Low-loss multi-mode self-adaptive aerolite rapid classification and verification system and method thereof

Abstract

The invention provides a low-loss multi-mode self-adaptive aerolite rapid classification and verification system and a method thereof, and belongs to the technical field of planetary science and material analysis. Comprising a multi-modal pre-screening module, a low-loss slide making module, an EDS data acquisition and feature extraction module, an optical microscopy and scanning electron microscope analysis module, a weathering grade perception adaptive discrimination module, an uncertainty driving verification module and a classification result output and sample database management module. Through the synergistic effect of technical means such as low-loss preparation, multi-point calibration, weathering self-adaption and probability verification closed loop, the key bottleneck in the prior art is overcome, and the stability, accuracy and expandability of the rapid classification process of the Antarctic meteorite are remarkably improved.

Description

Technical Field

[0001] This invention relates to the fields of planetary science and materials analysis technology, and in particular to a low-loss, multimodal adaptive rapid classification and verification system and method for meteorites. Background Technology

[0002] Antarctica is one of the most important known meteorite-rich regions in the world. Its unique polar ice sheet flow mechanism and low-weathering environment have preserved a large number of meteorites, which are concentrated and exposed in blue ice areas and drift ice zones. With the significant improvement of my country's polar scientific research capabilities, the annual number of Antarctic meteorites recovered has continued to increase. The samples are diverse in type and large in quantity, gradually creating an urgent need for large-scale, systematic identification and scientific research.

[0003] The traditional cataloging and classification process for Antarctic meteorites typically includes: (1) saw blade or diamond wire cutting and preparation of standard petrographic slices; (2) observation of texture structure using optical microscope and scanning electron microscope reflected electron (BSE); (3) high-precision mineral chemical determination using EPMA-WDS (electron probe microanalysis); and (4) manual judgment based on comparison of typical mineral chemical ranges and spherulitic structure characteristics.

[0004] Although the above process is the mainstream method in the current international meteoritics field, it has significant limitations in large-scale sample processing scenarios: (1) The process cycle is long. It often takes 2-4 weeks for a single sample to go from petrographic preparation, polishing, analysis to completing the classification report. Under the condition of a large number of Antarctic samples being returned, the processing cycle and efficiency are difficult to meet the needs of rapid scientific output and database construction. (2) The sample damage risk is high. Antarctic meteorites commonly contain metallic phases, brittle sulfide phases, and fine carbonaceous matrix, which are sensitive to water-cooled cutting and mechanical polishing. They are prone to irreversible damage such as oxidation, shedding, particle peeling and pyrolysis, which affect subsequent fine analysis and long-term preservation value. (3) There is a significant dependence on high-end instruments. The existing process relies on large-scale analysis platforms such as EPMA-WDS for full quantitative analysis. Such equipment is time-consuming and costly, making it difficult to quickly screen and prioritize a large number of suspected samples, which limits scientific research efficiency and resource allocation. (4) Significant weathering and impact interference: The weathering grade (W0–W6) and impact metamorphism grade (S1–S6) of Antarctic samples differ significantly, affecting the preservation of metallic phases, lithofacies structure, and element migration. This leads to the instability of traditional fixed threshold discrimination rules in highly weathered or highly impacted samples, resulting in insufficient classification accuracy and robustness. (5) Lack of a reliable quantitative confidence system: The existing discrimination process relies on the experience of professionals and the judgment of typical composition ranges, lacking a unified quantitative confidence index. This makes it difficult to standardize, track, and trace the classification results in large-scale sample processing. (6) Lack of quality closed-loop and verification mechanism: The existing process relies on manual verification for classification reliability. There is no uncertainty-driven automatic sampling and reanalysis mechanism, which easily leads to the accumulation of misjudgments or waste of resources, failing to meet the requirements of large-scale scientific database construction and high reliability in the real world. Therefore, it is necessary to design a low-loss, multimodal adaptive meteorite rapid classification and verification system and its method. Summary of the Invention

[0005] The purpose of this invention is to provide a low-destruction, multimodal, adaptive rapid meteorite classification and verification system and method, which solves the technical problems of high sample damage risk, large quantitative error in mineral chemistry, significant weathering interference, and lack of confidence criteria in existing rapid Antarctic meteorite classification technologies.

[0006] This system is designed for rapid primary identification of major meteorite types, including common chondrites (H, L, LL), enstatite meteorites (EH, EL), and carbonaceous chondrites (CM, CO, CV). It is also compatible with the intelligent cataloging needs of lunar samples (such as glassy cement, breccia, and samples with complex granular-intercalated structures), Martian meteorites, and asteroid particles. By introducing multimodal information fusion, low-destructive preparation processes, adaptive thresholds, and uncertainty closure, this invention addresses key challenges in current meteorite identification processes, such as sample preparation damage risks, weathering sensitivity, quantitative errors, and reliance on human experience. It achieves an efficient, traceable, and quantitatively confident rapid meteorite classification system.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A low-destructive, multimodal adaptive meteorite rapid classification and verification system includes a multimodal pre-screening module, a low-destructive slide preparation module, an EDS data acquisition and feature extraction module, an optical microscopy and scanning electron microscopy analysis module, a weathering grade sensing and adaptive discrimination module, an uncertainty-driven verification module, and a classification result output and sample database management module. The multimodal pre-screening module is connected to the low-destructive slide preparation module, which is used for non-destructive evaluation of sample composition, structure, and preservation status, determining slide preparation strategies and subsequent analysis routes. The low-destructive slide preparation module is connected to the EDS data acquisition and feature extraction module, which is used for... The characteristics of meteorites, such as metals, sulfides, and brittle carbonaceous matrix, are prone to oxidation, detachment, and fragmentation. The optical microscopy and scanning electron microscopy analysis module is connected to the EDS data acquisition and feature extraction module. The optical microscopy and scanning electron microscopy analysis module is used to obtain whole-rock, microscopic, and trace component evidence. The EDS data acquisition and feature extraction module is connected to the weathering grade perception and adaptive discrimination module. The weathering grade perception and adaptive discrimination module is used to construct a multi-dimensional feature model and dynamic discrimination system for meteorites. Both the weathering grade perception and adaptive discrimination module and the uncertainty-driven verification module are connected to the classification result output and sample database management module.

[0009] Furthermore, the multimodal pre-screening module includes a handheld magnetic susceptibility meter, a bulk density measurement device, a micro-CT, a miniature microscope, and a Raman / visible-near-infrared spectroscopy device. It identifies layered silicates, hydrous minerals, CAI enrichment areas, carbonaceous matrix, and hydration absorption bands, and obtains the magnetic susceptibility, bulk density, visible-near-infrared spectrum, microscopic appearance, and optional micro-CT structural information of the target meteorite sample. This information is used to identify metal content, carbonaceous matrix, degree of fracture development, and potential spherulitic structure, enabling preliminary classification of sample physical properties and mineral characteristics, as well as selection of slide preparation path.

[0010] Furthermore, the low-loss film preparation module selects ethanol or ethanol-water cooling cutting method, vacuum epoxy holding method, and load-adaptive polishing strategy based on pre-screening information to prepare thin films with a thickness of 20–35 µm. This allows the metallic phase, sulfide phase, and fine-grained matrix to maintain their in-situ structure. Ethanol / ethanol-water lubrication and cooling cutting is used to reduce iron-nickel phase oxidation and Fe migration. Vacuum epoxy holding and low-load polishing system protect the fine-grained and brittle mineral structure. The adaptive thickness control program is 20–35 µm to preserve impact veins, molten inclusions, spheroidal contours, and glass structure.

[0011] Furthermore, the optical microscopy and scanning electron microscopy analysis module performs optical microscopy observation and scanning electron microscopy reflection electron imaging to obtain texture features of the thin section, and performs energy dispersive spectroscopy point analysis and region analysis to extract mineral phase composition information, metal volume fraction and Fe / Mg ratio estimation of olivine and pyroxene, automatically identify metal, sulfide, olivine or pyroxene phase domains, and provide mineral chemical characteristics, texture features and mineral phase separation information.

[0012] Furthermore, the weathering grade perception adaptive discrimination module takes weathering grade and impact classification as independent variables and inputs them into the classification model. It combines mineral chemical ratios, physical properties and structural features to dynamically adjust the discrimination weights and confidence intervals, including but not limited to weathering sensitive parameters Mg / Si and Fe / Si ratios and EDS regression Fo-Fs values.

[0013] Furthermore, the uncertainty-driven verification module is used to output a confidence level p based on the classification. When p is less than a set threshold, it triggers an electron probe microanalysis (WDS) for precise verification and updates the classification parameters and calibration curve based on the verification results. The classification result output and sample database management module is used to output the meteorite chemical group, petrological type, impact level, weathering grade, and confidence level index of the sample, forming a multi-dimensional discrimination result with uncertainty evaluation.

[0014] A method for a low-loss, multimodal, adaptive rapid meteorite classification and verification system, the method comprising the following steps:

[0015] Step 1: Pre-screening using several modes to obtain information on the magnetic susceptibility, bulk density, visible-near-infrared spectrum, microscopic appearance, and optional micro-CT structure of the target meteorite sample, which is used to identify metal content, carbonaceous matrix, degree of fracture development, and potential spherulitic structure.

[0016] Step 2: Low-loss film preparation. Based on the pre-screening information, select ethanol or ethanol-water cooling cutting method, vacuum epoxy holding method and load adaptive grinding and polishing strategy to prepare thin films with a thickness of 20–35 µm, so that the metal phase, sulfide phase and fine matrix can maintain the in-situ structure.

[0017] Step 3: Acquire several modal signals, observe the thin section with an optical microscope, obtain texture features by scanning electron microscopy reflection electron imaging, and perform energy dispersive spectroscopy point analysis and region analysis to extract mineral phase composition information, metal volume fraction and estimate the Fe / Mg ratio of olivine and pyroxene;

[0018] Step 4: Adaptive classification based on weathering grade perception threshold. Weathering grade and impact classification are input into the classification model as independent variables. The discrimination weights and confidence intervals are dynamically adjusted by combining mineral chemical ratios, physical properties and structural features, including but not limited to weathering sensitive parameters Mg / Si and Fe / Si ratios and EDS regression Fo-Fs values.

[0019] Step 5: Uncertainty-driven verification and iteration. Output confidence level p based on classification. When p is less than the set threshold, trigger precise electron probe microanalysis (WDS) for verification, and update classification parameters and calibration curves based on the verification results.

[0020] Step 6: Output the five-dimensional classification results, including the meteorite chemical group to which the sample belongs, petrological type, impact level, weathering grade, and confidence index, forming a multi-dimensional discrimination result with uncertainty evaluation.

[0021] Furthermore, in step 2, the concentration of ethanol coolant is 70–100 wt% to reduce the risk of metal oxidation and sulfide hydration. In step 3, the thickness of the sheet is 20–35 µm, which is adaptively adjusted according to the sample cracks, carbon content and metal phase content.

[0022] Furthermore, in step 1, the target meteorite sample is vacuum-embedded with epoxy resin to enhance mechanical stability and prevent particles from being pulled out during polishing. In step 3, there are no fewer than 30 sampling points, including the background, mineral phase interface and typical particles, and drift correction or standard sample regression is automatically performed every 60–120 seconds.

[0023] Furthermore, in step 4, the adaptive classification includes a mineral ratio model, a feature parameter model, and an EDS-estimated Fo-Fs regression model.

[0024] The present invention, by adopting the above-described technical solution, has the following beneficial effects:

[0025] (1) This invention solves the problem of oxidation of high Fe-Ni minerals and detachment of fragile phases caused by traditional water-cooled cutting. This invention uses ethanol or a high ethanol-water system for cutting lubrication, and combines it with a low load pressure polishing strategy. By reducing water-metal reaction and cutting heat accumulation, it achieves the technical effect of maintaining the original morphology of the metal phase and reducing the propagation of microcracks.

[0026] (2) To solve the problems of large component fluctuations and regional biased sampling in rapid EDS measurement, this invention proposes a ≥30-point stratified sampling strategy and introduces periodic standard calibration and regression compensation methods. By improving the representative sampling density and drift correction accuracy, the technical effect of significantly improving the stability of semi-quantitative mineral chemistry results is achieved.

[0027] (3) To solve the problem of component discrimination threshold drift caused by the difference in weathering grade of Antarctic meteorites, this invention uses weathering grade W as the input parameter of the classification model and adopts a weathering sensitive threshold adaptive mechanism, so that the model can dynamically adjust the discrimination conditions in samples with different weathering degrees, thereby achieving the technical effect of enhancing classification accuracy under weathering interference conditions.

[0028] (4) To address the problem of the lack of confidence output and reliance on human experience in the existing meteorite identification system, this invention introduces a classification uncertainty assessment framework based on posterior probability and sets a confidence threshold to trigger the WDS sampling and verification mechanism, so that the classification results automatically enter the high-precision verification process in the low confidence region, thereby achieving the technical effects of interpretable classification results, improved reliability and reduced risk of error accumulation.

[0029] (5) This invention overcomes the key bottlenecks of existing technologies by using low-loss preparation, multi-point calibration, weathering adaptation and probabilistic verification closed loop, and significantly improves the stability, accuracy and scalability of the rapid classification process for Antarctic meteorites. Attached Figure Description

[0030] Figure 1 This is a system block diagram of the present invention;

[0031] Figure 2 This is a flowchart of the method of the present invention;

[0032] Figure 3 This is a flowchart of the adaptive weathering level threshold of the present invention;

[0033] Figure 4 This is a flowchart of the uncertainty-driven WDS verification closed loop of the present invention;

[0034] Figure 5 This is a flowchart of the multimodal feature extraction and fusion process of the present invention;

[0035] Figure 6 This is a flowchart of the online update and re-inspection process of this invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are merely to provide the reader with a thorough understanding of one or more aspects of the present invention, and these aspects of the invention can be implemented even without these specific details.

[0037] like Figure 1As shown, the low-destructive multimodal adaptive meteorite rapid classification and verification system includes a multimodal pre-screening module, a low-destructive slide preparation module, an EDS data acquisition and feature extraction module, an optical microscopy and scanning electron microscopy analysis module, a weathering grade sensing adaptive discrimination module, an uncertainty-driven verification module, and a classification result output and sample database management module. The multimodal pre-screening module is connected to the low-destructive slide preparation module. The multimodal pre-screening module is used for non-destructive evaluation of sample composition, structure, and preservation status, determining slide preparation strategies and subsequent analysis routes. The low-destructive slide preparation module is connected to the EDS data acquisition and feature extraction module. The low-destructive slide preparation module is used for... To address the characteristics of meteorites' metal, sulfide, and brittle carbonaceous matrix being easily oxidized, detached, and fractured, the optical microscopy and scanning electron microscopy analysis module is connected to the EDS data acquisition and feature extraction module. The optical microscopy and scanning electron microscopy analysis module is used to obtain whole-rock, microscopic, and trace component evidence. The EDS data acquisition and feature extraction module is connected to the weathering grade perception and adaptive discrimination module. The weathering grade perception and adaptive discrimination module is used to construct a multi-dimensional meteoritic feature model and dynamic discrimination system. Both the weathering grade perception and adaptive discrimination module and the uncertainty-driven verification module are connected to the classification result output and sample database management module.

[0038] This system consists of modules such as multi-source information acquisition, low-loss slide preparation, multimodal analysis, intelligent discrimination and adaptive correction, and quality control and verification, realizing a closed-loop process from initial sample screening to intelligent classification.

[0039] Sample pre-screening module

[0040] Used for non-destructive evaluation of sample composition, structure, and preservation status, determining slide preparation strategies and subsequent analytical routes, including: a handheld magnetic susceptibility meter (for rapid determination of metallic phase content), a volume density measurement device (to assist in distinguishing between metal-enriched and carbonaceous-enriched samples), a micro-CT (optional, for identifying spherulitic structures, fracture development, and metal aggregates), and a miniature microscope and Raman / visible-near-infrared spectroscopy device to identify layered silicates, hydrous minerals, CAI-enriched areas, carbonaceous matrices, and hydration absorption bands.

[0041] This module enables preliminary classification of sample physical properties and mineral characteristics, as well as selection of slide preparation methods (such as ethanol cooling, vacuum epoxidation, thick slide retention, etc.).

[0042] 2) Low-loss film preparation module

[0043] To address the characteristics of meteorites, such as the susceptibility to oxidation, breakage, and fragmentation of metals, sulfides, and brittle carbonaceous matrix, the following methods were employed: ethanol / ethanol-water lubrication and cooling cutting (to reduce iron-nickel phase oxidation and Fe migration), vacuum epoxy holding and low-load polishing system (to protect fine-grained and brittle mineral structures), and adaptive thickness control program (20–35 µm to preserve impact veins, melt inclusions, spherule profiles, and glassy structures).

[0044] This module enables the fabrication of thin films with stability, structural integrity, and low-loss operation.

[0045] 3) Multimodal analysis module

[0046] Used to obtain whole-rock, microscopic, and trace compositional evidence, including: optical microscopy (spherulitic structure, matrix proportions, impact characteristics), SEM-BSE imaging (mineral contrast, impact deformation, metal-sulfide distribution), EDS semi-quantitative analysis (conditions: ≥1 nA, 60–120 s / point, ≥30-point stratified sampling, standard regression and periodic drift calibration), and automatic identification of metal / sulfide / olivine / pyroxene phase domains.

[0047] This module provides mineral chemical characteristics, texture characteristics, and mineral phase separation information.

[0048] 4) Intelligent discrimination and adaptive processing module

[0049] Constructing a multidimensional feature model and dynamic discrimination system for meteorites: weathering grade identification (W-aware), feature threshold adaptive model (Mg / Si, Fe / Si, EDS-Fa / Fs regression), uncertainty assessment and sampling mechanism (p<0.8 triggering verification), and EDS→WDS bias correction model.

[0050] This module implements an intelligent classification mechanism that is interpretable, correctable, and self-learning.

[0051] 5) Verification and Quality Control Module

[0052] Used for critical sample validation and system stability maintenance: WDS precise sampling, sampling optimization strategy (minimum cost criterion), strategy iteration and dynamic threshold update.

[0053] (1) Fusion of multimodal meteoritic features

[0054] By comprehensively collecting multi-source data such as optical petrography (spherulitic structure, matrix ratio), SEM-BSE microtexture (weathering degree, impact melt microstructure), EDS mineral chemistry (olivine Fo, pyroxene Fs estimation and Fe / Mg ratio), VNIR / Raman spectroscopy (layered silicates, hydration characteristics, CAIs identification), magnetic susceptibility and bulk density, a four-dimensional evidence system of composition, structure, minerals and physical properties is constructed.

[0055] It can effectively identify: H / L / LL group iron-nickel-olivine-pyroxene system; EH / EL high sulfide-metal content characteristics; CM / CO / CV carbonaceous matrix, CAI content and hydration zone characteristics; impact veins, melt inclusions, and agglutinate structures.

[0056] (2) Low-loss sample preparation and adaptive sample state

[0057] Automatic selection based on meteorite sample metal content, matrix brittleness, and fracture density: ethanol lubrication for cutting to reduce iron-nickel oxidation and sulfide degradation; vacuum epoxy embedding to improve the fixation of carbonaceous matrix and fine-grained spherulites; progressive control of load and grain size to prevent pull-out. Thickness 20–35 µm. Adaptive preservation of shock veins and plagioclase twins.

[0058] Ensure that the key mineral structure and impact microstructure are not damaged.

[0059] (3) Weathering grade perception and threshold self-correction

[0060] The weathering grade (W0–W6) and shock stage (S1–S6) are input into the classification model. The elemental composition Mg / Si, Fe / Si, metal volume fraction and EDS regression olivine Fo value-pyroxene Fs value criteria are dynamically adjusted based on the degree of weathering, so that the system can maintain discrimination stability under iron oxidation, sulfide alteration and silicate edge alteration conditions.

[0061] (4) Uncertainty-driven verification closed loop

[0062] For each sample, the posterior probability p and confidence interval are output; for samples with low confidence, the WDS sampling and model iteration update mechanism is activated to achieve a closed loop of "rapid prediction → key verification → adaptive learning".

[0063] (5) Standardized multidimensional output

[0064] The system automatically outputs: Group (chemical group, H / L / LL / EH / EL / CM / CO / CV), Type (lithological type and / or degree of thermal metamorphism), Shock stage (S level), and Weathering grade (W level).

[0065] It supports scientific cataloging, database entry, and sample priority sorting.

[0066] This invention significantly improves the ability to rapidly identify extraterrestrial samples, allocate them for scientific research, and protect resources. It is applicable to various scenarios such as Antarctic meteorite recovery missions, planetary exploration sample laboratories, lunar soil and meteorite storage management, and scientific research and museum identification systems.

[0067] The system comprises an optical microscope, SEM-BSE, and EDS unit, forming a multimodal acquisition module. The slide preparation module includes a particle fixation device, an adjustable load polishing system, and a coolant management system. The WDS module performs high-precision verification on samples with insufficient classification confidence. The intelligent discrimination module includes an algorithm for processing weathering-sensitive features. The system features end-to-end sample tracking, data archiving, and label management. The model is applicable to various meteorite types, including common chondrites (H, L, LL), enstatite meteorites (EH, EL), and carbonaceous chondrites (CM, CO, CV).

[0068] like Figure 2 As shown, the method of low-loss multimodal adaptive rapid meteorite classification and verification system is outlined in the following steps:

[0069] (1) Multimodal pre-screening: Obtain the magnetic susceptibility, bulk density, visible-near infrared (VNIR) spectrum, microscopic appearance and optional micro-CT structure information of the target meteorite sample, which are used to identify metal content, carbonaceous matrix, degree of fracture development and potential spherule structure.

[0070] (2) Low-loss preparation: Based on the pre-screening information, select ethanol or ethanol-water cooling cutting method, vacuum epoxy holding method and load adaptive grinding and polishing strategy to prepare thin films with a thickness of 20–35 µm, so that the metal phase, sulfide phase and fine matrix maintain the in-situ structure.

[0071] (3) Multimodal signal acquisition: The thin section was observed by optical microscope and the texture features were obtained by scanning electron microscopy (BSE). Energy dispersive spectroscopy (EDS) point analysis and regional analysis were performed to extract mineral phase composition information, metal volume fraction and Fe / Mg ratio of olivine and pyroxene.

[0072] (4) Weathering grade perception threshold adaptive classification: Weathering grade (W0–W6) and impact grade (S1–S6) are used as independent variables to input the classification model. Combined with mineral chemical ratio, physical property characteristics and structural characteristics (such as spheroid morphology, impact vein structure, metal-sulfide residual morphology), the discrimination weight and confidence interval are dynamically adjusted, including but not limited to weathering sensitive parameters Mg / Si, Fe / Si ratio and EDS regression Fo-Fs value;

[0073] (5) Uncertainty-driven verification and iteration: Based on the classification output confidence level p, when p is less than the set threshold (e.g., 0.8), the electron probe WDS precise analysis is triggered for verification, and the classification parameters and calibration curve are updated according to the verification results;

[0074] (6) Five-dimensional classification results output: Output the meteorite chemical group (Group), petrological type (Type), impact level (S), weathering grade (W) and confidence index of the sample to form a multi-dimensional discrimination result with uncertainty evaluation.

[0075] The pre-screening module includes a magnetic susceptibility tester, a density measuring device, a portable VNIR spectrometer, a microscope, and optional micro-CT scanning equipment. Ethanol coolant concentration is 70–100 wt% to reduce the risk of metal oxidation and sulfide hydration. Slice thickness is 20–35 µm, adaptively adjusted based on sample fracture, carbonaceous content, and metallic phase content. Loose samples are vacuum-embedded in epoxy resin to enhance mechanical stability and prevent particle shedding during polishing. At least 30 EDS sampling points are used, including background, mineral phase interfaces, and typical particles. EDS automatically performs drift correction or standard regression every 60–120 seconds. Standard material calibration is performed every 10 EDS sampling points. Classification models include mineral ratio models, characteristic parameter models, and EDS-estimated Fo-Fs regression models.

[0076] The model input parameters include weathering grade W and impact grade S. Outputs include confidence level p, confidence interval, and verification suggestions. A WDS analysis queue is automatically triggered when p < 0.8. The classification model supports online parameter updates.

[0077] Classification process

[0078] Step 1: Pre-screening (completely low loss)

[0079] Measuring magnetic susceptibility and density → Rapid spectral identification of layered silicate / CAI / metal-enriched phases → Adopting different fabrication strategies based on characteristics

[0080] Step 2: Low-loss film preparation

[0081] Metal-enriched samples → Ethanol cooling + Anhydrous environment

[0082] Loose carbonaceous samples → Low loading, resin embedding, fine-grained grinding discs

[0083] Impact fracture sample → Thickness ≥ 32 μm

[0084] Step 3: Multimodal Analysis

[0085] Optical + BSE Form

[0086] ≥30-point stratified EDS analysis and standard sample regression

[0087] Automatic extraction of spheroidal features and metal volume fraction

[0088] Step 4: Intelligent discrimination; the model fusion information is shown in Table 1:

[0089] Table 1 shows the model fusion information.

[0090]

[0091] Output: Group + Type + S + W + Confidence level p.

[0092] Step 5: Adaptive verification, p<0.8 → Automatically add to EPMA verification queue.

[0093] Example:

[0094] Taking 100 meteorite samples recovered from the Antarctic region as an example, the samples cover a variety of types, including ordinary chondrites (H, L, LL), enstatite meteorites (EH, EL) and carbonaceous chondrites (CM, CO, CV), and include samples with different weathering grades (W0–W5) and impact metamorphism grades (S1–S6).

[0095] Steps and results:

[0096] 1. Multimodal pre-screening stage

[0097] The magnetic susceptibility, volume density, VNIR spectrum and microscopic features of all 100 samples were tested.

[0098] Based on information such as metal content, matrix characteristics, and weak absorption bands to identify potential carbonaceous components (CM / CO), 92 samples were selected to enter the thin section preparation stage, while the remaining 8 samples, due to their clear characteristics or obvious fragment morphology, entered the direct characterization and database recording process.

[0099] 2. Preparation of low-loss thin films

[0100] For metal-rich samples, ethanol cooling and cutting were used; for carbonaceous fine-grained matrix samples, vacuum epoxy resin embedding and low-load polishing were used; for samples with developed cracks, a thickness of ≥32 μm was set to preserve impact pulses and stress structures.

[0101] The success rate of adaptive grinding exceeded 95%, and the final thin section had an average thickness of 23–33 μm. Key microstructures (spheroidal boundaries, metal-sulfide phases, and glass veins) were clearly preserved, and no significant "mineral uplift" or secondary crack propagation was observed.

[0102] 3. Multimodal data acquisition

[0103] Optical petrography, SEM-BSE and EDS analysis were performed on each sample, and no less than 30 EDS points were collected, covering olivine, pyroxene, matrix, metal and sulfide regions. Standard regression and periodic drift correction were used to improve the semi-quantitative stability.

[0104] 4. Classification and Adaptive Optimization

[0105] The system extracts features such as Mg / Si ratio, Fe / Si ratio, Fe-Mg estimation of olivine / pyroxene (Fo-Fs), metal volume fraction, spectral characteristics of carbonaceous components, and spheroidal structure scale parameters, and inputs them into the weathering adaptive model.

[0106] Thirteen samples were automatically added to the EPMA-WDS sampling queue due to their classification confidence level p < 0.8, for use in validating the model output and updating the threshold. The model completed one cycle of parameter self-calibration and threshold optimization, improving its ability to distinguish highly weathered (W≥3) samples.

[0107] Algorithm description:

[0108] Inputs: Optical / BSE, EDS>=30, VNIR / Raman, κ, ρ; Meta: S, W

[0109] Params: tau=0.80, T_c(W), Calib_EDS2WDS, VerifyQueue

[0110] CLASSIFY(sample):

[0111] 1) Selection of pre-screening and slide preparation scheme (ethanol / epoxy / thickness 20–35μm)

[0112] 2) Collection and calibration (periodic reference materials → updating Calib_EDS2WDS)

[0113] 3) Feature construction X = {composition, structure, spectrum, physical properties}

[0114] 4) W-aware correction and threshold selection: thresholds = T_c(W_hat)

[0115] 5) Probabilistic classification p_vec → (label, p)

[0116] 6) If p ≥ tau: record and output; otherwise, enter the WDS verification queue.

[0117] WDS_VERIFY_AND_UPDATE():

[0118] Perform WDS on the queue samples; update Calib_EDS2WDS and T_c(W); retrain the lightweight model.

[0119] 5. Secondary Output and Recording

[0120] Based on the final classification results and confidence levels, output records with S-level, W-level, Group, and Type are generated for database standardization and storage, as shown in Table 2.

[0121] Table 2 shows examples of output templates.

[0122]

[0123] (The S and W grades in the table are determined according to the meteoritic standard system)

[0124] Sample A01 exhibits a typical LL-type olivine-low-iron pyroxene assemblage with a clear spheroidal structure, low metal content, and well-developed local impact veins; Sample A18 shows a carbonaceous matrix dominance, obvious Raman hydration characteristic absorption bands, low CAI content, and significant weathering. Therefore, further EPMA verification is needed to eliminate the influence of high-level alteration on the EDS results.

[0125] This invention can be deployed in meteorite classification institutions, meteorite collection institutions, planetary science and extraterrestrial sample research institutions, universities and research institutes, museums and collection appraisal laboratories, etc., and is particularly suitable for large-scale Antarctic meteorite recovery missions, planetary exploration sample return programs (lunar, Martian, asteroid samples), and existing and planned meteorite database systems. Through low-destructive preparation, multimodal information acquisition, adaptive intelligent discrimination, and uncertainty verification mechanisms, this invention shortens the cycle time, standardizes, and automates the rapid meteorite classification process, significantly improving sample processing throughput and data reliability. It is suitable for scientific research analysis, public sample platform operation, and high-value sample protection and management, and has broad application value and promising prospects.

[0126] By employing pre-screening criteria such as magnetic susceptibility, volume density, spectral characteristics, and microstructure, combined with low-destructive thin-section preparation techniques such as ethanol cooling and vacuum embedding, the system acquires multimodal data including optical microscopy, SEM-BSE, EDS mineral chemistry, and texture features. The system utilizes a weathering grade sensing threshold adaptive model to jointly analyze the Fe-Mg composition of olivine and pyroxene, metal and sulfide content, chondrite structure retention, and carbonaceous matrix characteristics, enabling rapid identification of typical meteorites such as ordinary chondrites (H, L, LL), enstatite meteorites (EH, EL), and carbonaceous chondrites (CM, CO, CV). An uncertainty-driven sampling mechanism triggers WDS verification when confidence levels decrease, achieving automatic bias correction and accuracy maintenance. The system ultimately outputs the meteorite's chemical group (Group), rock type (Type), impact metamorphic grade (S), weathering grade (W), and classification confidence level, achieving high-throughput, low-damage, and traceable intelligent meteorite identification. This invention significantly shortens the meteorite classification cycle, improves standardization and reliability, and is suitable for primary scientific cataloging and research in Antarctic meteorite recovery and other extraterrestrial rock and sample return missions.

[0127] Matters not covered in this invention are common knowledge.

[0128] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A low-loss, multimodal, adaptive rapid meteorite classification and verification system, characterized by: The system includes a multimodal pre-screening module, a low-destructive slide preparation module, an EDS data acquisition and feature extraction module, an optical microscopy and scanning electron microscopy analysis module, a weathering grade sensing and adaptive discrimination module, an uncertainty-driven verification module, and a classification result output and sample database management module. The multimodal pre-screening module is connected to the low-destructive slide preparation module, used for non-destructive evaluation of sample composition, structure, and preservation status, determining slide preparation strategies and subsequent analysis routes. The low-destructive slide preparation module is connected to the EDS data acquisition and feature extraction module, used for analyzing metals, sulfides, and brittle materials in meteorites. Due to the susceptibility of carbonaceous matrices to oxidation, shedding, and pyrolysis, the optical microscopy and scanning electron microscopy analysis module is connected to the EDS data acquisition and feature extraction module. The optical microscopy and scanning electron microscopy analysis module is used to obtain whole-rock, microscopic, and trace component evidence. The EDS data acquisition and feature extraction module is connected to the weathering grade perception and adaptive discrimination module. The weathering grade perception and adaptive discrimination module is used to construct a multidimensional feature model and dynamic discrimination system for meteorites. Both the weathering grade perception and adaptive discrimination module and the uncertainty-driven verification module are connected to the classification result output and sample database management module.

2. The low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 1, characterized in that: The multimodal pre-screening module includes a handheld magnetic susceptibility meter, a bulk density measurement device, a micro-CT, a miniature microscope, and a Raman / visible-near-infrared spectroscopy device. It identifies layered silicates, hydrous minerals, CAI enrichment areas, carbonaceous matrix, and hydration absorption bands, and obtains the magnetic susceptibility, bulk density, visible-near-infrared spectrum, microscopic appearance, and optional micro-CT structural information of the target meteorite sample. This information is used to identify metal content, carbonaceous matrix, degree of fracture development, and potential spherulitic structure, enabling preliminary classification of sample physical properties and mineral characteristics, as well as selection of slide preparation path.

3. The low-loss multimodal adaptive rapid meteorite classification and verification system and method according to claim 1, characterized in that: The low-loss film preparation module selects ethanol or ethanol-water cooling cutting method, vacuum epoxy holding method, and load-adaptive polishing strategy based on pre-screening information to prepare films with a thickness of 20–35 µm. This preserves the in-situ structure of the metal phase, sulfide phase, and fine-grained matrix. Ethanol / ethanol-water lubrication and cooling cutting is used to reduce iron-nickel phase oxidation and Fe migration. Vacuum epoxy holding and low-load polishing system protect the fine-grained and brittle mineral structure. The adaptive thickness control program is 20–35 µm to preserve impact veins, molten inclusions, spheroidal contours, and glass structure.

4. The low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 1, characterized in that: The optical microscopy and scanning electron microscopy analysis module performs optical microscopy observation and scanning electron microscopy reflection electron imaging to obtain texture features. It also performs energy dispersive spectroscopy point analysis and region analysis to extract mineral phase composition information, metal volume fraction and Fe / Mg ratio estimation of olivine and pyroxene. It automatically identifies metal, sulfide, olivine or pyroxene phase domains and provides mineral chemical characteristics, texture features and mineral phase separation information.

5. The low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 1, characterized in that: The weathering grade perception adaptive discrimination module takes weathering grade and impact classification as independent variables and inputs them into the classification model. It combines mineral chemical ratios, physical properties and structural features to dynamically adjust the discrimination weights and confidence intervals, including but not limited to weathering sensitive parameters Mg / Si and Fe / Si ratios and EDS regression Fo-Fs values.

6. The low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 1, characterized in that: The uncertainty-driven verification module is used to output a confidence level p based on the classification. When p is less than a set threshold, it triggers an electron probe microanalysis (WDS) for precise verification and updates the classification parameters and calibration curves based on the verification results. The classification result output and sample database management module is used to output the meteorite chemical group, petrological type, impact level, weathering grade, and confidence index of the sample, forming a multi-dimensional discrimination result with uncertainty evaluation.

7. A method for a low-loss, multimodal adaptive rapid meteorite classification and verification system, characterized in that, The method includes the following steps: Step 1: Pre-screening using several modes to obtain information on the magnetic susceptibility, bulk density, visible-near-infrared spectrum, microscopic appearance, and optional micro-CT structure of the target meteorite sample, which is used to identify metal content, carbonaceous matrix, degree of fracture development, and potential spherulitic structure. Step 2: Low-loss film preparation. Based on the pre-screening information, select ethanol or ethanol-water cooling cutting method, vacuum epoxy holding method and load adaptive grinding and polishing strategy to prepare thin films with a thickness of 20–35 µm, so that the metal phase, sulfide phase and fine matrix can maintain the in-situ structure. Step 3: Acquire several modal signals, observe the thin section with an optical microscope, obtain texture features by scanning electron microscopy reflection electron imaging, and perform energy dispersive spectroscopy point analysis and region analysis to extract mineral phase composition information, metal volume fraction and estimate the Fe / Mg ratio of olivine and pyroxene; Step 4: Adaptive classification based on weathering grade perception threshold. Weathering grade and impact classification are input into the classification model as independent variables. The discrimination weights and confidence intervals are dynamically adjusted by combining mineral chemical ratios, physical properties and structural features, including but not limited to weathering sensitive parameters Mg / Si and Fe / Si ratios and EDS regression Fo-Fs values. Step 5: Uncertainty-driven verification and iteration. Output confidence level p based on classification. When p is less than the set threshold, trigger precise electron probe microanalysis (WDS) for verification, and update classification parameters and calibration curves based on the verification results. Step 6: Output the five-dimensional classification results, including the meteorite chemical group to which the sample belongs, petrological type, impact level, weathering grade, and confidence index, forming a multi-dimensional discrimination result with uncertainty evaluation.

8. The method of the low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 7, characterized in that: In step 2, the concentration of ethanol coolant is 70–100 wt% to reduce the risk of metal oxidation and sulfide hydration. In step 3, the thickness of the sheet is 20–35 µm, which is adaptively adjusted according to the sample cracks, carbon content and metal phase content.

9. The method of the low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 7, characterized in that: In step 1, the target meteorite sample is encapsulated in vacuum epoxy resin to enhance mechanical stability and prevent particles from being pulled out during polishing. In step 3, there are no fewer than 30 sampling points, including the background, mineral phase interface and typical particles, and drift correction or standard sample regression is automatically performed every 60–120 seconds.

10. The method of the low-loss multimodal adaptive rapid meteorite classification and verification system according to claim 7, characterized in that: In step 4, the adaptive classification includes a mineral ratio model, a feature parameter model, and an EDS-estimated Fo-Fs regression model.

Images