Water tank granularity and mineral characterization method based on laser scattering spots and Raman spectrum

Abstract

The invention provides a water tank granularity and mineral characterization method based on laser scattering spots and a Raman spectrum, and belongs to the technical field of underwater detection. A seabed fan deposition process is reproduced and a deposition body is generated; a multi-mode optical detection system integrating laser ranging, elastic scattering imaging and Raman spectrum acquisition is adopted, a probe scans along a preset path, and bottom type tracking and maximum signal acquisition are realized in combination with real-time bottom type feedback and a dynamic optical compensation mechanism; performing multi-scale denoising and feature enhancement on the scattering image and the Raman spectrum through discrete wavelet transform; particle size distribution is inversed based on a space contrast method, and a three-dimensional particle size diagram is generated in combination with Kriging interpolation; pure mineral Raman signals are extracted through spectrum baseline correction and water interference deduction, and high-precision recognition of mineral components is achieved through spectrum library matching; and establishing a granularity-mineral correlation model through clustering analysis and partial least squares regression. The invention provides a systematic and reliable method for collaborative analysis of the sedimentary structure and the material composition.

Description

TECHNICAL FIELD

[0001] The application provides a water tank particle size and mineral characterization method based on laser scattering spot and Raman spectrum, and belongs to the technical field of underwater detection. BACKGROUND

[0002] The particle size and mineral characterization of underwater sedimentary systems are key links in reservoir evaluation and resource exploration. However, due to the complexity of the underwater environment and the variability of the sedimentary dynamic process, traditional single detection methods are difficult to fully characterize the multimodal characteristics of sediments. In the prior art, laser particle size analysis can only provide particle distribution information without mineralogical basis, and underwater Raman spectrum can identify mineral composition but cannot correlate sediment structure characteristics. Both of them have the defects of serious signal attenuation in high turbidity environment and insufficient data coordination. This technical bottleneck is highly similar to the challenges faced by the geological sedimentary field - just as reservoir modeling needs to integrate multiple sources of data such as seismic, logging, and numerical simulation, underwater sediment characterization also urgently needs to establish a fusion mechanism for multimodal data. Current underwater detection technology has not effectively combined optical scattering characteristics and molecular spectral information, and has not been able to integrate expert experience into intelligent analysis systems like advanced geological knowledge graphs, resulting in obvious limitations in the inversion of sedimentary dynamic processes and the evaluation of reservoir quality.

[0003] In the prior art, underwater particle characterization laser diffraction and dynamic light scattering can only achieve particle size analysis but cannot identify mineral composition, while underwater Raman spectrum can identify minerals but cannot correlate particle distribution characteristics, and both are susceptible to water turbidity interference; although there are attempts to fuse multimodal sensors, there are problems such as system complexity and insufficient collaborative algorithms.

[0004] The shortcomings of the prior art include:

[0005] 1) Underwater particle characterization mainly relies on traditional optical methods such as laser diffraction and dynamic light scattering, which can achieve particle size analysis but cannot simultaneously obtain mineral composition information, and generally have sampling interference and turbidity sensitivity problems.

[0006] 2) Although underwater Raman spectrum technology can identify mineral composition, it is difficult to correlate particle distribution characteristics when used alone, and signal attenuation is serious in turbid environments.

[0007] 3) Imaging techniques based on scattering spots are mostly limited to motion monitoring in laboratory environments, lacking a collaborative mechanism with composition analysis.

[0008] 4) Fusion of multimodal sensors has defects such as system complexity, high cost, and imperfect data coordination algorithms. SUMMARY

[0009] The present application realizes in-situ synchronous precise characterization of underwater particle size and mineral composition by innovative coupling of laser scattering spot and Raman spectrum, to solve the following technical problems:

[0010] First, the "composition-structure" data split problem caused by independent use of traditional laser particle size analysis and Raman spectrum technology, which cannot simultaneously obtain the correlation characteristics of particle size distribution and mineral composition of sediments;

[0011] Second, the technical bottleneck of serious signal interference and significant decline in detection accuracy of existing methods in high turbidity water environment;

[0012] Third, the lack of intelligent analysis system similar to geological knowledge graph, which makes it difficult to effectively integrate optical detection data with geological knowledge such as sediment dynamic process and reservoir evaluation.

[0013] The underwater laser scattering spot-Raman spectrum combined technology proposed by the present application analyzes sediment structure through optical scattering characteristics and identifies mineral composition through Raman spectrum, and constructs a cross-modal correlation model similar to a geological knowledge graph, providing a new technical path for in-situ precise characterization of complex underwater sediment environment. The adaptive signal processing mechanism can be compared to the interactive tool in geological modeling, and finally realizes the leap from single parameter detection to multi-modal intelligent evaluation.

[0014] The present application realizes in-situ synchronous precise characterization of underwater particle size and mineral composition by innovative coupling of laser scattering spot and Raman spectrum, to solve the following technical problems, which provides a new multi-dimensional analysis tool for reservoir quality evaluation and sediment environment analysis. Its technical effect is similar to converting the professional experience of geologists into quantifiable optical detection indicators, significantly improving the reliability and practicality of sediment multi-parameter detection in complex underwater environment.

[0015] The present application reproduces the submarine fan deposition process and generates sedimentary bodies with natural levees and distributary channels and other micro-landforms by high-precision adjustable slope tank system and controllable circulating water environment; a multi-modal optical detection system integrating laser ranging, elastic scattering imaging and Raman spectrum acquisition is adopted, the underwater probe is driven by a mechanical arm to scan along the preset path, the real-time bottom feedback and dynamic optical compensation mechanism are combined to realize bottom tracking and signal maximization acquisition; the scattering images and Raman spectra are denoised and feature enhanced by discrete wavelet transform, and the signal-to-noise ratio and data quality are significantly improved; the spatial contrast method is used to invert the particle size distribution, and the three-dimensional particle size map is generated by combining the Kriging interpolation to display the spatial differentiation characteristics highly coupled with the micro-landform; the pure mineral Raman signal is extracted by spectral baseline correction and water interference subtraction, and the mineral composition is identified with high precision by spectral library matching; finally, the particle size-mineral correlation model is established by cluster analysis and partial least squares regression to reveal the variation of mineral composition in different particle size intervals. The technical system realizes the whole-chain analysis from physical simulation, multi-modal sensing, signal processing to feature fusion, and provides a systematic and reliable method basis for the collaborative analysis of sedimentary structure and material composition.

[0016] The following is a specific technical solution of the scheme, a tank particle size and mineral characterization method based on laser scattering spot and Raman spectrum, comprising the following steps:

[0017] Step one, tank device construction

[0018] A transparent tank is built. Jacks are arranged at the four corners and the midpoint of the long side of the bottom of the tank;

[0019] Two submersible pumps are installed in the water tank and the bottom of the tank respectively, and a closed loop waterway is formed by connecting them through PVC pipes. The outlet of the pump is provided with a filter;

[0020] A turbidimeter, a pH sensor and a thermometer are installed at a depth of 1 / 3 of the middle part of the tank respectively to form a water body monitoring system, and data is transmitted to a computer in real time;

[0021] A material source supply system is built, which includes a mixing bin, a screw sand feeder, a peristaltic pump and a mixing funnel. The mixing bin is rigidly connected with the screw sand feeder, and the peristaltic pump is used to draw experimental water from the water tank. The sand and water are fully mixed through the conical mixing funnel, and then delivered to the upstream inlet of the tank through the transparent PVC pipeline.

[0022] A liftable rigid baffle is arranged at the rear end of the tank as a water level control boundary. A reserved overflow storage area is arranged behind the baffle, and a submersible pump is arranged in the overflow area to pump water to a waste water tank at regular intervals.

[0023] Step two, physical model construction

[0024] An adjustable-thickness sand base is laid at the bottom of the water tank, and the surface flatness is controlled by a leveling device. The sand layer is left to stand naturally to stabilize.

[0025] Adjust the electric jack to make the water tank angle 5°, start the submersible pump in the water storage tank to fill the water to the deepest point of the rigid baffle 1.2m, and then simultaneously adjust the height of the rigid baffle to 1.2m.

[0026] A mixture source was prepared using 100-mesh quartz sand, 325-mesh calcite, and 1000-mesh kaolin. The mixture was then transported via pipeline to the upstream inlet of the water tank.

[0027] The overflow submersible pump was started to drain the water, and a typical submarine fan sedimentary body was obtained under the condition of a 5° tilt in the flume. Its surface developed natural dikes with a height difference of about 3-5 cm and distributary channels with a depth of 4 cm.

[0028] Step 3: Setup and Calibration of the Laser Scanning Device

[0029] The laser integrated instrument is fixed to the end effector of the robotic arm, and the underwater probe is fixed by a rigid connection mechanism. The height of the probe is adjusted to maintain a constant distance of 15±5cm from the bottom surface. The optical axis of the focusing lens is calibrated to be consistent with the normal direction of the bottom surface, forming a stable measurement optical path that is vertically downward.

[0030] The laser beam is reflected by a dichroic mirror to a high-speed galvanometer. The galvanometer quickly deflects to adjust the vertical incident direction of the laser beam, ensuring that the laser is focused on the substrate at different depths.

[0031] Optical system calibration: The field of view of the high-speed camera and spectrometer was calibrated using a standard silicon wafer, with the high-speed camera aligned with the same focal area; then the standard silicon wafer was placed at the laser focal point, and the spectrometer was used to observe the field of view at 520 cm. -1 Raman peak, fine-tuning the lens position to maximize signal intensity, and completing the system optical calibration.

[0032] Laser timing control: The timing output of the dual-mode laser is precisely controlled by a digital delay generator. First, the sample is irradiated with a 50mW continuous laser for 100ms to acquire the scattering spot image. Then, it is immediately switched to a 200mW pulsed laser for Raman spectroscopy excitation. The switching interval between the two modes is controlled within 1μs and the synchronization error is less than 50ns. This achieves interference-free alternating acquisition of the scattering spot and Raman signal.

[0033] Step 4: Real-time automatic fine-tuning of the high-speed camera

[0034] Start the water tank circulation system and adjust the water to the test standard conditions.

[0035] Set up a robot full-coverage pre-scanning path program to drive the laser ranging module to quickly pre-scan the bottom of the tank, collect bottom elevation data in real time and generate a 3D bottom elevation map, and simultaneously record the spatial coordinates (x, y, z) of each measurement point.

[0036] Based on the slope data of the 3D elevation map, the automatic focusing program of the high-speed camera lens is adjusted in real time by the control system to achieve dynamic focal length compensation according to the undulation of the bottom shape; at the same time, the deflection angle of the galvanometer is controlled to correct the horizontal displacement deviation caused by the change of the bottom shape, so as to ensure the stability of the measurement optical path.

[0037] Step 5: Acquisition of laser scattering images and Raman signals

[0038] The robot starts moving along a preset path, and simultaneously the 532nm laser is activated to emit laser light. After being expanded by the beam expander, the laser light is reflected by the dichroic mirror and enters the main optical path channel downwards.

[0039] The 532nm laser, reflected by the dichroic mirror, is incident perpendicularly onto the surface of the water tank substrate through the focusing lens of the underwater probe, forming a 5cm*5cm measurement spot on the surface of the water tank substrate.

[0040] When a 532nm laser irradiates the surface of the substrate particles, the resulting elastically scattered light returns along the original optical path and is reflected laterally at a 90° angle by a dichroic mirror to the high-speed camera for imaging. A 532nm narrowband filter is installed in front of the camera lens.

[0041] The Raman signal (wavelength > 540 nm) generated by the substrate particles excited by a 785 nm laser is collected by the underwater probe, first passes through a dichroic mirror, then undergoes spectral purification through a long-pass filter, and finally couples into the spectrometer.

[0042] Step Six: 3D Visualization of Granularity Distribution

[0043] Scattering spot image preprocessing: The system first acquires scattering spot images, removes background light interference, removes noise using wavelet transform, and then enhances contrast through image enhancement.

[0044] Spatial contrast calculation: After laser irradiation of suspended particles, the scattered light forms random interference speckles on a high-speed camera. The particle size is calculated using the spatial contrast method, and the contrast ratio C of the scattered speckles is calculated: C = , : Standard deviation of light intensity W / m²; I: Average light intensity W / m².

[0045] Establishing a particle size-scattering spot contrast curve: The least squares method was used to fit the measured data of the known particle size d of the standard sample with the corresponding scattering spot contrast C to establish a C−d calibration curve.

[0046] Particle size calibration: Calculate the contrast C of the laser scattering spot image obtained in the experiment, substitute the C value into the calibration curve Cd for inversion calculation, and obtain the particle size d of the particle to be tested. The measured inversion error is ±3%.

[0047] Establish a spatial correlation model: Based on the granular data of each measurement point obtained by the laser scattering spot analysis system, combined with the spatial coordinates (x, y, z) recorded by the laser rangefinder, the Kriging interpolation method is used to perform spatial interpolation calculations on the discrete measurement points, and the semi-variogram analysis shows the best fitting model.

[0048] Granularity space visualization: A generated 3m*1.5m gridded 3D granularity distribution map;

[0049] Step 7: Construction of Characteristic Spectral Library and Mineral Identification

[0050] Spectral preprocessing: After acquiring the raw Raman spectra, the spectral data were first decomposed into multiple scales using wavelet transform; then, the spectral baseline was fitted using the asymmetric least squares method; and the pure mineral Raman characteristic spectra were obtained by simultaneously acquiring pure water Raman spectra and subtracting characteristic peaks.

[0051] Establish a characteristic spectral library: Construct a characteristic spectral library of target minerals in the flume based on a standard mineral Raman database.

[0052] Mineral calibration: The preprocessed Raman spectral data are matched with the characteristic peaks of a standard mineral database.

[0053] In step seven, the specific method for matching characteristic peaks between the preprocessed Raman spectral data and the standard mineral database is as follows:

[0054] The Pearson correlation coefficient R is used to quantitatively assess spectral similarity and screen candidate minerals. When the matching threshold R ≥ 0.85, it is considered a valid match, and the mineral type is determined when the confidence level is ≥ 90%. Otherwise, the candidate mineral with the highest correlation coefficient is selected as the final identification result, and the mineral category and its correlation coefficient are output, with R ≥ 0.75.

[0055]

[0056]

[0057] R: Pearson correlation coefficient. The closer the R value is to 1, the stronger the positive linear correlation.

[0058] n: Number of bands, i.e., how many wavelength channels each spectral curve consists of;

[0059] i: Band index, indicating that the current calculation is for the i-th band;

[0060] x i: The value of the standard mineral spectrum in the i-th band;

[0061] y i : The value of the target spectrum in the i-th band;

[0062] : The average reflectance of a standard mineral spectrum across all wavelengths;

[0063] : The average reflectance of the target spectrum across all bands;

[0064] P: Confidence level.

[0065] Step 8: Establish a grain size-mineral correlation model

[0066] A particle size-mineral correlation model was established: Based on particle size distribution data obtained from laser scattering spot analysis and mineral composition data identified by Raman spectroscopy, the particle size data was first clustered to divide characteristic particle size intervals. Then, for the particle groups within each size interval, the corresponding mineral types and their content proportions were statistically analyzed to establish a correspondence matrix between particle size intervals and mineral composition. Finally, partial least squares regression was used to establish a quantitative relationship model between particle size characteristic parameters and mineral composition.

[0067] The technical effects of this invention are as follows:

[0068] 1. It innovatively solves the industry problem of the disconnect between "composition-structure" data of underwater sediments, and for the first time achieves in-situ synchronous and accurate measurement of particle size distribution and mineral composition, improving the detection accuracy compared with traditional methods.

[0069] 2. By constructing a particle size-mineral correlation model, the detection results can be directly used for reservoir evaluation, providing a new multi-dimensional analysis paradigm for underwater resource exploration. Attached Figure Description

[0070] Figure 1 This is a structural diagram of the water tank device of the present invention;

[0071] Figure 2 This is a structural diagram of the laser device of the present invention. Detailed Implementation

[0072] The following is a clear and complete description of the technical solution of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments:

[0073] I. Water Tank Installation

[0074] 1. For example Figure 1A transparent water tank measuring 5m x 1.5m x 1.5m was constructed, with 20mm thick high-transmittance tempered glass used as the bottom and surrounding enclosure. The glass seams were sealed with waterproof silicone and reinforced with a stainless steel frame.

[0075] 2. Six high-precision mechanical jacks are installed at the four corners and the midpoint of the long side of the bottom of the water tank. Each jack has a load capacity of 2 tons and an adjustment accuracy of ±0.1mm. The lifting stroke of each jack is adjusted synchronously through the central control system, and the slope can be precisely controlled from 0° to 15° within 30 seconds.

[0076] 3. Establish a circulating water supply system: Install two 10m³ / h corrosion-resistant submersible pumps at the bottom of the water tank and water trough respectively, and connect them through PVC pipes to form a closed-loop water circuit. Install a filter at the outlet of the water injection pump to pre-treat particulate matter and other impurities.

[0077] 4. Install the turbidity meter, pH sensor, and thermometer at the middle 1 / 3 depth of the water tank to form a water monitoring system, transmitting data to the computer in real time. The system trial run showed that the water was under standard testing conditions: pH change ±0.2, water temperature 25±1℃, and turbidity change <5%.

[0078] 5. Construct a material supply system, mainly consisting of a 50L mixing chamber, a screw-type sand feeder, a peristaltic pump, and a mixing funnel. The mixing chamber is rigidly connected to the screw-type sand feeder. Simultaneously, the peristaltic pump draws experimental water from the water tank. After the sand and water are thoroughly mixed through the conical mixing funnel, they are transported to the upstream inlet of the water tank via a transparent PVC pipe.

[0079] 6. A liftable rigid baffle is installed at the rear of the water tank, with a height adjustment range of 0-1.3m, serving as the water level control boundary. A 1m*1m*1.5m overflow storage area is reserved behind the baffle. When the water level exceeds the set value, the overflow water automatically crosses the baffle and enters the storage area, thereby maintaining a stable water surface. A 10m³ / h corrosion-resistant submersible pump is built into the overflow area to periodically pump water to the wastewater tank.

[0080] II. Physical Model Construction

[0081] 7. Lay an adjustable sand base at the bottom of the water tank. The sand particle size is 0.75mm and the thickness is controlled at 5±1cm. Use a leveling device to control the surface flatness so that the undulation is <3mm. After laying, let it stand for 12 hours to allow the sand layer to stabilize naturally.

[0082] 8. Adjust the electric jack to make the water tank angle 5°, start the submersible pump of the water storage tank to fill the water to the deepest part of the rigid baffle 1.2m, and then simultaneously adjust the height of the rigid baffle to 1.2m.

[0083] 9. Prepare a mixture source using 100-mesh quartz sand, 325-mesh calcite, and 1000-mesh kaolin. The volume ratio is 1:3:6 under the same compaction degree. Mix the mixture in a mixing chamber at 200 rpm for 15 minutes until homogeneous.

[0084] 10. Set the material supply parameters: set the sand feeder rate to 2.4 kg / min and the peristaltic pump water supply rate to 4.8 L / min. The mixture flows through the pipeline to the upstream inlet of the water tank, with an outlet flow velocity of about 0.1 m / s and a mixture concentration of about 80 g / L.

[0085] 11. After starting the submersible pump in the overflow area to drain the water and continuously supplying sediment for 1 hour, a typical submarine fan sedimentary body with a length, width and thickness of about 3*0.8*0.1m was obtained under the condition of a 5° tilt in the water tank. Its surface has developed micro-landforms such as natural dikes with a height difference of about 3-5cm and distributary channels with a depth of 4cm.

[0086] III. Setup and Calibration of Laser Scanning Device

[0087] 12. For example Figure 2 The laser integrated instrument is fixed to the end effector of the robotic arm, and the underwater probe is fixed by a rigid connection mechanism. The height of the probe is adjusted to maintain a constant distance of 15±5cm from the bottom surface. An O-ring seal is used to ensure waterproof performance, and the optical axis of the focusing lens is calibrated to be consistent with the normal direction of the bottom surface, forming a stable vertical downward measurement optical path.

[0088] 13. The laser beam is reflected by a dichroic mirror to a high-speed galvanometer. The galvanometer quickly deflects to adjust the vertical incident direction of the laser beam, ensuring that the laser is focused on the substrate at different depths. If the substrate of the tank is tilted, the galvanometer can deflect in real time to keep the laser vertically incident, maximizing the signal.

[0089] 14. Optical System Calibration: The high-speed camera and spectrometer field of view were calibrated using a standard silicon wafer. The high-speed camera was aligned with the same focal area to verify that the speckle contrast was ≥0.3. Subsequently, the standard silicon wafer was placed at the laser focal point, and the spectrometer was used to observe the field of view at 520 cm⁻¹. -1 Raman peak, fine-tuning the lens position to maximize signal intensity, and completing the system optical calibration.

[0090] 15. Laser Timing Control: The timing output of the dual-mode laser is precisely controlled by a digital delay generator. First, a 50mW continuous laser (532nm, 10mm spot size) is used to irradiate the sample for 100ms to acquire the scattering spot image. Then, it is immediately switched to a 200mW pulsed laser (785nm, 10ns pulse width, 1kHz repetition rate) for Raman spectroscopy excitation. The switching interval between the two modes is controlled within 1μs and the synchronization error is less than 50ns. The laser power and pulse width are monitored and fed back in real time by the built-in photodiode to ensure that the power fluctuation of the continuous mode is <±1% and the pulse energy stability is >98%, ultimately achieving interference-free alternating acquisition of the scattering spot and Raman signal.

[0091] IV. Real-time automatic fine-tuning of high-speed camera

[0092] 16. Start the water tank circulation system and adjust the water to the test standard conditions of pH value change ±0.2, water temperature 25±1℃, and turbidity change <5%.

[0093] 17. Set up a full-coverage pre-scanning path program for the robot, drive the laser ranging module to perform a rapid pre-scan of the bottom of the tank, collect bottom elevation data in real time and generate a 3D elevation map of the bottom, and simultaneously record the spatial coordinates (x, y, z) of each measurement point.

[0094] 18. Based on the slope data of the 3D elevation map, the automatic focusing program of the high-speed camera lens is adjusted in real time by the control system to achieve dynamic focal length compensation according to the undulation of the bottom shape; at the same time, the deflection angle of the galvanometer is controlled to correct the horizontal displacement deviation caused by the change of the bottom shape, so as to ensure the stability of the measurement optical path.

[0095] V. Acquisition of Laser Scattering Images and Raman Signals

[0096] 19. Set the robot bottom scanning program: the scanning speed is 10mm / s, pauses every 5cm to check once, and scans in an S-shaped path to achieve systematic measurement coverage of the bottom of the water tank.

[0097] 20. Start the robot to move along the preset path, and simultaneously start the 532nm laser to emit laser light. After the laser beam is expanded by the beam expander, it is incident on the dichroic mirror and reflected downwards into the main optical path channel.

[0098] 21. The 532nm laser reflected by the dichroic mirror is incident perpendicularly onto the surface of the water tank substrate through the focusing lens of the underwater probe, forming a 5cm*5cm measurement spot on the surface of the water tank substrate.

[0099] 22. When a 532nm laser irradiates the surface of the substrate particles, the generated elastically scattered light returns along the original optical path and is laterally reflected at a 90° angle by a dichroic mirror to the high-speed camera for imaging. The 532nm narrowband filter installed in front of the camera lens effectively suppresses interference signals from ambient light and other wavelengths. The optical system maintains more than 98% elastically scattered light transmittance under ambient light interference of 500W / m². When the camera exposure time is 1μs, the signal-to-noise ratio of the elastically scattered image exceeds 40dB.

[0100] 23. The Raman signal (wavelength > 540 nm) generated by the substrate particles excited by a 785 nm laser is collected by the underwater probe, first passes through a dichroic mirror, then undergoes spectral purification through a long-pass filter, and finally is coupled into the spectrometer. This optical path system can effectively suppress 90% of Rayleigh scattering light at an excitation power of 500 mW, so that the signal-to-noise ratio of the characteristic Raman peak reaches 50:1.

[0101] VI. Three-dimensional visualization of granular distribution

[0102] 24. Scattering Spot Image Preprocessing: The system first acquires scattering spot images, removes background light interference, removes noise using wavelet transform, and then enhances contrast through image enhancement, ultimately improving the image signal-to-noise ratio from 28.6dB to 41.2dB and increasing particle feature clarity by 3.8 times.

[0103] 25. Spatial Contrast Calculation: After laser irradiation of suspended particles, the scattered light forms random interference speckles on a high-speed camera. The spatial contrast method is used to calculate the particle size and the contrast of the scattered speckles, C: C = , : Standard deviation of light intensity W / m²; I: Average light intensity W / m².

[0104] 26. Establishing the particle size-scattering spot contrast curve: The least squares method was used to fit the measured data of the known particle size d of the standard sample with the corresponding scattering spot contrast C, and a C-d calibration curve was established. The calibration curve exhibits a typical three-segment characteristic: significant nonlinearity in the small particle size (d<30μm) range, with the contrast increasing rapidly with increasing particle size; the change slows down in the medium particle size (30-60μm) range; and it tends to saturate in the large particle size (d>60μm) range.

[0105] 27. Particle size calibration: Calculate the contrast C of the laser scattering spot image obtained in the experiment, substitute the C value into the calibration curve Cd for inversion calculation, and obtain the particle size d of the particle to be tested. The measured inversion error is ±3%.

[0106] 28. Establishing a spatial correlation model: Based on the particle size data of 2400 measurement points obtained by the laser scattering spot analysis system, combined with the spatial coordinates (x, y, z) recorded by the laser rangefinder, the Kriging interpolation method was used to perform spatial interpolation calculations on the discrete measurement points. The semi-variogram analysis showed that the best-fit model was the spherical model, with a nugget value of 0.12, a sill value of 1.85, a range of 35 cm, an anisotropy ratio of 1.8, and stronger correlation along the water flow direction.

[0107] 29. Spatial visualization of grain size: The generated 3m*1.5m gridded 3D grain size distribution map shows that the median grain size D50 in the study area ranges from 15 to 160 μm. Among them, the coarse grain area has D50>60 μm and is concentrated at the root of the fan, accounting for 18% of the area; the fine grain area has D50<30 μm and is mainly distributed at the edge of the fan, accounting for 32% of the area. The mean absolute error is 4.2 μm and the root mean square error is 6.8 μm. The 3D visualization shows that the grain size distribution and the micro-topography of the seafloor fan have significant spatial coupling characteristics.

[0108] VII. Construction of Characteristic Spectral Library and Mineral Identification

[0109] 30. Spectral Preprocessing: After acquiring the raw Raman spectra, the spectral data were first decomposed into multiple scales using wavelet transform to filter out high-frequency noise components and retain the effective signal frequency bands, improving the signal-to-noise ratio by 8.2 dB. Subsequently, the spectral baseline was fitted using the asymmetric least squares method with a weighting parameter λ=10. 5 This method eliminates baseline tilt caused by fluorescence and instrument drift. Addressing the unique underwater environment's water interference, it simultaneously acquires pure water Raman spectra and subtracts characteristic peaks to eliminate the superposition effect of water background on mineral characteristic peaks, achieving an elimination rate of 98.7%. This results in pure mineral Raman characteristic spectra with a deviation of less than ±0.5 cm⁻¹. -1 The relative error of peak intensity is less than 3%.

[0110] 31. Establish a characteristic spectral library: Construct a characteristic spectral library of the target minerals in the water tank based on a standard mineral Raman database. Extract the characteristic Raman peak positions of each mineral and accurately calibrate their shift values ​​(cm). -1 To establish a structured spectral database containing information such as mineral name, chemical formula, crystal structure, characteristic peak position, and intensity ratio, and relative intensity, we will collect data on minerals and their relative intensities.

[0111] 32. Mineral Identification: The preprocessed Raman spectral data were matched with characteristic peaks in a standard mineral database. The Pearson correlation coefficient (R) was used to quantitatively assess spectral similarity and screen candidate minerals. A valid match was considered to have a matching threshold R ≥ 0.85, and the mineral type was determined with a confidence level ≥ 90%. Otherwise, the candidate mineral with the highest correlation coefficient was selected as the final identification result, and the mineral category and its correlation coefficient (R ≥ 0.75) were output. Experimental Results: Quartz characteristic peak at 465 cm⁻¹. -1 The matching degree R=0.92, and the confidence level is 93.5%; calcite 1086cm -1 The matching degree R=0.86, and the confidence level 90.3%; kaolin 3620cm -1 The matching degree R=0.91, with a confidence level of 92.6%. Experimental results show that the method achieves an effective matching rate of 91.3% (R≥0.85) and an estimated recognition accuracy of 88.7%.

[0112]

[0113]

[0114] R: Pearson correlation coefficient. The closer the R value is to 1, the stronger the positive linear correlation.

[0115] n: Number of bands, i.e., how many wavelength channels each spectral curve consists of;

[0116] i: Band index, indicating that the current calculation is for the i-th band;

[0117] x i : The value of the standard mineral spectrum in the i-th band;

[0118] y i : The value of the target spectrum in the i-th band;

[0119] : The average reflectance of a standard mineral spectrum across all wavelengths;

[0120] : The average reflectance of the target spectrum across all bands;

[0121] P: Confidence level.

[0122] 8. Establish a grain size-mineral correlation model

[0123] 33. Establishing a Particle Size-Mineral Correlation Model: Based on particle size distribution data obtained from laser scattering spot analysis and mineral composition data identified by Raman spectroscopy, the particle size data is first clustered to divide characteristic particle size intervals. Then, for each particle group within a size interval, the corresponding mineral types and their content proportions are statistically analyzed to establish a correspondence matrix between particle size intervals and mineral composition. Finally, partial least squares regression is used to establish a quantitative relationship model between particle size characteristic parameters and mineral composition.

Claims

1. A water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy, characterized in that, By employing a high-precision adjustable slope flume system and a controllable circulating water environment, the depositional process of submarine fans was reproduced, and sedimentary bodies with micro-topography were generated. A multi-modal optical detection system integrating laser ranging, elastic scattering imaging, and Raman spectroscopy was used. The underwater probe scanned along a preset path, and combined with real-time bottom shape feedback and dynamic optical compensation mechanisms, bottom shape tracking and maximum signal acquisition were achieved. Discrete wavelet transform was used to perform multi-scale denoising and feature enhancement on the scattering images and Raman spectra. The particle size distribution is inverted based on the spatial contrast method, and a three-dimensional particle size map is generated by combining Kriging interpolation to show the spatial differentiation characteristics that are highly coupled with micro-topography. By utilizing spectral baseline correction and water interference subtraction to extract pure mineral Raman signals, high-precision identification of mineral components is achieved through spectral library matching; a particle size-mineral correlation model is established using cluster analysis and partial least squares regression.

2. The water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy according to claim 1, characterized in that, Includes the following steps: Step 1: Setting up the water tank assembly A jack is installed at the bottom of the water tank; the water tank and water trough are connected by a submersible pump and pipeline to form a closed-loop water circuit; a turbidity meter, pH sensor and thermometer are installed at the middle 1 / 3 depth of the water tank, and the data is transmitted to the computer in real time. After the sand and water are mixed, they are transported through pipes to the upstream inlet of the water tank; A liftable rigid baffle is installed at the rear of the water tank as a water level control boundary. An overflow storage area is reserved behind the baffle, and a submersible pump is built into the overflow area. Step 2: Physical Model Construction Lay an adjustable-thickness sand base at the bottom of the water tank, ensuring the surface is flat, and let it stand. Adjust the jack to make the water tank angle 5°, and fill the tank with water to the deepest point of the rigid baffle. A mixture of quartz sand, calcite, and kaolin is prepared and transported to the upstream inlet of the water tank. The overflow area was drained, and under the condition of a 5° tilt in the flume, a typical submarine fan sedimentary body was obtained, and its surface developed into natural dikes and distributary channel micro-topography. Step 3: Setup and Calibration of the Laser Scanning Device The laser integrated instrument is fixed to the end effector of the robotic arm, and the underwater probe is fixed through a rigid connection mechanism. The optical axis of the focusing lens is aligned with the normal direction of the bottom surface to form a stable measurement optical path that is vertically downward. The laser beam is reflected by a dichroic mirror to a high-speed galvanometer, which rapidly deflects to adjust the vertical incident direction of the laser beam. The high-speed camera and spectrometer field of view were calibrated using a standard silicon wafer, with the high-speed camera aligned with the same focal area. The standard silicon wafer was then placed at the laser focus, and the spectrometer was used to observe the field of view at 520 cm⁻¹. -1 Raman peaks; fine-tuning the lens position to maximize signal strength. By precisely controlling the timing output of the dual-mode laser using a digital delay generator, interference-free alternating acquisition of the scattered spot and Raman signal is achieved. Step 4: Real-time automatic fine-tuning of the high-speed camera Start the water tank circulation system and adjust the water to the test standard conditions; The laser ranging module is driven to perform a rapid pre-scan of the bottom of the tank, collect bottom elevation data in real time and generate a 3D elevation map of the bottom, and simultaneously record the spatial coordinates (x, y, z) of each measurement point; the control system adjusts the autofocus of the high-speed camera lens in real time to achieve dynamic focal length compensation as the bottom undulates; at the same time, the galvanometer deflection angle is controlled to correct the horizontal displacement deviation caused by changes in the bottom shape. Step 5: Acquisition of laser scattering images and Raman signals The robot starts moving along the preset path, and the 532nm laser is activated simultaneously to emit laser light. After being expanded by the beam expander, the laser light is reflected by the dichroic mirror and enters the main optical path channel downwards. The 532nm laser, reflected by the dichroic mirror, is incident perpendicularly onto the surface of the water tank substrate through the focusing lens of the underwater probe, forming a 5cm*5cm measurement spot on the surface of the water tank substrate. When a 532nm laser irradiates the surface of the substrate particles, the resulting elastic scattered light returns along the original optical path, is reflected laterally at a 90° angle by a dichroic mirror, and is imaged by a high-speed camera. A 532nm narrowband filter is installed in front of the camera lens. The Raman signal generated by the substrate particles excited by the 785nm laser is collected by the underwater probe, first passes through the dichroic mirror, then passes through the long-pass filter for spectral purification, and finally couples into the spectrometer. Step Six: 3D Visualization of Granularity Distribution First, images of the scattered spots are acquired. After removing background light interference, wavelet transform is used to remove noise, and then image enhancement is used to improve contrast. After the suspended particles are irradiated with laser light, the scattered light forms random interference speckles on a high-speed camera; the particle size is calculated using the spatial contrast method, and the contrast of the scattered speckles is calculated. The least squares method was used to fit the measured data of the known particle size d of the standard sample and the corresponding scattering spot contrast C to establish the C−d calibration curve. The contrast C of the laser scattering spot image obtained in the experiment was calculated. The C value was substituted into the calibration curve Cd for inversion calculation to obtain the particle size d of the particle to be tested. The measured inversion error was ±3%. Based on the granular data of each measurement point obtained by the laser scattering spot analysis system, combined with the spatial coordinates (x, y, z) recorded by the laser rangefinder, the Kriging interpolation method is used to perform spatial interpolation calculation on the discrete measurement points, and the semi-variogram analysis shows the best fitting model. Granularity space visualization: A generated 3m*1.5m gridded 3D granularity distribution map; Step 7: Construction of Characteristic Spectral Library and Mineral Identification After acquiring the raw Raman spectra, the spectral data were first decomposed into multiple scales using wavelet transform; then, the spectral baseline was fitted using the asymmetric least squares method; and the pure mineral Raman characteristic spectra were obtained by simultaneously acquiring pure water Raman spectra and subtracting characteristic peaks. A characteristic spectral library of target minerals in the flue was constructed based on a standard mineral Raman database. The preprocessed Raman spectral data were matched with characteristic peaks in a standard mineral database. Step 8: Establish a grain size-mineral correlation model.

3. The water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy according to claim 2, characterized in that, In step two, sand with a particle size of 0.75mm is laid at the bottom of the water tank, with a thickness controlled at 5±1cm; A mixture source was prepared using 100-mesh quartz sand, 325-mesh calcite, and 1000-mesh kaolin, with a volume ratio of 1:3:6 under the same compaction degree.

4. The water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy according to claim 2, characterized in that, In step three, the laser timing control method is as follows: first, the sample is irradiated with a 50mW continuous laser for 100ms to acquire the scattering spot image, and then immediately switched to a 200mW pulsed laser for Raman spectroscopy excitation. The switching interval between the two modes is controlled within 1μs and the synchronization error is less than 50ns. The laser power and pulse width are monitored and fed back in real time through the built-in photodiode to ensure that the power fluctuation of the continuous mode is <±1% and the pulse energy stability is >98%, and finally, the interference-free alternating acquisition of the scattering spot and Raman signal is achieved.

5. The water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy according to claim 2, characterized in that, In step seven, the specific method for matching the characteristic peaks of the preprocessed Raman spectral data with the standard mineral database is as follows: Pearson correlation coefficient R was used to quantitatively assess spectral similarity and screen candidate minerals. When the matching threshold R ≥ 0.85, it was considered a valid match, and the mineral type was determined when the confidence level was ≥ 90%. Otherwise, the candidate mineral with the highest correlation coefficient is selected as the final identification result, and the mineral category and its correlation coefficient are output, requiring R≥0.75; ； ； R: Pearson correlation coefficient. The closer the R value is to 1, the stronger the positive linear correlation. n: Number of bands, i.e., how many wavelength channels each spectral curve consists of; i: Band index, indicating that the current calculation is for the i-th band; x i : The value of the standard mineral spectrum in the i-th band; y i : The value of the target spectrum in the i-th band; : The average reflectance of a standard mineral spectrum across all wavelengths; : The average reflectance of the target spectrum across all bands; P: Confidence level.

6. The water tank particle size and mineral characterization method based on laser scattering spots and Raman spectroscopy according to claim 2, characterized in that, Step eight is as follows: Based on particle size distribution data obtained from laser scattering spot analysis and mineral composition data identified by Raman spectroscopy, the particle size data is first clustered to divide characteristic particle size intervals. Then, for the particle groups in each particle size interval, the corresponding mineral types and content proportions are statistically analyzed to establish a correspondence matrix between particle size intervals and mineral composition. Finally, partial least squares regression is used to establish a quantitative relationship model between particle size characteristic parameters and mineral composition.

Images