Method for tracing copper concentrate samples based on X-ray fluorescence spectroscopy and isotope detection

Abstract

The copper concentrate sample traceability method based on X-ray fluorescence spectroscopy and isotope detection proposed in this application relates to the technical field of copper concentrate traceability. Data sampling is carried out on subjective and objective factors during the sampling process of copper concentrate samples, and the data is saved based on the blockchain to improve the security and accuracy of the data. The data analysis results of copper concentrate samples are subjected to discretization, standardization, outlier correction, and principal component analysis processing, thereby ensuring the standardization of the data. Based on the first mining area characteristic fingerprint library set obtained by historical data mining, the target mining area to which the copper concentrate sample belongs and the corresponding first similarity are determined. Through the technical solution of this application, the standardization processing of the copper concentrate analysis results can be realized, and then the consistent rules can be extracted, and the confidence probability of the traceability result can be given in combination with the sampling process data, so that the obtained traceability result is more accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of mineral product traceability, and particularly relates to a method for tracing copper concentrate samples based on X-ray fluorescence spectroscopy and isotope detection. Background Art

[0002] In recent years, with the continuous expansion of China's foreign trade, the import volume of imported mineral product raw materials has been continuously increasing. Among them, copper concentrate, as an important industrial raw material, the issues of its quality and origin traceability have become increasingly prominent. There are many technical problems to be solved urgently in the field of copper concentrate traceability in the existing technology: First, the origin traceability of copper concentrate mainly relies on the declaration of exporters, lacking effective technical means for verification. Especially when there are deviations in samples or malicious concealment of declarations, it is difficult to accurately trace its true source; Second, copper concentrate is usually transported in bulk, and there are untraceable factors during the sample transfer process, resulting in difficulties in identifying the origin of the samples; In addition, the existing traceability technical means mostly focus on elemental analysis, such as X-ray fluorescence spectrometry. Although it can measure the elemental composition, the limitations of a single technical means are relatively large, and it is difficult to comprehensively reflect the origin characteristics of the samples. Although certain progress has been made in the origin traceability technology in the fields of animal and plant products, agricultural products, etc. at home and abroad, its application in the field of mineral products such as copper concentrate is still in its initial stage, lacking a systematic traceability model and technical means. Therefore, developing a method for tracing the origin of copper concentrate that combines multiple analysis techniques (such as X-ray fluorescence spectroscopy and isotope detection) has important practical significance and application value. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for tracing copper concentrate samples based on X-ray fluorescence spectroscopy and isotope detection to solve the above-mentioned technical problems existing in the existing technology.

[0004] The present application proposes a method for tracing copper concentrate samples based on X-ray fluorescence spectroscopy and isotope detection, and the method includes:

[0005] S0: Obtain the first multi-source data of the copper concentrate sample, generate a first credibility value for each of the first multi-source data, and generate first block data according to the first multi-source data and the first credibility value and store it on the chain;

[0006] S1: Analyze the copper concentrate sample to obtain a plurality of first analysis results, and perform preprocessing and discretization on each of the first analysis results to obtain a first transaction database;

[0007] S2: Extract a plurality of first data items from the first transaction database, and determine a target mining area characteristic fingerprint library and a first similarity from the plurality of first data items according to the first mining area characteristic fingerprint library set; wherein, the first mining area characteristic fingerprint library is obtained by analyzing historical data of a plurality of different copper concentrates.

[0008] S3: Determine a first target mining area and a first confidence probability corresponding to the copper concentrate sample according to the first block data and the first similarity.

[0009] Preferably, the first multi-source data in S0 is obtained through the following steps:

[0010] S01: Obtain first subjective sampling data and first objective sampling data of the copper concentrate sample;

[0011] S02: Generate the first multi-source data according to the first subjective sampling data and the first objective sampling data.

[0012] Preferably, the first subjective sampling data includes the accumulative working hours of the sampler and the number of historical errors, and the first objective sampling data includes sampling position point information and the static time of the sample before sampling.

[0013] Preferably, S0 further includes the following steps:

[0014] S03: Input the first multi-source data into a first credibility determination model to obtain a first credibility corresponding to the first multi-source data;

[0015] S04: Obtain first block data according to the first multi-source data and the first credibility, and store the first block data on the chain.

[0016] Preferably, S1 includes the following steps:

[0017] S11: Store the first analysis result in a table form to obtain a first data table;

[0018] S12: Perform normalization processing on each first data value stored in the first data table to obtain a first normalized data table; wherein, each first analysis result is composed of multiple first data values;

[0019] S13: Perform abnormal correction processing on the first normalized data table to obtain a second normalized data table;

[0020] S14: Perform principal component analysis and dimensionality reduction processing on the second normalized data table to obtain a third normalized data table;

[0021] S15: Perform discretization processing on the third normalized data table to obtain a first transaction database.

[0022] Preferably, the first analysis result includes XRF element content and lead isotope ratio.

[0023] Preferably, the principal component analysis and dimensionality reduction processing in S14 include:

[0024] Perform dimensionality reduction on the standardized isotope ratios to generate principal component features to replace the original features;

[0025] Replace the original features in the second standardized data table according to the generated principal component features, so as to obtain the third standardized data table.

[0026] Preferably, in S0, the first data item includes a data combination of an element interval and an isotope interval.

[0027] Preferably, in S2, the first mine area characteristic fingerprint library set is obtained by analyzing historical data of multiple different copper concentrates, and specifically includes the following steps:

[0028] Taking different copper concentrates as units, divide the sample data into multiple groups; among them, each sample data includes an element interval, an isotope interval, and a mine area label;

[0029] For multiple groups of sample data, respectively obtain multiple first mine area characteristic fingerprint libraries, and obtain the first mine area characteristic fingerprint library set according to the multiple first mine area characteristic fingerprint libraries.

[0030] Preferably, the first mine area characteristic fingerprint library includes:

[0031] Multiple frequent item sets composed of element contents and isotope intervals in specified copper concentrates.

[0032] The copper concentrate sample traceability method based on X-ray fluorescence spectroscopy and isotope detection proposed in this application relates to the technical field of copper concentrate traceability. Data sampling is carried out on subjective and objective factors in the sampling process of copper concentrate samples, and data is saved based on the blockchain to improve the security and accuracy of data. The data analysis results of copper concentrate samples are discretized, standardized, outlier corrected, and principal component analysis is performed, thus ensuring data standardization. Based on the first mine area characteristic fingerprint library set obtained by historical data mining, the target mine area to which the copper concentrate sample belongs and the corresponding first similarity are determined. Through the technical solution of this application, standardized processing of copper concentrate analysis results can be realized, and then consistent rules can be extracted, and confidence probabilities are given to the traceability results in combination with sampling process data, so that the obtained traceability results are more accurate. Description of the Drawings

[0033] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only exemplary. For those of ordinary skill in the art, without creative efforts, other implementation drawings can be obtained based on the provided drawings.

[0034] Figure 1 It is the execution flowchart of the copper concentrate sample traceability method based on X-ray fluorescence spectroscopy and isotope detection in the present invention.

[0035] Figure 2 It is the determination flowchart of the first transaction database in step S1 of the present invention. Specific Embodiments

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the protection scope of the present invention.

[0037] The following will detail the present invention in combination with the accompanying drawings and specific embodiments, where the illustrative embodiments and explanations are only used to explain the present invention, but not to limit the present invention.

[0038] The following will detail the copper concentrate sample traceability method based on X-ray fluorescence spectroscopy and isotope detection of the present invention, specifically as Figure 1 shown.

[0039] S0: Obtain the first multi-source data of the copper concentrate sample, generate a first credibility value for each of the first multi-source data, and generate first block data based on the first multi-source data and the first credibility value and store it on the chain.

[0040] The accuracy of the sampling data of copper concentrate samples will directly affect the accuracy of subsequent data analysis and traceability results. During the traceability process of copper concentrate, the behaviors that may lead to inaccurate sampling due to human factors mainly include the following aspects: 1. Sampling deviation: Inhomogeneity problem: If the copper concentrate material itself is inhomogeneous (such as containing different proportions of metals and other impurities), and this is not taken into account during sampling, it may result in the sample not being able to represent the characteristics of the entire batch. 2. Improper sampling location: Selecting the wrong sampling point or depth, and failing to fully cover the entire material pile, may overlook the compositional changes in certain areas. 3. Improper use of tools: Using sampling tools that are not suitable for the current material characteristics. For example, for materials with large particles or high viscosity, specially designed samplers should be used instead of ordinary shovels. 4. The tools are not cleaned properly, and the residues from the previous sampling will affect the purity of the current sample. 5. Improper time management: Sampling at the wrong time. For example, sampling when the material has just arrived at the port and is not yet stable may lead to distorted results due to the rapid evaporation of surface moisture. 6. Operational errors: Human errors during the sample processing, including but not limited to inaccurate weighing, sample loss, uneven mixing, etc. 7. Not following the standard operating procedures (SOP), such as not achieving the required particle size during sample grinding, or randomly discarding small particles and debris during sieving. 8. Recording and marking errors: unclear recording or incorrect marking of sample information may lead to sample confusion and affect the accuracy of subsequent analysis results. 9. Data entry errors, especially transcription errors that are likely to occur when manually recording data. 10. Deliberate behavior: Out of economic interests, deliberately tampering with samples or forging test reports to cover up the true situation of low-quality products. 11. Lack of training: If the staff lacks the necessary professional training and does not understand the correct sampling techniques and procedures, even with perfect equipment and technical support, it is impossible to ensure the accuracy of sampling.

[0041] Among the above reasons, there are various factors including the distribution of sampling points of copper concentrate samples, the static time, and the work experience of samplers. In order to make the data analysis and traceability results of copper concentrate more accurate, in this step, a credible probability value is first given for the credibility of the sampling samples, so as to make the subsequent big data analysis more accurate and targeted.

[0042] The S0 includes the following sub-steps:

[0043] S01: Obtain the first subjective sampling data and the first objective sampling data of the copper concentrate sample.

[0044] The first subjective sampling data includes information such as the cumulative working hours of the sampler and the historical error times. The first objective sampling data may include information such as the sampling position points when the sampler conducts data sampling and the static time of the sample before sampling. The sampling position point information includes a plurality of first coordinate values, and each of the first coordinate values is a three-dimensional coordinate value of each sampling point in a coordinate system with the geometric center of the copper concentrate sample as the origin.

[0045] S02: Generate the first multi-source data according to the first subjective sampling data and the first objective sampling data.

[0046] Preferably, the first multi-source data is in the form of a multi-dimensional vector, which contains all the information in the first subjective sampling data and the first objective sampling data.

[0047] S03: Input the first multi-source data into the first credibility determination model to obtain the first credibility corresponding to the first multi-source data.

[0048] The first credibility determination model is obtained by training through CNN. During the training process, the sampling data of historical copper concentrate samples is used as the input data of the model, and the corresponding credibility is used as the output data to train and obtain the first credibility determination model.

[0049] S04: Obtain the first block data according to the first multi-source data and the first credibility, and store the first block data on the chain.

[0050] To ensure the accuracy and security of the sampling information, the first block data is stored in a blockchain manner. Specifically, the first block data is consensus in the blockchain and stored in the blockchain ledger after successful consensus to prevent data from being tampered with.

[0051] S1: Analyze the copper concentrate sample to obtain a plurality of first analysis results, and perform preprocessing and discretization on each of the first analysis results to obtain the first transaction database.

[0052] The first analysis result is the XRF element content (such as Cu, Fe, Zn) and lead isotope ratio (206Pb / 204Pb, 207Pb / 204Pb) obtained by analyzing the copper concentrate sample. Since there are multiple copper concentrate samples, multiple first analysis results can be obtained. The preprocessing and discretization process includes data standardization and binning. Specifically, Z-score standardization is performed on the XRF element content (such as Cu, Fe, Zn) and lead isotope ratio (206Pb / 204Pb, 207Pb / 204Pb) to eliminate the dimension difference.

[0053] Such asFigure 2 As shown, S1 includes the following sub-steps:

[0054] S11: Store the first analysis result in tabular form to obtain a first data table.

[0055] The first multi-source data includes XRF element content and lead isotope ratio information. The data sources are as follows:

[0056] XRF element content: The percentage content of Cu, Fe, and Zn in the copper concentrate sample (such as Cu = 16.85%, Fe = 16.35%, etc.).

[0057] Lead isotope ratio: The measured values of 206Pb / 204Pb and 207Pb / 204Pb (such as 206Pb / 204Pb = 18.656975). Original data format:

[0058] Stored in tabular form, each row represents a sample, and the columns include fields such as Cu(%), Fe(%), Zn(%), 206Pb / 204Pb, 207Pb / 204Pb, etc.

[0059] S12: Perform normalization processing on each first data value stored in the first data table to obtain a first normalized data table.

[0060] Before performing the normalization processing, it is necessary to first calculate the mean and standard deviation for each feature (such as Cu content Cu i ). The calculation formulas for the mean and standard deviation are as follows:

[0061] Mean, where N is the number of samples:

[0062] ;

[0063] Standard deviation:

[0064] ;

[0065] After calculating the mean and standard deviation, apply the Z-score normalization formula to obtain the normalized value of each first data value in the first data table. Taking the Cu content as an example, its normalization formula is as follows:

[0066] ;

[0067] where Cu i is the copper content, μCu is the mean of the copper content, and σCu is the standard deviation of the copper content.

[0068] After that, for each of the first data values in the first data table, they are replaced with the standardized data values, so that the first standardized data table can be obtained.

[0069] S13: Perform anomaly correction processing on the first standardized data table to obtain a second standardized data table.

[0070] In the first standardized data table, some data values may be abnormal. Therefore, it is necessary to detect the abnormal values and eliminate or correct the identified abnormal values.

[0071] The method for detecting abnormal values is as follows:

[0072] Box plot analysis: Check the lower and upper quartiles (Q1 = 25%, Q3 = 75%) of each feature, and samples outside the range of Q1 - 1.5I QR , Q3 + 1.5I QR are regarded as abnormal values (I QR = Q3 - Q1).

[0073] Processing method: If the abnormal value is caused by measurement error, directly eliminate it; if it is real data, use robust standardization (such as median and interquartile range instead of mean standard deviation).

[0074] S14: Perform principal component analysis and dimensionality reduction processing on the second standardized data table to obtain a third standardized data table.

[0075] Among the multiple data values in the second standardized data table, the lead isotope ratios (such as 206Pb / 204Pb and 207Pb / 204Pb) may be highly correlated. To avoid the problem of multicollinearity, principal component analysis needs to be performed on each type of data value in the second standardized data table, and dimensionality reduction processing is carried out on the data types with high collinearity.

[0076] The solution is as follows:

[0077] Principal component analysis (PCA): Reduce the dimension of the standardized isotope ratios to generate principal components to replace the original features. Replace the original features in the database table according to the generated principal component features, so as to obtain the third standardized data table.

[0078] S15: Perform discretization processing on the third standardized data table to obtain the first transaction database.

[0079] For facilitating subsequent association rule mining, in this step, continuous data needs to be divided into intervals (for example, Cu content: 10%-15% is "low", 15%-20% is "medium", 20%+ is "high"), and converted into categorical variables, that is, discretized according to preset rules.

[0080] Feature labeling: Construct a "element interval - isotope interval" transaction database. That is, for the discretized third standardized data table, each row is used as a unit to form a data item, and thus the first transaction database is obtained according to the determined multiple data items.

[0081] The first transaction database is used to represent the association relationship between the element interval and the isotope interval.

[0082] S2: Extract multiple first data items from the first transaction database, and determine the target mining area feature fingerprint library and the first similarity from the multiple first data items according to the first mining area feature fingerprint library set.

[0083] The first data item can be each row of data in the transaction database, that is, it includes a data combination of an element interval and an isotope interval. In subsequent steps, in order to determine whether there is an association relationship between the data combination and a specific mining area, each first data item needs to be input into the first frequent item set determination model one by one, so as to mine the matching degree between each data combination and a specific mining area through big data.

[0084] The first mining area feature fingerprint library is obtained by analyzing historical data of multiple different copper concentrates, specifically as follows:

[0085] Taking different copper concentrates as units, the sample data is divided into multiple groups, that is, the sample data in each group comes from the same copper concentrate. In each group, each sample data contains an element interval, an isotope interval, and a mining area label. In each sample, the element interval, the isotope interval, and the mining area label are all discretized values.

[0086] Set a dynamic support degree adjustment strategy: Since the historical data samples of copper concentrates often have the problem of insufficient quantity, for small quantity samples (such as 24), set a lower support degree (such as 10%) and introduce a weighted support degree, and preferentially mine highly discriminatory combinations (such as "high Cu content + specific Pb isotope ratio").

[0087] The calculation method of the weighted support degree is as follows:

[0088] Discriminative weight: Calculate the weight based on the Chi-square value between the item set and the mining area. The calculation method of the Chi-square value is as follows: Select multiple sample data from the same copper concentrate, square the difference between the actual value and the expected value of each sample data, then divide by the expected value, and finally sum up all the results to obtain the Chi-square value.

[0089] The calculation formula of weighted support is as follows:

[0090] Weighted support = Original support × (1 + Chi-square value / Chi-square threshold);

[0091] Prioritize mining high-weight item sets: For example, if the Chi-square value of the item set {Cu = high, 206Pb / 204Pb = 18.6~18.7} is 15 (the Chi-square threshold is 3.84), its weighted support is increased to 4 times the original value.

[0092] Rule screening and confidence optimization:

[0093] Screen rules significantly related to the mining area through Chi-square test or Lift (such as: {Cu = high, 206Pb / 204Pb = 18.6~18.7} → COLLAHUASI (confidence 95%)).

[0094] Retain rules with confidence > 90% to form the first mining area feature fingerprint database.

[0095] Thus, for multiple groups of sample data, multiple such first mining area feature fingerprint databases are obtained respectively, and the first mining area feature fingerprint database set is obtained based on the multiple first mining area feature fingerprint databases.

[0096] Thus, in subsequent steps, based on the first similarity between the first transaction database and each first mining area feature fingerprint database in the first mining area feature fingerprint database set, the first mining area feature fingerprint database with the highest similarity to the copper concentrate sample can be determined and used as the target mining area feature fingerprint database.

[0097] Among them, the first similarity is calculated as follows: For all data items in the first transaction database, calculate the second similarity with each first mining area feature fingerprint database, and take the average value of all the second similarities as the first similarity.

[0098] S3: Determine the first target mining area and the first confidence probability according to the first block data and the first similarity.

[0099] In the target mining area characteristic fingerprint database, the copper concentrate mining area corresponding to the target mining area characteristic fingerprint database is indicated. Therefore, the determined copper concentrate mining area can be used as the first target mining area corresponding to the copper concentrate sample.

[0100] Furthermore, in order to determine the first confidence probability that the copper concentrate sample originates from the first target mining area, based on a machine learning model, the first confidence probability can also be predicted according to the sampling-related information characterized in the first block data and the first similarity. When training the machine learning model for predicting the first confidence probability, historical data of other copper concentrates can be selected as samples, and the sample data of the same data type as the first block data and its similarity with the mining area characteristic fingerprint database with the highest similarity are used as input data, and the corresponding confidence probability is used as output data to train and obtain a prediction model.

[0101] As an alternative embodiment, the sampling information characterized in the first block data and the first similarity can also be evaluated based on the experience of artificial experts, so as to determine the first confidence probability based on preset rules.

[0102] The method for tracing the origin of copper concentrate samples based on X-ray fluorescence spectroscopy and isotope detection proposed in this application relates to the technical field of copper concentrate origin tracing. Data sampling is carried out on subjective and objective factors in the sampling process of copper concentrate samples, and data is stored based on the blockchain to improve the security and accuracy of the data. The data analysis results of copper concentrate samples are discretized, standardized, outlier corrected, and principal component analysis is performed, thereby ensuring the standardization of the data. Based on the first mining area characteristic fingerprint database set obtained by historical data mining, the target mining area to which the copper concentrate sample belongs and the corresponding first similarity are determined. Through the technical solution of this application, the standardization processing of the copper concentrate analysis results can be realized, and then the consistent rules can be extracted, and the confidence probability is given to the tracing result in combination with the sampling process data, so that the obtained tracing result is more accurate.

[0103] The above are only the preferred embodiments of the present invention. Therefore, all equivalent changes or modifications made according to the structure, characteristics, and principles described in the scope of this invention patent application are included in the scope of this invention patent application.

Claims

1. A copper concentrate sample traceability method based on X-ray fluorescence spectroscopy and isotope detection, characterized in that: The method includes: S0: Acquire first multi-source data of the copper concentrate sample, generate a first credible value for each of the first multi-source data, generate first block data according to the first multi-source data and the first credible value, and store the data on the chain; S1: Analyze the copper concentrate sample to obtain a plurality of first analysis results, and preprocess and discretize each of the first analysis results to obtain a first transaction database; S2: extracting a plurality of first data items from the first transaction database, and determining a target mining area characteristic fingerprint library and a first similarity from the plurality of first data items according to a first mining area characteristic fingerprint library set; wherein the first mining area characteristic fingerprint library is obtained after historical data analysis of a plurality of different copper concentrates; S3: determining a first target mining area and a first confidence probability corresponding to the copper concentrate sample according to the first block data and the first similarity; The first multi-source data in S0 is obtained by the following steps: S01: Acquire first subjective sampling data and first objective sampling data of the copper concentrate sample; S02: generating the first multi-source data according to the first subjective sampling data and the first objective sampling data; The first subjective sampling data includes the cumulative working time and the number of historical errors of the sampler, and the first objective sampling data includes the sampling location information and the sample static time before sampling; The S1 comprises the following steps: S11: storing the first analysis result in a table form to obtain a first data table; S12: performing standardization processing on each first data value stored in the first data table to obtain a first standardized data table; wherein each first analysis result is composed of a plurality of the first data values; S13: performing anomaly correction processing on the first standardized data table to obtain a second standardized data table; S14: performing principal component analysis and dimensionality reduction processing on the second standardized data table to obtain a third standardized data table; S15: discretizing the third standardized data table to obtain a first transaction database; In S2, the first mining area feature fingerprint library set is obtained by analyzing historical data of multiple different copper concentrates, and specifically includes the following steps: The sample data is divided into multiple groups based on different copper concentrates; each piece of sample data contains element interval, isotope interval, and mining area label; For multiple groups of sample data, multiple first mining area feature fingerprint libraries are obtained respectively, and the first mining area feature fingerprint library set is obtained according to the multiple first mining area feature fingerprint libraries.

2. The copper concentrate sample tracing method based on X-ray fluorescence spectroscopy and isotope detection according to claim 1 is characterized in that: The S0 further comprises the following steps: S03: inputting the first multi-source data into a first credibility determination model to obtain a first credibility corresponding to the first multi-source data; S04: Obtain first block data according to the first multi-source data and the first credibility, and store the first block data on-chain.

3. The copper concentrate sample tracing method based on X-ray fluorescence spectroscopy and isotope detection according to claim 2 is characterized in that: The first analysis result includes XRF element content and lead isotope ratio.

4. The copper concentrate sample tracing method based on X-ray fluorescence spectroscopy and isotope detection according to claim 3 is characterized in that: The principal component analysis and dimensionality reduction processing in S14 includes: The standardized isotope ratio is reduced in dimension to generate principal component features to replace the original features; The original features in the second standardized data table are replaced according to the generated principal component features, thereby obtaining the third standardized data table.

5. The copper concentrate sample tracing method based on X-ray fluorescence spectroscopy and isotope detection according to claim 4 is characterized in that: In the S0, the first data item is a data combination including an element interval and an isotope interval.

6. The copper concentrate sample tracing method based on X-ray fluorescence spectroscopy and isotope detection according to claim 5 is characterized in that: The first mining area feature fingerprint library includes: Multiple frequent item sets consisting of element contents and isotope intervals in specified copper concentrates.

Images