An on-line chromatographic monitoring system for a tailings thickener

Abstract

The application discloses an online tomography monitoring system for a tailings thickener, which realizes online monitoring of slurry in the thickener by installing an ultrasonic array and a data acquisition system provided by the application on the inner wall of the thickener; the simultaneous transmission and reception of the ultrasonic sensor array are realized through the cooperation of a signal filter and a multiplexer; the marking and differentiation of the pulse tomography path and the multi-frequency source tomography are realized by adopting a wideband ultrasonic transmitting probe and a receiving probe; after the acoustic characteristics of the slurry are calibrated, the multi-frequency source slurry tomography data are collected; a slurry background sound velocity model is established based on low-frequency tomography data, and a slurry disturbance sound velocity model is established based on high-frequency tomography data; the background sound velocity model and the disturbance sound velocity model of different frequency bands are trained to establish a reliable slurry tomography image resolution improvement model, and finally, a slurry super-resolution sound velocity model is obtained, thereby realizing fine monitoring of the internal running state and slurry thickening condition of the thickener.

Description

Technical Field

[0001] This invention relates to the field of measurement technology for the thickening effect of tailings slurry in mines, and in particular to an online chromatography monitoring system for tailings thickeners. Background Technology

[0002] A tailings thickener is a piece of equipment that uses gravity settling to separate solids and liquids and increase the concentration of tailings slurry. It is currently widely used in surface stockpiling and underground backfilling. The slurry concentration and stratification height inside the thickener are key parameters affecting its operation and related production.

[0003] Although the sedimentation principle and efficiency of thickeners have greatly improved since the early days of ordinary thickeners to today's deep cone thickeners with added flocculants, the density of the tailings slurry inside the thickener is still not visible. Existing measurement methods cannot achieve accurate online monitoring, making it difficult to accurately optimize and control the thickener's operation and hindering the advancement of related production processes. Therefore, there is an urgent need for a system capable of real-time online monitoring of the thickener's internal operating status.

[0004] As a typical non-destructive testing technology, ultrasound can perform tomographic measurements on various media such as solids, liquids, and gases. However, existing electronic array sensors are costly and difficult to apply to engineering equipment such as tailings thickeners, while electronic linear array sensors can only perform sequential excitation and cannot efficiently perform tomographic scanning on large equipment like thickeners. Therefore, an online tomographic monitoring system for tailings thickeners is needed to solve these problems. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an online chromatography monitoring system for tailings thickeners, which can accurately measure and analyze the concentration distribution and stratification within the thickener in real time.

[0006] The solution of the present invention is:

[0007] An online tomographic monitoring system for tailings thickeners includes hardware and a computer device with running software. The hardware refers to an integrated ultrasonic sensor array and a data acquisition system, while the software running on the computer device refers to a matching data processing and tomographic imaging algorithm. The hardware uses a wideband ultrasonic transmitter, receiver, and multiplexing technology to form a sensor array, enabling rapid tomographic analysis of the thickener slurry. The software includes steps for measuring the acoustic characteristics of the slurry, acquiring multi-frequency source tomographic signals, performing multi-frequency source mixed-domain full waveform inversion imaging, and performing tomographic image super-resolution reconstruction based on deep learning, thereby enabling real-time monitoring of the thickener's operating status.

[0008] As a preferred technical solution, the integrated ultrasonic sensor array includes a transmitting transducer array, a receiving transducer array, a signal filter, a multiplexer, signal lines, power lines, and a DC power supply; the transmitting transducer array contains several ultrasonic transmitting transducers, which are integrated into the signal filter via signal lines; the receiving transducer array contains several receiving transducers, which are integrated into the multiplexer via signal lines; both the ultrasonic transmitting transducers and receiving transducers are connected in parallel to the DC power supply via power lines.

[0009] As a preferred technical solution, the ultrasonic receiving transducers all employ broadband ultrasonic sensors; the ultrasonic transmitting transducers are controlled by a computer to excite pulses of different frequencies to achieve signal marking and avoid mutual interference; by changing the excitation frequency of the ultrasonic transducers and performing multiple scans, multi-frequency ultrasonic tomography is achieved; the ultrasonic receiving transducers all employ broadband ultrasonic sensors to achieve response to ultrasonic pulses of different frequencies.

[0010] As a preferred technical solution, the data acquisition system includes a signal filter, a multiplexer, signal lines, a computer, and a memory; the signal filter of the data acquisition system is connected to the multiplexer of the receiving transducer array via signal lines; the multiplexer of the data acquisition system is connected to the signal filter of the transmitting transducer array via signal lines.

[0011] As a preferred technical solution, the data processing and tomographic imaging algorithm is as follows:

[0012] S1, Measurement of acoustic properties of slurry; by establishing an experimental device, the propagation speed of ultrasonic waves of different frequencies in slurries of different concentrations is measured, and a standard characteristic database is established.

[0013] S2, acquire multi-frequency ultrasonic tomography data; by changing the ultrasonic pulse frequency, obtain the basic data of signal waveform, delay, and amplitude for inversion imaging;

[0014] S3, Multi-frequency source hybrid domain full waveform inversion imaging; Employing a frequency source hybrid domain full waveform inversion algorithm, high-resolution, high signal-to-noise ratio tomographic images based on different frequencies are obtained;

[0015] S4, super-resolution image deep learning; uses convolutional neural networks to process tomographic images at different time series to obtain super-resolution tomographic images, namely slurry super-resolution sound velocity model.

[0016] As a preferred technical solution, the standard characteristic database established in S1 is as follows:

[0017] S101, Construct an experimental device for measuring the mapping relationship between ultrasonic wave velocity and slurry concentration. The experimental device includes a slurry container, an ultrasonic controller, a timer, a transmitting probe, and a receiving probe.

[0018] S102, inject a slurry of known concentration into the slurry container; the ultrasonic controller generates a signal of a specific frequency and transmits it to the transmitting probe to be converted into ultrasonic pulses, and at the same time starts the timer; the ultrasonic waves are sensed by the receiving probe after passing through the slurry medium, and at the same time the timer is turned off.

[0019] S103, the acoustic properties of the slurry are determined by the constant density acoustic wave equation. The time domain and frequency domain forms of the constant density acoustic wave equation are as follows:

[0020]

[0021]

[0022] Where v(x) is the speed of sound, ρ(x) is the density, t is time, ω is the angular frequency, and p(x,t;x) is the velocity of sound. t f is the wave field pressure value. t (x,t;x t The epicenter was ( ).

[0023] S104, Change the signal frequency generated by the ultrasonic controller to measure the propagation speed of ultrasonic waves of different frequencies in the slurry of the current concentration.

[0024] S105, change the slurry with different concentrations, and repeat steps S102, S103, and S104 until the required test results are obtained. Establish a standard database of slurry acoustic properties and determine the correlation between sound velocity, density, time difference, and frequency.

[0025] As a preferred technical solution, the multi-frequency ultrasonic tomography data collected in S2 is as follows:

[0026] S201, the computer of the data acquisition system encodes a set of control signals, which are sent to the ultrasonic transmitting array by the multiplexer, and the corresponding excitation waveform, frequency, time and amplitude data are recorded;

[0027] S202, the signal filter of the ultrasonic transmitting array receives the control signal, filters it into multiple control signals, and sends them to the corresponding ultrasonic transmitters;

[0028] S203, the ultrasonic transmitter excites ultrasonic pulses of a corresponding frequency, the pulses pass through the slurry and are transmitted to the ultrasonic receiver;

[0029] S204, the ultrasonic receiver converts the received pulse signals of different frequencies into multiple digital signals, including waveform, frequency, arrival time and amplitude data, and transmits them to the multiplexer;

[0030] S205, the multiplexer combines multiple digital signals into one wideband digital signal and transmits it to the signal filter of the data acquisition system.

[0031] S206, the signal filter of the data acquisition system filters the broadband digital signal into multiple digital signals and transmits them to the computer for storage of the corresponding data;

[0032] S207, change the excitation frequency scheme of the ultrasonic transmitter, and repeat steps S201, S202, S203, S204, S205, and S206 to perform multiple excitations, scans, and recordings to obtain multi-frequency source ultrasonic tomography data.

[0033] As a preferred technical solution, the multi-frequency source hybrid domain full waveform inversion algorithm in S3 is as follows:

[0034] S301, Ultrasonic Time Domain Inversion; First, using data such as arrival time difference and amplitude from low-frequency ultrasonic tomography, the time domain full waveform inversion of the slurry's sound velocity model is performed; through iterative optimization, a high signal-to-noise ratio background sound velocity model based on different low frequencies is established.

[0035] The objective function for time-domain full waveform inversion is:

[0036]

[0037] Where m represents the velocity model parameters, T represents the total duration, and p1 and p2 represent the calculated and observed wavefields, respectively; the model update gradient is:

[0038]

[0039]

[0040]

[0041]

[0042] In the formula: E is the velocity model, m is the velocity model parameter, v(x) is the speed of sound, t is time, and p1, p2 and p3 are the calculated wave field, the observed wave field and the accompanying wave field, respectively.

[0043] Choose an appropriate iteration step size, repeatedly calculate the gradient, and update the velocity model to minimize the objective function;

[0044] S302, Ultrasonic Frequency Domain Inversion: Through Fast Fourier Transform, the ultrasonic time-domain signal is converted into a frequency-domain signal, and then the frequency domain full waveform inversion is performed using the frequency characteristics of high-frequency ultrasonic tomography; through iterative optimization, a high-resolution perturbation velocity model based on different high frequencies is established.

[0045] The objective function for frequency domain full waveform inversion is:

[0046]

[0047] The model update gradient is:

[0048]

[0049]

[0050]

[0051]

[0052] In the formula: E is the velocity model, m is the velocity model parameter, v(x) is the speed of sound, ω is the angular frequency, t is time, and p1, p2 and p3 are the calculated wave field, the observed wave field and the accompanying wave field, respectively.

[0053] As a preferred technical solution, the super-resolution image deep learning in S4 is as follows:

[0054] S401, Model Training: First, a sound velocity model resolution enhancement model is trained and established using a convolutional neural network with a portion of the low-frequency background sound velocity model as input and a portion of the high-frequency perturbation sound velocity model as output, to obtain a super-resolution reconstructed image.

[0055] S402, Model Evaluation and Validation; The quality of the model output is evaluated by calculating the peak signal-to-noise ratio (PSNR) and structural similarity of the reconstructed image. The higher the PSNR and structural similarity, the better the image quality. The parameters are adjusted to obtain a model that meets the requirements.

[0056] The formula for calculating peak signal-to-noise ratio is as follows:

[0057]

[0058]

[0059] Where MSE is the mean square error between the reconstructed image SR(i,j) and the low-resolution image LR(i,j), H and W are the height and width of the image, respectively, n is the image bit rate, and R is the peak signal-to-noise ratio in dB;

[0060] The formula for calculating structural similarity is as follows:

[0061]

[0062] Where S is the structural similarity, μ HR ,μ SR The average gray values ​​of the high-resolution image HR and the reconstructed image SR are respectively, σ HR ,σ SR These are the corresponding image variances, σ HR,SR Let C1 and C2 be the covariance of the two images, and C1 and C2 be constants.

[0063] S403, Model Testing; The remaining background model and perturbation model are used as the test set to ensure that the established model can run reliably and obtain high-quality reconstructed images;

[0064] S404 Finally, the background model and disturbance model obtained by inversion are input together to establish a super-resolution model for in-thickness slurry chromatography.

[0065] An online tomographic monitoring system for tailings thickeners, employing the aforementioned technical solution, comprises a computer device with hardware and operating software. The hardware refers to an integrated ultrasonic sensor array and data acquisition system, while the software running on the computer device refers to a matching data processing and tomographic imaging algorithm. The hardware uses a wideband ultrasonic transmitter, receiver, and multiplexing technology to form a sensor array, enabling rapid tomographic analysis of the thickener slurry. The software includes steps such as measuring the acoustic characteristics of the slurry, acquiring multi-frequency source tomographic signals, performing multi-frequency source mixed-domain full-waveform inversion imaging, and performing deep learning-based tomographic image super-resolution reconstruction, thereby achieving real-time monitoring of the thickener's operating status.

[0066] Advantages of this invention:

[0067] Therefore, this invention provides an ultrasonic sensor array and data acquisition system for tailings thickeners, and provides supporting data analysis and tomographic imaging methods. The technology is reasonable and feasible, and has broad application prospects in both research and production.

[0068] (1) According to the system provided by the present invention, it can not only determine the concentration and stratification of the slurry in the thickener to meet the basic detection needs, but also realize real-time online monitoring and visualization of the slurry compression in the thickener. By accurately grasping the slurry concentration and sedimentation, it is beneficial to optimize and control its operating status, and improve production efficiency and process level.

[0069] (2) The integrated ultrasonic sensor array provided by the present invention uses ultrasonic transmitting transducers of a specific frequency and wideband ultrasonic receiving transducers. Multiple ultrasonic transducers are integrated through multiplexers and signal filters, which can avoid ultrasonic interference at the same frequency, reduce AC noise, and realize the simultaneous transmission and reception of multiple transducers, significantly improving the speed of tomography measurement.

[0070] (3) The present invention adopts an online installation design, which can realize remote detection and real-time analysis of the slurry concentration in the thickener, and can monitor and record the operating status of the thickener for a long time to ensure its efficient and reliable operation. In addition, the present invention is based on the ultrasonic measurement principle, which has low measurement cost, high accuracy, and is harmless to the human body and environmentally friendly. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of an integrated ultrasonic array sensor and data acquisition system.

[0072] Figure 2 This is a basic flowchart of data processing and tomographic imaging algorithms;

[0073] Figure 3 A schematic diagram of the experimental setup required to establish a database of slurry acoustic properties;

[0074] Figure 4 The background sound velocity model of the slurry obtained by time-domain inversion;

[0075] Figure 5 The slurry disturbance sound velocity model obtained by frequency domain inversion;

[0076] Figure 6 A super-resolution sound velocity model of slurry obtained through deep learning;

[0077] Among them, 1-ultrasonic transmitting array; 2-ultrasonic receiving array; 3-signal filter; 4-multiplexer; 5-signal line; 6-power line; 7-DC power supply; 8-memory; 9-computer; 101-transmitter; 201-receiver; 10-slurry container; 11-ultrasonic controller; 12-counter. Detailed Implementation

[0078] This invention provides an online chromatography monitoring system for tailings thickeners.

[0079] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to specific embodiments.

[0080] Example

[0081] The hardware refers to an integrated ultrasonic sensor array and data acquisition system, while the "part" refers to the supporting data processing and tomographic imaging algorithms. The hardware employs a wideband ultrasonic transmitter and receiver, along with multiplexing technology, to form a sensor array that enables rapid tomographic analysis of thickener slurry. The software includes steps such as measuring the acoustic properties of the slurry, acquiring multi-frequency source tomographic signals, performing multi-frequency source mixed-domain full-waveform inversion imaging, and deep learning-based tomographic image super-resolution reconstruction, enabling real-time monitoring of the thickener's operating status.

[0082] The integrated ultrasonic sensor array includes: a transmitting transducer array 1, a receiving transducer array 2, a signal filter 3, a multiplexer 4, a signal line 5, a power line 6, and a DC power supply 7; the transmitting transducer array 1 includes a plurality of ultrasonic transmitting transducers 101, which are integrated into the signal filter 3 via the signal line 5; the receiving transducer array 2 includes a plurality of receiving transducers 201, which are integrated into the multiplexer 4 via the signal line; the ultrasonic transmitting transducers 101 and the receiving transducers 201 are both connected in parallel to the DC power supply 7 via the power line 6.

[0083] The ultrasonic receiving transducers 101 all employ wideband ultrasonic sensors; the ultrasonic transmitting transducers are controlled by a computer to excite pulses of different frequencies to achieve signal marking and avoid mutual interference; by changing the excitation frequency of the ultrasonic transducers and performing multiple scans, multi-frequency ultrasonic tomography is achieved; the ultrasonic receiving transducers 201 all employ wideband ultrasonic sensors to achieve response to ultrasonic pulses of different frequencies.

[0084] The data acquisition system includes: a signal filter 3, a multiplexer 4, a signal line 5, a computer 8, and a memory 9; the signal filter 3 of the data acquisition system is connected to the multiplexer 4 of the receiving transducer array via the signal line 5; the multiplexer 4 of the data acquisition system is connected to the signal filter 3 of the transmitting transducer array via the signal line 5.

[0085] Data processing and tomographic imaging algorithms include the following steps:

[0086] S1, Measurement of acoustic properties of slurry. An experimental setup was established to measure the propagation speed of ultrasonic waves of different frequencies in slurries of different concentrations, and a standard characteristic database was created.

[0087] S2, acquires multi-frequency ultrasound tomography data. By changing the ultrasound pulse frequency, basic data such as signal waveform, delay, and amplitude are obtained for inversion imaging;

[0088] S3, Multi-frequency source hybrid domain full waveform inversion imaging. Using the algorithm provided in this invention, high-resolution, high signal-to-noise ratio tomographic images based on different frequencies are obtained;

[0089] S4, Super-resolution Image Deep Learning. A convolutional neural network is used to process tomographic images at different time points to obtain super-resolution tomographic images, i.e., a slurry super-resolution sound velocity model.

[0090] The specific steps for establishing the standard characteristic database in step S1 are as follows:

[0091] S101, Construct an experimental device for measuring the mapping relationship between ultrasonic wave velocity and slurry concentration. The experimental device includes a slurry container 10, an ultrasonic controller 11, a timer 12, a transmitting probe 13, and a receiving probe 14.

[0092] S102, a slurry of known density is injected into the slurry container 10; the ultrasonic controller generates a signal of a specific frequency and transmits it to the transmitting probe 13 to be converted into an ultrasonic pulse, and at the same time the timer 12 is turned on; the ultrasonic wave is sensed by the receiving probe 14 after passing through the slurry medium, and at the same time the timer 12 is turned off.

[0093] S103, the acoustic properties of the slurry are determined by the constant density acoustic wave equation. The time domain and frequency domain forms of the constant density acoustic wave equation are as follows:

[0094]

[0095]

[0096] Where v(x) is the speed of sound, ρ(x) is the density, t is time, ω is the angular frequency, and p(x,t;x) is the velocity of sound. t f is the wave field pressure value. t (x,t;x t The epicenter was ( ).

[0097] S104, Change the signal frequency generated by the ultrasonic controller to measure the propagation speed of ultrasonic waves of different frequencies in the slurry of the current concentration.

[0098] S105, change to slurry of different densities, repeat steps S102 to S104 until the required test results are obtained, establish a standard database of slurry acoustic properties, and determine the correlation between sound velocity, density and frequency;

[0099] The specific steps for acquiring multi-frequency ultrasound tomography data in step S2 are as follows:

[0100] S201, the computer of the data acquisition system encodes a set of control signals, such as a pulse excitation frequency of 10KHz to 20KHz; the signals are sent to the ultrasonic transmitting array by a multiplexer, and the corresponding excitation waveform, frequency, time, amplitude and other data are recorded.

[0101] S202, the signal filter of the ultrasonic transmitting array receives the control signal, filters it into multiple control signals, and sends them to the corresponding ultrasonic transmitters;

[0102] S203, the ultrasonic transmitter excites ultrasonic pulses of a corresponding frequency, the pulses pass through the slurry and are transmitted to the ultrasonic receiver;

[0103] S204, the ultrasonic receiver converts the received pulse signals of different frequencies into multiple digital signals, including data such as waveform, frequency, arrival time, and amplitude, and transmits them to the multiplexer;

[0104] S205, the multiplexer combines multiple digital signals into one wideband digital signal and transmits it to the signal filter of the data acquisition system.

[0105] S206, the signal filter of the data acquisition system filters the broadband digital signal into multiple digital signals and transmits them to the computer for storage of the corresponding data;

[0106] S207, change the excitation frequency scheme of the ultrasonic transmitter, for example, each excitation scheme step size is 10KHz, and sequentially complete the pulse excitation, scanning and recording of 20KHz~30KHz, 30KHz~40KHz, 40KHz~50KHz, 50KHz~60KHz, 60KHz~70KHz, 70KHz~80KHz, 80KHz~90KHz, and 90KHz~100KHz to obtain multi-frequency source ultrasonic tomography data.

[0107] The specific implementation steps of the multi-frequency source hybrid domain full waveform inversion algorithm in step S3 are as follows:

[0108] S301, Ultrasonic Time-Domain Inversion. First, using data such as arrival time difference and amplitude from low-frequency ultrasonic tomography, the sound velocity model of the slurry is inverted in the time domain as a whole waveform. Through iterative optimization, a high signal-to-noise ratio background sound velocity model based on different low frequencies is established.

[0109] The objective function for time-domain full waveform inversion is:

[0110]

[0111] Where m is the velocity model parameter, T is the total duration, and p1 and p2 are the calculated wavefield and the observed wavefield, respectively;

[0112] The model update gradient is:

[0113]

[0114]

[0115]

[0116]

[0117] In the formula: E is the velocity model, m is the velocity model parameter, v(x) is the speed of sound, t is time, and p1, p2 and p3 are the calculated wave field, the observed wave field and the accompanying wave field, respectively.

[0118] Choose an appropriate iteration step size, repeatedly calculate the gradient, and update the velocity model to minimize the objective function.

[0119] S302, Ultrasonic Frequency Domain Inversion. The ultrasonic time-domain signal is converted into a frequency-domain signal using a Fast Fourier Transform, and then full-waveform inversion in the frequency domain is performed based on the frequency characteristics of high-frequency ultrasonic tomography. Through iterative optimization, a high-resolution perturbation velocity model based on different high frequencies is established.

[0120] The objective function for frequency domain full waveform inversion is:

[0121]

[0122] The model update gradient is:

[0123]

[0124]

[0125]

[0126]

[0127] In the formula: E is the velocity model, m is the velocity model parameter, v(x) is the speed of sound, ω is the angular frequency, t is time, and p1, p2 and p3 are the calculated wave field, the observed wave field and the accompanying wave field, respectively.

[0128] The super-resolution image deep learning steps in step S4 are as follows:

[0129] S401, Model Training: First, a sound velocity model resolution enhancement model is trained and established using a convolutional neural network with a portion of the low-frequency background sound velocity model as input and a portion of the high-frequency perturbation sound velocity model as output, to obtain a super-resolution reconstructed image.

[0130] S402, Model Evaluation and Validation; The quality of the model output is evaluated by calculating the peak signal-to-noise ratio (PSNR) and structural similarity of the reconstructed image. The higher the PSNR and structural similarity, the better the image quality. The parameters are adjusted to obtain a model that meets the requirements.

[0131] The formula for calculating peak signal-to-noise ratio is as follows:

[0132]

[0133]

[0134] Where MSE is the mean square error between the reconstructed image SR(i,j) and the low-resolution image LR(i,j), H and W are the height and width of the image, respectively, n is the image bit rate, and R is the peak signal-to-noise ratio in dB.

[0135] The formula for calculating structural similarity is as follows:

[0136]

[0137] Where S is the structural similarity, μ HR ,μ SR The average gray values ​​of the high-resolution image HR and the reconstructed image SR are respectively, σ HR ,σ SR These are the corresponding image variances, σ HR,SR Let C1 and C2 be the covariance of the two images, and C1 and C2 are constants.

[0138] For example, the output of the established model is required to have a peak signal-to-noise ratio greater than 100dB and a structural similarity greater than 0.9.

[0139] S403, Model Testing; The remaining background and perturbation models are used as the test set to ensure that the established model can run reliably and obtain high-quality reconstructed images;

[0140] S404. Finally, the background model and disturbance model obtained by inversion are input together to establish a super-resolution model for in-thickness slurry chromatography.

[0141] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An online chromatography monitoring system for tailings thickeners, characterized in that: The system includes a computer device with hardware and running software. The hardware refers to an integrated ultrasonic sensor array and data acquisition system, while the software running on the computer device refers to the supporting data processing and tomographic imaging algorithms. The hardware uses a wideband ultrasonic transmitter, receiver, and multiplexing technology to form a sensor array, enabling rapid tomographic analysis of thickener slurry. The software includes steps for measuring the acoustic properties of the slurry, acquiring multi-frequency source tomographic signals, performing multi-frequency source mixed-domain full-waveform inversion imaging, and performing tomographic image super-resolution reconstruction based on deep learning, thereby enabling real-time monitoring of the thickener's operating status. The data processing and tomographic imaging algorithm is as follows: S1, Measurement of acoustic properties of slurry; by establishing an experimental device, the propagation speed of ultrasonic waves of different frequencies in slurries of different concentrations is measured, and a standard characteristic database is established. S2, acquire multi-frequency ultrasonic tomography data; by changing the ultrasonic pulse frequency, obtain the basic data of signal waveform, delay, and amplitude for inversion imaging; S3, Multi-frequency source hybrid domain full waveform inversion imaging; Employing a frequency source hybrid domain full waveform inversion algorithm, high-resolution, high signal-to-noise ratio tomographic images based on different frequencies are obtained; S4, super-resolution image deep learning; uses convolutional neural networks to process tomographic images at different time series to obtain super-resolution tomographic images, namely slurry super-resolution sound velocity model.

2. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that: The integrated ultrasonic sensor array includes a transmitting transducer array, a receiving transducer array, a signal filter, a multiplexer, signal lines, power lines, and a DC power supply. The transmitting transducer array contains several ultrasonic transmitting transducers, which are integrated into the signal filter via signal lines. The receiving transducer array contains several receiving transducers, which are integrated into the multiplexer via signal lines. Both the ultrasonic transmitting and receiving transducers are connected in parallel to the DC power supply via power lines.

3. The online chromatography monitoring system for tailings thickeners as described in claim 2, characterized in that: The ultrasonic receiving transducers all employ broadband ultrasonic sensors; the ultrasonic transmitting transducers are controlled by a computer to excite pulses of different frequencies to achieve signal marking and avoid mutual interference; by changing the excitation frequency of the ultrasonic transducers and performing multiple scans, multi-frequency ultrasonic tomography is achieved; the ultrasonic receiving transducers all employ broadband ultrasonic sensors to achieve response to ultrasonic pulses of different frequencies.

4. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that: The data acquisition system includes a signal filter, a multiplexer, signal lines, a computer, and a memory; the signal filter of the data acquisition system is connected to the multiplexer of the receiving transducer array via signal lines; the multiplexer of the data acquisition system is connected to the signal filter of the transmitting transducer array via signal lines.

5. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that, The standard characteristic database established in S1 is as follows: S101, Construct an experimental device for measuring the mapping relationship between ultrasonic wave velocity and slurry concentration. The experimental device includes a slurry container, an ultrasonic controller, a timer, a transmitting probe, and a receiving probe. S102, inject a slurry of known concentration into the slurry container; the ultrasonic controller generates a signal of a specific frequency and transmits it to the transmitting probe to be converted into ultrasonic pulses, and at the same time starts the timer; the ultrasonic waves are sensed by the receiving probe after passing through the slurry medium, and at the same time the timer is turned off. S103, the acoustic properties of the slurry are determined by the constant density acoustic wave equation. The time domain and frequency domain forms of the constant density acoustic wave equation are as follows: , , in, For the speed of sound, For density, For time, Angular frequency, This represents the wave field pressure value. The epicenter; S104, Change the signal frequency generated by the ultrasonic controller to measure the propagation speed of ultrasonic waves of different frequencies in the slurry of the current concentration. S105, change the slurry with different concentrations, and repeat steps S102, S103, and S104 until the required test results are obtained. Establish a standard database of slurry acoustic properties and determine the correlation between sound velocity, density, time difference, and frequency.

6. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that: The multi-frequency ultrasonic tomography data collected in S2 are as follows: S201, the computer of the data acquisition system encodes a set of control signals, which are sent to the ultrasonic transmitting array by the multiplexer, and the corresponding excitation waveform, frequency, time and amplitude data are recorded; S202, the signal filter of the ultrasonic transmitting array receives the control signal, filters it into multiple control signals, and sends them to the corresponding ultrasonic transmitters; S203, the ultrasonic transmitter excites ultrasonic pulses of a corresponding frequency, the pulses pass through the slurry and are transmitted to the ultrasonic receiver; S204, the ultrasonic receiver converts the received pulse signals of different frequencies into multiple digital signals, including waveform, frequency, arrival time and amplitude data, and transmits them to the multiplexer; S205, the multiplexer combines multiple digital signals into one wideband digital signal and transmits it to the signal filter of the data acquisition system. S206, the signal filter of the data acquisition system filters the broadband digital signal into multiple digital signals and transmits them to the computer for storage of the corresponding data; S207, change the excitation frequency scheme of the ultrasonic transmitter, and repeat steps S201, S202, S203, S204, S205, and S206 to perform multiple excitations, scans, and recordings to obtain multi-frequency source ultrasonic tomography data.

7. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that, The frequency source hybrid domain full waveform inversion algorithm in S3 is as follows: S301, Ultrasonic Time Domain Inversion; First, using low-frequency ultrasonic tomography time difference and amplitude data, the time domain full waveform inversion of the slurry's sound velocity model is performed; through iterative optimization, a high signal-to-noise ratio background sound velocity model based on different low frequencies is established. The objective function for time-domain full waveform inversion is: , in, For velocity model, For velocity model parameters, For time, Total duration and These are the calculated wavefield and the observed wavefield, respectively. The model update gradient is: , , , ， In the formula: For velocity model, For velocity model parameters, For the speed of sound, For time, , and These are the calculated wavefield, the observed wavefield, and the adjoint wavefield, respectively. Choose an appropriate iteration step size, repeatedly calculate the gradient, and update the velocity model to minimize the objective function; S302, Ultrasonic Frequency Domain Inversion: Through Fast Fourier Transform, the ultrasonic time-domain signal is converted into a frequency-domain signal, and then the frequency domain full waveform inversion is performed using the frequency characteristics of high-frequency ultrasonic tomography; through iterative optimization, a high-resolution perturbation velocity model based on different high frequencies is established. The objective function for frequency domain full waveform inversion is: , The model update gradient is: , , , ， In the formula: For velocity model, For velocity model parameters, For the speed of sound, Angular frequency, For time, , and These are the calculated wavefield, the observed wavefield, and the accompanying wavefield, respectively.

8. The online chromatography monitoring system for tailings thickeners as described in claim 1, characterized in that, The deep learning of super-resolution images in S4 is as follows: S401, Model Training: First, a sound velocity model resolution enhancement model is trained and established using a convolutional neural network with a portion of the low-frequency background sound velocity model as input and a portion of the high-frequency perturbation sound velocity model as output, to obtain a super-resolution reconstructed image. S402, Model Evaluation and Validation; The quality of the model output is evaluated by calculating the peak signal-to-noise ratio (PSNR) and structural similarity of the reconstructed image. The higher the PSNR and structural similarity, the better the image quality. The parameters are adjusted to obtain a model that meets the requirements. The formula for calculating peak signal-to-noise ratio is as follows: , , in, To reconstruct the image With low-resolution images The mean square error, and These are the height and width of the image, respectively. For image bit rate, Peak signal-to-noise ratio, in dB; The formula for calculating structural similarity is as follows: , in, For structural similarity, , High-resolution images With reconstructed images Average gray value, , These are the corresponding image variances. Let covariance be the variance of the two images. and It is a constant; S403, Model Testing; The remaining background model and perturbation model are used as the test set to ensure that the established model can run reliably and obtain high-quality reconstructed images; S404 Finally, the background model and disturbance model obtained by inversion are input together to establish a super-resolution model for in-thickness slurry chromatography.

Images