude and phase characteristics under different mineralization degrees. Simultaneously, fix the mineralization degree and observe the stability of the correlation between amplitude, phase characteristics, and mineralization degree when the gas-liquid ratio changes. Key determination: Under different gas-liquid ratio disturbances, whether the monotonicity of amplitude, phase characteristics, and mineralization degree remains stable, i.e., changes in the gas-liquid ratio do not cause a reversal or excessive fluctuation in the trend of this physical quantity and mineralization degree. If the amplitude characteristic maintains a stable monotonic increasing / decreasing relationship with mineralization degree under different gas-liquid ratios, while the phase characteristic shows monotonic disorder with changes in the gas-liquid ratio—for example, within a certain mineralization degree range, an increase in the gas-liquid ratio causes the increasing trend of phase and mineralization degree to become decreasing—then the amplitude characteristic is preferentially determined to meet the monotonicity requirement. 3. Superimpose and compare the amplitude sweep curves and phase sweep curves corresponding to each gradient mineralization degree within the preset mineralization degree variation range, and visually observe the separation degree and consistency of the changing trends of the curves. If the amplitude characteristic curves show a clear and continuous separation trend with the change of mineralization, and the monotonicity of all curves is consistent, i.e., all increasing or all decreasing, then the monotonicity of the amplitude characteristic and mineralization is further confirmed to be excellent; if the phase characteristic curves intersect or overlap after being superimposed, or the monotonicity of the curves corresponding to different mineralizations is inconsistent, then it indicates that the monotonicity of the phase characteristic and mineralization cannot meet the measurement requirements.

[0095] Sensitivity comparison is a key basis for determining the optimal physical quantity to be measured. The core logic is that, under the same frequency, gas-liquid ratio, and mineralization gradient, the more sensitive the physical quantity to changes in mineralization, the higher its measurement accuracy, and the more suitable it is as the core physical quantity for a sensor. Sensitivity here is defined as the change in amplitude or phase characteristics caused by a unit change in mineralization; the larger the change, the higher the sensitivity. The specific comparison method combines experimental data and frequency sweep curves, and is carried out on three levels: 1. For data with different mineralization gradients in static experiments and anti-interference experiments, calculate the amplitude characteristic sensitivity and phase characteristic sensitivity respectively. Set the mineralization gradient as... The corresponding change in amplitude characteristic is The change in phase characteristics is Then amplitude sensitivity Phase sensitivity Calculate the values ​​for each mineralization gradient. and 1. Calculate the average of all gradients to obtain the average amplitude sensitivity and average phase sensitivity, which serve as the core indicators for quantitative comparison. 2. Divide the mineralization into three intervals: low, medium, and high. Compare the magnitude and stability of amplitude sensitivity and phase sensitivity within each interval. Key observations: Whether the amplitude sensitivity is consistently greater than the phase sensitivity across the entire mineralization interval, with minimal fluctuation; if the phase sensitivity is greater than the amplitude sensitivity in a certain interval, but this interval is not within the pre-set mineralization measurement range, its influence can be ignored; if the amplitude sensitivity is significantly greater than the phase sensitivity across all pre-set measurement ranges of mineralization, then the amplitude sensitivity is considered superior. 3. Under different gas-liquid ratio interferences, calculate the amplitude sensitivity and phase sensitivity corresponding to each mineralization gradient and compare their stability. If the amplitude sensitivity fluctuation is less than the preset percentage when the gas-liquid ratio changes, while the phase sensitivity fluctuation is greater than the preset percentage, it indicates that the amplitude characteristics still have high sensitivity and stability in response to changes in salinity under interference conditions and are not easily affected by gas-liquid ratio interference. Conversely, if the phase sensitivity fluctuation is too large, it indicates that its response to changes in salinity is easily interfered with, the sensitivity stability is poor, and the measurement accuracy cannot be guaranteed.

[0096] In summary, based on the comprehensive analysis of the two judgment criteria mentioned above, it can be concluded that the amplitude characteristic and the mineralization degree have good monotonicity. Moreover, under different mineralization degree ranges and different anti-interference conditions, the amplitude characteristic is significantly more sensitive to changes in mineralization degree than the phase characteristic and has better stability. Therefore, the measured physical quantity of this bilinear sensor is finally determined to be the amplitude characteristic.

[0097] In this embodiment, monotonicity and sensitivity are the two core dimensions used as the judgment criteria, covering all experimental scenarios, including static and anti-interference scenarios. This ensures both the correlation between the measured physical quantity and mineralization and the response sensitivity and stability, avoiding measurement errors caused by unreasonable measurement of physical quantities from the source. Closely integrated with the experimental conditions of the peristaltic pump system, the performance under gas-liquid ratio interference is considered. The selected amplitude characteristics can withstand interference from complex operating conditions and are suitable for measurement needs of different mineralization and gas-liquid ratios, avoiding the limitations of screening results in a single scenario. Monotonicity observation and sensitivity calculation both have clear implementation methods, requiring no complex professional equipment, making it easy for experimental personnel to implement. At the same time, the comparative verification process further improves the accuracy of the judgment results. Subsequent measurements lay a solid foundation. The determined amplitude characteristics have advantages in correlation, sensitivity, and stability, providing reliable support for optimal frequency point selection and mineralization back-calculation, ensuring the accuracy and reliability of the entire mineralization measurement process, and forming a complete logical closed loop from judgment to screening to application.

[0098] The method also includes: calculating and comparing the amplitude characteristic correlation coefficient, data range and standard deviation under different mineralization and different gas-liquid ratios, and taking the frequency point corresponding to the amplitude correlation coefficient closest to 1, the largest data range and the smallest standard deviation.

[0099] Specifically, after determining that the physical quantity measured by the sensor is the amplitude characteristic, it is necessary to further screen the optimal frequency point to ensure that the sensor has the most stable response and the highest measurement accuracy at that frequency point for different mineralization and gas-liquid ratios. The core logic is that the optimal frequency point must simultaneously satisfy "the strongest correlation between amplitude characteristics and mineralization, the smallest data dispersion, and the best measurement stability". Therefore, by calculating and comparing the correlation coefficient, data range, and standard deviation of amplitude characteristics under different mineralization and gas-liquid ratios, and combining the comprehensive performance of the three indicators, the final frequency point is determined. The specific implementation steps are as follows: 1. Basic data extraction and organization: Based on the amplitude characteristic sweep curve obtained from the dynamic experiment and anti-interference experiment, extract the amplitude characteristic data corresponding to each frequency point in the full frequency range, and classify and organize them according to the combination of different mineralization and different gas-liquid ratios. Among them, mineralization covers all the gradients preset in the experiment, and gas-liquid ratio covers all the gradients controlled in the anti-interference experiment, ensuring that each frequency point has complete multi-condition amplitude data to support it, and avoiding calculation deviations due to missing data. 2. For each frequency point, the correlation coefficient, data range, and standard deviation of the amplitude characteristics were calculated. The calculation methods for these three indicators closely match the characteristics of the experimental data, as follows: The amplitude characteristic correlation coefficient reflects the strength of the linear correlation between amplitude characteristics and mineralization at that frequency point. The closer the absolute value of the correlation coefficient is to 1, the stronger the correlation between the two, and the higher the accuracy of inferring mineralization from the amplitude data at that frequency point. In the calculation, mineralization was used as the independent variable, and the amplitude characteristic data under the corresponding operating conditions was used as the dependent variable. The Pearson correlation coefficient calculation method was employed to calculate the amplitude-mineralization correlation coefficient for each frequency point under different gas-liquid ratios. The average value of the correlation coefficients under all gas-liquid ratios was then taken as the final correlation coefficient for that frequency point. , The range of values ​​is , focus The frequency points were ultimately selected. The frequency point closest to 1. The amplitude characteristic data range reflects the distribution span of amplitude characteristic data under different salinity and gas-liquid ratio conditions at that frequency point. The larger the data range, the higher the discrimination of amplitude characteristics against changes in salinity and gas-liquid ratio at that frequency point, and the clearer the reflection of differences between different operating conditions, avoiding measurement ambiguity due to data overlap. The calculation method is as follows: For each frequency point, extract amplitude data under different salinity and gas-liquid ratio conditions, and calculate the difference between the maximum and minimum values ​​in the data set. , , The larger the standard deviation, the better the amplitude data discrimination at that frequency point. 3. The core characteristic is the dispersion of amplitude characteristic data under the same salinity and gas-liquid ratio conditions at that frequency point. The smaller the standard deviation, the better the repeatability and stability of the amplitude measurement at that frequency point, the less affected by experimental random noise, and the more reliable the measurement results. The calculation method is as follows: For each frequency point, group the amplitude data according to the same salinity and gas-liquid ratio, calculate the standard deviation of each group, and then take the average of the standard deviations of all groups as the final standard deviation for that frequency point. , The smaller the value, the better the stability of the amplitude measurement. This is generally required. 3. Multi-indicator comprehensive comparison and frequency point selection: Three indicators are compared for each frequency point across the entire frequency range, namely... , and A comprehensive comparison was conducted, and the optimal frequency points were determined using priority ranking and threshold filtering to ensure that all three indicators simultaneously met the requirements. The specific filtering rules are as follows: First, frequency points that do not meet the basic requirements were eliminated, i.e., the remaining frequency points were... , The first step is to select frequency points that meet the threshold requirements, ensuring that the selected frequency points have basic correlation and stability. The second step is to rank the frequency points according to their correlation coefficients. The closest to 1 to the data range Maximum to standard deviation Sort by lowest priority and select first. The frequency point closest to 1; if multiple frequency points exist Difference Then choose one of them. The highest frequency point; if and If they are all similar, then choose The third step is to substitute the selected optimal frequency point into the original amplitude characteristic sweep curve to verify the correlation, discrimination, and stability of the amplitude data at that frequency point. After confirming that there are no anomalies, the frequency point is finally determined as the operating frequency point of the bilinear sensor.

[0100] The above method can be used to select the optimal frequency point that takes into account correlation, discrimination and stability, so as to ensure that the sensor can achieve accurate and stable measurement of mineralization based on amplitude characteristics at this frequency point, while effectively resisting gas-liquid ratio interference and adapting to the experimental measurement requirements of peristaltic pump system.

[0101] In this embodiment, correlation, discriminability, and stability are the three core indicators for selection. Quantitative calculations replace subjective judgment, and each step has clear calculation standards and selection rules, avoiding randomness in frequency selection from the outset and ensuring that the selected frequencies meet experimental requirements. By fully integrating multi-condition data from dynamic and anti-interference experiments, covering different mineralization and gas-liquid ratio scenarios, the selected optimal frequency can resist gas-liquid ratio interference and adapt to the actual measurement conditions of the peristaltic pump system, preventing the frequency selected under a single condition from failing in complex scenarios. The optimal frequency has advantages such as strong correlation, low data dispersion, and high stability, enabling the sensor to accurately capture changes in mineralization based on amplitude characteristics, reducing measurement errors, and providing reliable frequency support for subsequent mineralization back-calculation. Fourthly, it is highly operable, with clear steps for data extraction, indicator calculation, and frequency selection. The calculation methods are simple and easy to understand, requiring no complex professional equipment, making it easy for experimental personnel to implement. Simultaneously, the verification process further avoids selection errors, ensuring the accuracy and reliability of frequency selection.

[0102] Experimental Example Salt solutions with concentrations of 5 g / L, 10 g / L, 20 g / L, 30 g / L, and 40 g / L, as well as a 5% change in salinity, were prepared. Dynamic experiments and bubble anti-interference experiments were conducted using the microwave sensor described in Example 1. Static experiments, dynamic experiments, and two-phase dynamic experiments with different gas-liquid ratios were performed. The amplitude characteristic test results are as follows: Depend on Figures 6-7 It can be seen that when the mineralization is 5 g / L and 5.25 g / L, the amplitude characteristic curves of the sensor under static, dynamic and different gas-liquid ratios almost overlap between 52.7685 MHz and 67.7595 MHz. This indicates that the relationship between amplitude characteristics and mineralization in this frequency range is only slightly affected by the water flow rate and different gas-liquid ratios, and can be ignored.

[0103] Depend on Figure 8 It can be seen that when the mineralization changes by 5%, that is, from 5 g / L to 5.25 g / L, the measured amplitude characteristic curve varies greatly between 52.7685 MHz and 67.7595 MHz, showing obvious differentiation. This indicates that within this frequency range, the sensor is relatively sensitive to amplitude characteristics when the mineralization is 5 g / L and changes by 5%.

[0104] Depend on Figures 9-10 It can be seen that when the mineralization is 10 g / L and 10.5 g / L, the amplitude characteristic curves of the sensor under static, dynamic and different gas-liquid ratios almost overlap between 52.7685 MHz and 67.7595 MHz. This indicates that the relationship between amplitude characteristics and mineralization in this frequency range is only slightly affected by the water flow rate and different gas-liquid ratios, and can be ignored.

[0105] Depend on Figure 11 It can be seen that when the mineralization changes by 5%, that is, from 10 g / L to 10.5 g / L, the measured amplitude characteristic curve varies greatly between 52.7685 MHz and 67.7595 MHz, showing obvious differentiation. This indicates that within this frequency range, the sensor is relatively sensitive to amplitude characteristics when the mineralization is 10 g / L and changes by 5%.

[0106] Depend on Figures 12-13 It can be seen that when the salinity is 20 g / L and 21 g / L, the amplitude characteristic curves of the sensor under static, dynamic and different gas-liquid ratios almost overlap between 52.7685 MHz and 67.7595 MHz. This indicates that the relationship between amplitude characteristics and salinity in this frequency range is only slightly affected by the water flow rate and different gas-liquid ratios, and can be ignored.

[0107] Depend on Figure 14 It can be seen that when the mineralization changes by 5%, that is, from 20 g / L to 21 g / L, the measured amplitude characteristic curve varies greatly between 52.7685 MHz and 67.7595 MHz, showing obvious differentiation. This indicates that within this frequency range, the sensor is relatively sensitive to amplitude characteristics when the mineralization is 20 g / L and changes by 5%.

[0108] Depend on Figures 15-16 It can be seen that when the salinity is 30 g / L and 31.5 g / L, the amplitude characteristic curves of the sensor under static, dynamic and different gas-liquid ratios almost overlap between 52.7685 MHz and 67.7595 MHz. This indicates that the relationship between amplitude characteristics and salinity in this frequency range is only slightly affected by the water flow rate and different gas-liquid ratios, and can be ignored.

[0109] Depend on Figure 17 It can be seen that when the mineralization changes by 5%, that is, from 30 g / L to 31.5 g / L, the measured amplitude characteristic curve varies greatly between 52.7685 MHz and 67.7595 MHz, showing obvious differentiation. This indicates that within this frequency range, the sensor is relatively sensitive to amplitude characteristics when the mineralization is 30 g / L and changes by 5%.

[0110] Depend on Figures 18-19 It can be seen that when the mineralization is 40 g / L and 41.5 g / L, the amplitude characteristic curves of the sensor under static, dynamic and different gas-liquid ratios almost overlap between 52.7685 MHz and 67.7595 MHz. This indicates that the relationship between amplitude characteristics and mineralization in this frequency range is only slightly affected by the water flow rate and different gas-liquid ratios, and can be ignored.

[0111] Depend on Figure 20 It can be seen that when the mineralization changes by 5%, that is, from 40 g / L to 41.5 g / L, the measured amplitude characteristic curve varies greatly between 52.7685 MHz and 67.7595 MHz, showing obvious differentiation. This indicates that within this frequency range, the sensor is relatively sensitive to amplitude characteristics when the mineralization is 40 g / L and changes by 5%.

[0112] The experiment included static tests with pure water, dynamic tests, and two-phase dynamic tests with different gas-liquid ratios. The phase characteristic test results are as follows: within the measurement range, the relationship between the phase characteristics and salinity of this sensor is minimally affected by different gas-liquid ratios.

[0113] Depend on Figures 21-22 It can be seen that when the mineralization is 5 g / L and 5.25 g / L, the phase characteristic curves of the sensor almost overlap under static, dynamic and different gas-liquid ratios within the measurement range. This indicates that the relationship between the phase characteristic and mineralization is only slightly affected by the water flow rate and different gas-liquid ratios within this frequency range, and can be ignored.

[0114] Depend on Figure 23 It can be seen that when the mineralization changes by 5%, that is, from 5 g / L to 5.25 g / L, the measured phase characteristic curve changes significantly in the low frequency range and has obvious distinguishability. This indicates that in the low frequency range, the sensor is more sensitive to the phase characteristics when the mineralization is 5 g / L and the mineralization changes by 5%.

[0115] Depend on Figures 24-25 It can be seen that when the mineralization is 10 g / L and 10.5 g / L, the phase characteristic curves of the sensor almost overlap under static, dynamic and different gas-liquid ratios within the measurement range. This indicates that the relationship between the phase characteristic and mineralization is only slightly affected by the water flow rate and different gas-liquid ratios within this frequency range, and can be ignored.

[0116] Depend on Figure 26 It can be seen that when the mineralization changes by 5%, that is, from 10 g / L to 10.5 g / L, the measured phase characteristic curve changes little in the low frequency range and does not have obvious distinguishability. This indicates that in the low frequency range, the sensor is not sensitive to the phase characteristics when the mineralization is 10 g / L and the mineralization changes by 5%.

[0117] Depend on Figures 27-28It can be seen that when the mineralization is 20 g / L and 21 g / L, the phase characteristic curves of the sensor almost overlap under static, dynamic and different gas-liquid ratios within the measurement range. This indicates that the relationship between the phase characteristic and mineralization is only slightly affected by the water flow rate and different gas-liquid ratios within this frequency range, and can be ignored.

[0118] Depend on Figure 29 It can be seen that when the mineralization changes by 5%, that is, from 20 g / L to 21 g / L, the measured phase characteristic curve changes little in the low frequency range and does not have obvious distinguishability. This indicates that in the low frequency range, the sensor is not sensitive to the phase characteristics when the mineralization is 20 g / L and the mineralization changes by 5%.

[0119] Depend on Figures 30-31 It can be seen that when the mineralization is 30 g / L and 31.5 g / L, the phase characteristic curves of the sensor almost overlap under static, dynamic and different gas-liquid ratios within the measurement range. This indicates that the relationship between the phase characteristic and mineralization is only slightly affected by the water flow rate and different gas-liquid ratios within this frequency range, and can be ignored.

[0120] Depend on Figure 32 It can be seen that when the mineralization changes by 5%, that is, from 30 g / L to 31.5 g / L, the measured phase characteristic curve changes little in the low frequency range and does not have obvious distinguishability. This indicates that in the low frequency range, the sensor is not sensitive to the phase characteristics when the mineralization is 30 g / L and the mineralization changes by 5%.

[0121] Depend on Figures 33-34 It can be seen that when the mineralization is 40 g / L and 42 g / L, the phase characteristic curves of the sensor almost overlap under static, dynamic and different gas-liquid ratio conditions within the measurement range. This indicates that the relationship between the phase characteristic and mineralization is only slightly affected by the water flow rate and different gas-liquid ratios within this frequency range, and can be ignored.

[0122] Depend on Figure 35 It can be seen that when the mineralization changes by 5%, that is, from 40 g / L to 42 g / L, the measured phase characteristic curve changes little in the low frequency range and does not have obvious distinguishability. This indicates that in the low frequency range, the sensor is not sensitive to the phase characteristics when the mineralization is 40 g / L and the mineralization changes by 5%.

[0123] Based on the above observations and analysis, it can be concluded that microwave sensors... Compared to the phase characteristics, both the amplitude and phase characteristics exhibit monotonicity with respect to salinity. At different salinities, the amplitude characteristics are more sensitive than the phase characteristics and are unaffected by different gas-liquid ratios; therefore, the amplitude is determined to be the measured physical quantity of the sensor. The correlation coefficient, data range, and standard deviation of the amplitude characteristics at each frequency were calculated using the above methods, resulting in Table 1. Based on the above screening rules, the microwave sensor frequency is determined to be 55.267MHz.

[0124] Table 1. Amplitude characteristics of the microwave sensor between 52.7685MHz and 67.7595MHz Group Correlation coefficient Data range Standard deviation Frequency Frequency (MHz) (MHz) Group 1 0.9925 3.543 1.211 52.7685 Group 2 0.9839 3.639 1.256 55.267 Group 3 0.9827 3.886 1.295 57.7655 Group 4 0.9801 3.729 1.24 60.264 Group 5 0.9774 3.575 1.183 62.7625 Group 6 0.9595 3.509 1.13 65.261 Group 7 0.9101 3.5 1.121 67.7595 Example 4 like Figure 5 As shown, the mineralization measurement method includes: using the amplitude characteristics and frequency points of the microwave sensor described in Example 3 above. Specifically, firstly, the assembly and debugging of the measurement system are completed to ensure that each component works normally and the parameters are matched. Using the amplitude characteristics of the microwave sensor described in Example 4 above and the selected optimal frequency point, the frequency point parameter is preset into the control module of the microwave sensor, and the working mode of the sensor is set to continuous measurement mode to ensure that the sensor can operate stably at the optimal frequency point after startup, avoiding measurement errors caused by frequency deviation. At the same time, the connection status of the transmitting unit 1, the detection unit 2, and the antenna unit 3 of the microwave sensor is checked to ensure that the lines are unobstructed and that the transmitting and receiving sensitivity of the antenna unit 3 meets the experimental requirements; the detection system is debugged to enable it to normally receive and process the amplitude data transmitted by the detection unit 2, and to call the optimal vector machine regression fitting model described in Example 3 to prepare for the subsequent mineralization back-calculation.

[0125] The measurement method also includes immersing the antenna element 3 of the microwave sensor into the water to be tested. Specifically, the antenna element 3 of the microwave sensor is smoothly immersed into the water to be tested, ensuring that both the transmitting and receiving ends of the antenna element 3 are completely submerged in the water. If the water to be tested is the water in the peristaltic pump system pipeline, the antenna element 3 needs to be fixed at the preset detection position in the pipeline, ensuring that the antenna is not in contact with or obstructed by the inner wall of the pipeline, and is compatible with the operating conditions of the peristaltic pump to avoid antenna displacement caused by vibration generated by the peristaltic pump, which would affect the measurement stability. After deployment, allow it to stand for 3-5 minutes to allow the antenna element 3 and the water to be tested to reach temperature equilibrium, reducing the impact of temperature differences on microwave signal transmission.

[0126] The microwave sensor is activated and operates continuously at a predetermined frequency. It generates a microwave signal within a preset range via transmitting unit 1, which is then transmitted to the water sample to be tested via antenna unit 3. Antenna unit 3 simultaneously receives the reflected microwave signal and transmits it to detection unit 2. Detection unit 2 converts the reflected signal into corresponding amplitude data and transmits it to the detection system in real time. Specifically, the microwave sensor is activated and operates stably at a preset optimal frequency. Transmitting unit 1 generates a microwave signal within a preset range according to preset parameters. This microwave signal is transmitted to antenna unit 3 via a transmission line and then directionally transmitted to the water sample to be tested. As the microwave signal propagates in the water sample, it interacts with mineral ions in the water, producing physical phenomena such as reflection and refraction. Antenna unit 3 simultaneously receives the reflected microwave signal and quickly transmits it to detection unit 2. The detection unit 2 has a built-in signal conversion module that converts the received microwave reflection signal into corresponding digital amplitude data. At the same time, it performs preliminary filtering on the converted amplitude data to remove slight noise generated during signal transmission. Then, it transmits the processed amplitude data to the detection system in real time and continuously to ensure the timeliness and integrity of data transmission.

[0127] The detection system performs outlier removal and normalization preprocessing on the received amplitude data. Substituting this data into the optimal vector machine regression fitting model described in Example 4, the system uses the nonlinear correlation between amplitude characteristics and salinity in the optimal vector machine regression fitting model to calculate the salinity value of the water to be tested. Specifically, after receiving the amplitude data, the detection system first performs the same preprocessing operation as during the training of the vector machine regression fitting model in Example 4 to ensure that the data format, accuracy, and model compatibility are maintained. This includes two steps: first, outlier removal, using... The criteria involve two steps: first, removing outliers from the amplitude data; second, data normalization, which maps the effective amplitude data after outlier removal to... Within the interval, the influence of dimensions is eliminated to ensure consistency with the training data preprocessing standards of the optimal vector machine regression fitting model.

[0128] After preprocessing, the effective amplitude data is substituted into the optimal vector machine regression fitting model described in Example 4. This model has been trained and verified through dynamic experiments and anti-interference experiments, and can accurately capture the nonlinear correlation between amplitude characteristics and mineralization. The nonlinear mapping of the model is used for back-calculation, i.e., the externally processed amplitude data is used as the output variable to solve for the corresponding mineralization independent variable value. Simultaneously, the error range of this mineralization value is calculated to ensure the reliability of the calculation results. The detection system outputs the mineralization value of the water quality to be tested in real time, which can be displayed intuitively through the display module or stored in the terminal device through the data transmission module for subsequent data retrieval and analysis. If higher measurement accuracy is required, parallel measurement verification can be performed: repeat the above steps at different locations and time points of the same water quality to be tested, performing 3-5 parallel measurements, and calculating the average value of the multiple measurement results as the final mineralization measurement value to further reduce random errors. Meanwhile, the measurement result can be compared and verified with a standard salt solution with known mineralization. If the deviation between the measured value and the standard value is preset, the measurement result is confirmed to be qualified. If the deviation exceeds the allowable range, the sensor frequency parameters, antenna deployment status, or model call status need to be checked, and the measurement should be repeated after adjustment.

[0129] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A microwave sensor, comprising: Transmitting unit (1), which is used to generate microwave signals within a preset range; The detection unit (2) is used to convert the received microwave signal into an electrical signal; The sensor is characterized in that it further includes two antenna units (3) for transmitting microwave signals and receiving reflected signals; one end of each antenna unit (3) is connected to the corresponding transmitting unit (1) and detection unit (2), and the other end extends vertically into the water to be detected; The antenna unit (3) is an enameled wire. The use of enameled wire in the antenna unit (3) is to prevent its end face from coming into contact with salt water and causing a chemical reaction that would cause corrosion. The frequency of the sensor was determined through dynamic experiments and anti-interference experiments.

2. A mineralization measurement system, characterized in that, This mineralization measurement system is used in dynamic experiments and anti-interference experiments. The system includes: Pipe (4); At least two peristaltic pumps (5) are connected to the pipe (4); one of the peristaltic pumps (5) pumps brine solutions of different mineralization into the pipe; the other peristaltic pump (5) pumps air into the pipe; The microwave sensor of claim 1 is installed on the side wall of the pipe (4); the microwave sensor transmits the collected data to the preset detection system.

3. A method for determining the physical quantity and frequency point to be measured by a microwave sensor, characterized in that, This method is applicable to the mineralization measurement system of claim 2, and the method includes: The mineralization measurement system introduces salt water solutions or air with different mineralization and preset mineralization variations into the pipeline; When conducting dynamic experiments, a peristaltic pump in the mineralization measurement system is turned on to continuously pump salt water solutions with different mineralization into the pipeline, and the amplitude and phase data under different mineralization and preset mineralization changes are recorded by a microwave sensor. During the anti-interference experiment, one peristaltic pump in the mineralization measurement system continuously pumps salt water solutions of different mineralization into the pipeline, while another peristaltic pump continuously pumps air into the pipeline. By controlling the rotation speed of the two peristaltic pumps, different gas-liquid ratios are controlled, and amplitude and phase data under different mineralization and preset mineralization changes are recorded by a microwave sensor. The amplitude and phase data obtained from dynamic experiments and anti-interference experiments under different mineralization degrees and preset mineralization degree changes were respectively subjected to curve fitting by vector machine regression fitting strategy. This yielded amplitude characteristic sweep curves and phase characteristic sweep curves under static conditions and different gas-liquid ratios, as well as comparison charts of amplitude characteristic sweep curves and phase characteristic sweep curves under preset mineralization degree changes. The amplitude and phase sweep frequency curves under static conditions and different gas-liquid ratios with different mineralization degrees, as well as the amplitude and phase sweep frequency curves under preset mineralization degree variations, were observed and analyzed. By observing the monotonicity between amplitude and phase characteristics and mineralization degree, and comparing the sensitivity of amplitude characteristics to phase characteristics under different mineralization degrees, the measured physical quantity of the sensor was determined to be amplitude characteristics. By calculating and comparing the amplitude characteristic correlation coefficient, data range, and standard deviation under different mineralization and gas-liquid ratios, the frequency corresponding to the amplitude correlation coefficient closest to 1, the largest data range, and the smallest standard deviation is taken as the frequency point.

4. The method for determining the physical quantity and frequency point of a microwave sensor according to claim 3, characterized in that, The vector machine regression fitting strategy includes: The raw data collected in the dynamic experiment and anti-interference experiment were preprocessed to remove outliers in the discrete amplitude data and phase data; the effective amplitude data and phase data after removing outliers were normalized to map the amplitude data, phase data, mineralization parameter, gas-liquid ratio parameter and frequency parameter to the [0,1] interval. The radial basis function kernel is defined as the kernel function of the vector machine regression model. It is used to fit the effective mineralization, gas-liquid ratio, frequency and amplitude, and phase data into a nonlinear relationship, and introduces the penalty coefficient and insensitive loss function parameters. By combining grid search and five-fold cross-validation, the optimal penalty coefficient, kernel function parameters, and insensitive loss function parameters are obtained, and the optimal vector machine regression fitting model is constructed. Using the mineralization degree, preset mineralization degree change, and frequency measured in the dynamic experiment as input variables, and the corresponding amplitude data and phase data as output variables, the data are substituted into the optimal vector machine regression fitting model. Through the nonlinear mapping capability of the model, the discrete amplitude and phase data are fitted into continuous curves, resulting in amplitude characteristic sweep curves and phase characteristic sweep curves under different mineralization degrees, as well as comparative data of amplitude characteristic sweep curves and phase characteristic sweep curves under preset mineralization degree changes. Using the mineralization degree, preset mineralization degree change, gas-liquid ratio, and frequency measured in the anti-interference experiment as input variables, and the corresponding amplitude data and phase data as output variables, the discrete data are continuously fitted into the same optimal vector machine regression fitting model to obtain amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under different mineralization degrees and different gas-liquid ratios, as well as comparative data of amplitude characteristic sweep frequency curves and phase characteristic sweep frequency curves under preset mineralization degree changes.

5. The method for determining the physical quantity and frequency point of a microwave sensor according to claim 4, characterized in that, The preprocessed effective amplitude and phase data are divided into training and test sets according to a preset ratio. The training set is used to build a vector machine regression fitting model, and the test set is used to verify the model's fitting accuracy and generalization ability.

6. The method for determining the physical quantity and frequency point of a microwave sensor according to claim 5, characterized in that, Substitute the divided test set data into the fitting model, calculate the fitting error and coefficient of determination of the test set data, and verify the fitting results; When the coefficient of determination is greater than the preset threshold and the fitting error is less than the preset range, the accuracy of the fitting model meets the experimental requirements and the fitting curve is correct. If the coefficient of determination is less than the preset threshold and the fitting error is greater than the preset range, then adjust the values ​​of the penalty coefficient, kernel function parameter and insensitive loss function parameter, and perform fitting and verification again until the coefficient of determination is greater than the preset threshold and the fitting error is less than the preset range.

7. The method for determining the physical quantity and frequency point of a microwave sensor according to claim 6, characterized in that, If the fitted curve fails to capture the data variation pattern, increase the value of the penalty coefficient; if the training set fits successfully but the test set has a large deviation, decrease the value of the penalty coefficient. If the fitted curve fails to capture the complex relationship between mineralization, gas-liquid ratio, amplitude, or phase, gradually increase the value of the kernel function parameter; if the fitted curve fluctuates greatly and has poor fit, decrease the value of the kernel function parameter. If the actual fitting deviation frequently exceeds the limit, increase the value of the insensitive loss function parameter; if the fitting deviation is large in some high-frequency data, decrease the value of the insensitive loss function parameter.

8. The method for determining the physical quantity and frequency point to be measured by a microwave sensor according to claim 7, characterized in that, Each time, only one of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter is adjusted, while the values ​​of the other two parameters are fixed. The optimal value of the parameter is selected by five-fold cross-validation. After adjusting the values ​​of the three parameters in sequence, the values ​​of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter are fine-tuned as a whole to obtain the optimal values ​​of the penalty coefficient parameter, kernel function parameter, and insensitive loss function parameter.

9. A method for measuring mineralization, characterized in that, The measurement method includes: The amplitude characteristics and frequency points of the microwave sensor are determined using the method for determining the physical quantity and frequency point of the microwave sensor as described in claim 8. The antenna unit (3) of the microwave sensor is introduced into the water to be tested; The microwave sensor is activated and operates continuously at a determined frequency. It generates a microwave signal within a preset range through the transmitting unit (1) and transmits it to the water quality to be tested through the antenna unit (3). The antenna unit (3) synchronously receives the microwave reflection signal and transmits it to the detection unit (2). The detection unit (2) converts the reflection signal into the corresponding amplitude data and transmits it to the detection system in real time. The detection system performs outlier removal and normalization preprocessing on the received amplitude data, and substitutes it into the optimal vector machine regression fitting model in the method for determining the physical quantity and frequency point of microwave sensor as described in claim 8. Through the nonlinear correlation between amplitude characteristics and mineralization in the optimal vector machine regression fitting model, the mineralization value of the water quality to be detected is calculated in reverse.

Images