Method for improving accuracy of bauxite ore component detection data

Abstract

The invention relates to the technical field of improvement of data accuracy, in particular to a method for improving accuracy of bauxite ore component detection data, which comprises the following specific steps: (1) sampling: sampling by adopting a dynamic sampling method; (2) sample division operation; (3) detection operation: detecting and analyzing main component indexes and test component data values of bauxite by using an NIR technology online detection intelligent analysis system; (4) checking data; and (5) after data comparison, the deviation rates of all aluminum and effective aluminum of the ore are both smaller than four thousandths and are far lower than laboratory analysis permissible difference, and ore matching can be completely guided. The method has the advantages that 84 groups of data are obtained for nearly 3 months, results show that the data obtained by the method is matched with the data obtained by the original method, the maximum deviation rate is only 2.90%, the data stability is better, and the index condition of the total ore component grade can be objectively reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of improving data accuracy, and in particular to a method for improving the accuracy of bauxite ore component detection data. Background Art

[0002] With the increase in domestic alumina production in recent years, domestic bauxite resources are becoming increasingly scarce, and the dependence on imported bauxite is increasing year by year. In 2022, China imported 125 million tons of bauxite, a year-on-year increase of 16.8%. Compared with 2017, the import volume increased by 82.9%, and the 5-year CAGR was 12.8%. Structurally, China mainly imports bauxite from three countries: Guinea, Australia and Indonesia.

[0003] Due to the large fluctuations in the content of imported ores, a large amount of sampling, preparation, laboratory analysis, and data output are required. Conventional laboratory analysis is carried out by chemical manual methods and laboratory XRF. Imported ores have the characteristics of large fluctuations in active silicon and effective aluminum indicators. The sampling representativeness of each batch of samples is poor, and the data results are seriously lagging. Summary of the invention

[0004] In order to solve the defects in the prior art, the present invention discloses a method for accurately establishing a model, and its technical solution is as follows:

[0005] A method for improving the accuracy of bauxite ore component detection data comprises the following steps:

[0006] (1) Sampling: dynamic sampling method is used for sampling;

[0007] (2) Sample reduction operation: the bauxite obtained in the sampling stage is placed in a container as sample 1, sample 1 is evenly reduced and divided into sample 2 and sample 3, sample 2 is evenly divided into sample 4 and sample 5, and sample 5 is merged with sample 3; this operation is repeated 5 times, and the samples after the first round of reduction are tested and analyzed;

[0008] Next, a second round of subdivision operation is performed on the remaining samples, and a second test and analysis is performed on the subdivided samples; the second round of subdivision operation is the same as the first round of subdivision operation;

[0009] Finally, the remaining samples that have not been reduced are double-packed in sealed bags and the sample labels are sealed for processing;

[0010] (3) Detection operation: Use NIR technology to detect the intelligent analysis system online to detect and analyze the main component indicators and test component data values ​​of bauxite;

[0011] (4) Data assessment:

[0012] The main component index and chemical component data of bauxite detected and analyzed in the previous step R 2 >0.9;

[0013] The standard deviation of the main component index and the test component data of bauxite detected and analyzed in the previous step and the main component index and the test component data of bauxite analyzed by laboratory XRF is within ±2σ;

[0014] The assessment data in the following table is established with reference to the national standard for alumina:

[0015]

[0016] (5) Data comparison:

[0017] The main component indicators of bauxite obtained by the online detection and analysis of the system described in the previous step and the analysis using the existing technical method are compared with the test component data results.

[0018] In the present invention, it is further described that the specific requirements for dynamic sampling in step (1) are as follows:

[0019] 1) Sampling location: above the belt transporting aluminum ore;

[0020] 2) Sample quantity: The sampling quantity is 12, each sample is about 5kg;

[0021] 3) Sampling cycle: 4 samples are taken each time, once every 8 hours;

[0022] 4) Sampling requirements: Use a sampling shovel for sampling. A single sample is completed by sampling more than 10 times. Each shovel samples about 500g. After sampling, each sample is about 5kg. The steps are as follows:

[0023] Time: 00:00:15, the first shovel is placed into the first bucket or sealed bag;

[0024] Time: 00:00:30, the second shovel is placed into the second bucket or sealed bag;

[0025] Time: 00:00:45, the third shovel is placed into the third bucket or sealed bag;

[0026] Time: 00:01:00, the fourth shovel is placed in the fourth bucket or sealed bag;

[0027] Time: 00:01:10, the fifth shovel is placed into the first bucket or sealed bag;

[0028] Time: 00:01:30, the sixth shovel is placed into the second bucket or sealed bag;

[0029] Time: 00:01:45, the seventh shovel is placed into the third bucket or sealed bag;

[0030] Time: 00:02:00, the eighth shovel is placed into the fourth bucket or sealed bag;

[0031] The cycle is repeated until sampling is completed.

[0032] In the present invention, it is further described that in step (3), the main component indicators and the assay components are: activated alumina, reactive silicon, total alumina, silicon dioxide, ferric oxide, total carbon, sodium oxide and water.

[0033] In the present invention, it is further stated that in step (3), the standard deviation is SEP / σ, wherein SEP represents standard error prediction and σ represents standard deviation.

[0034] In the present invention, it is further described that in step (5), the method of obtaining the main component index and the assay component data of bauxite by the analysis of the prior art method is as follows:

[0035] S1. Sampling:

[0036] 1) Sampling location: Sampling is carried out at the ore yard;

[0037] 2) Sample quantity: The sampling quantity is 3, each sample is about 10kg;

[0038] 3) Sampling range: Sampling is carried out according to the different particle sizes, colors, hardness and content of different indicators of the ore to ensure full coverage of the model data;

[0039] 4) Sampling requirements: Each sample is sampled for at least 20 shovels, each shovel is about 500g, and one sample is 10kg;

[0040] S2, the original reduction operation used the traditional quartering method to reduce the sample;

[0041] S3. Testing operation uses laboratory XRF method to test and analyze the main component indicators and chemical components of bauxite.

[0042] In the present invention, it is further described that in step (5), the data standard for comparison is as follows:

[0043] ① The laboratory XRF analysis value has sampling representative error (ER) and assay analysis error (EA), which will be determined during the performance test and deducted from the difference between the laboratory XRF analysis value (VL) and the online detection analysis value (VS);

[0044] ② The absolute error between the main component index of bauxite and the tested component, evaluation method:

[0045] Absolute error = (laboratory XRF analysis value or analysis average - average value of online detection analysis value during the corresponding sampling time) - (sampling representative error + laboratory XRF analysis error) < STDEV value defined in the assessment table in step (4).

[0046] In the present invention, it is further described that the sampling representative error (ER) is determined according to a dynamic sampling step;

[0047] The sampling representative error (ER) is calculated as ER = (Xmax-Xmin) / 2;

[0048] Xmax is the maximum measured value, and Xmin is the minimum measured value.

[0049] In the present invention, it is further described that the laboratory analysis error (EA) is to mix the samples and analyze each sample twice;

[0050] The analytical error EA is calculated as EA = (Xmax - Xmin) / 2;

[0051] Xmax is the maximum measured value, and Xmin is the minimum measured value.

[0052] Compared with the prior art, the present invention has the following advantages:

[0053] (1) Inspection and detection of abnormal materials. It can react sensitively to the monitoring of materials with abnormal components and play an early warning role.

[0054] (2) Comparison between the method of the present application and the original method: The data of the present application and the original method are nearly 3 months, with a total of 84 sets of data. The data detected by the two methods are consistent with each other, with the maximum deviation rate of only 2.90%. In addition, the method of the present application has better stability and can objectively reflect the overall ore composition grade indicators.

[0055] (3) Comparison of the acceptance of sampling data at the same location: The committed standard deviation of the analysis values ​​of the proposed method and the original method meets the standard requirements, as shown in Table 1. The results show that the total aluminum, effective aluminum, silicon, iron content and other items of the two methods are within the allowable range.

[0056] (4) The analysis comparison between the present application method and the original method shows that the deviation rates of total aluminum and effective aluminum are both less than 0.4%, which is far lower than the allowable error of laboratory analysis and can fully guide ore matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is the on-site sampling diagram for the original sampling method.

[0058] Figure 2 The present invention is a specific flow chart of sample reduction for the present application method.

[0059] Figure 3 This is the on-site picture of the online testing for this application.

[0060] Figure 4 A comparison chart of the methods of this application. DETAILED DESCRIPTION

[0061] The specific implementation modes of the present invention are further described below in conjunction with examples.

[0062] Step 1: Sampling operation;

[0063] The sampling operation used in this application is dynamic sampling, the specific contents are as follows:

[0064] 1) Sampling location: above the belt transporting aluminum ore;

[0065] 2) Sample quantity: The sampling quantity is 12, each sample is about 5kg;

[0066] 3) Sampling cycle: 4 samples are taken each time, once every 8 hours;

[0067] 4) Sampling requirements: Use a sampling shovel for sampling. A single sample is completed by sampling more than 10 times. Each shovel samples about 500g. After sampling, each sample is about 5kg. The steps are as follows:

[0068] Time: 00:00:15, the first shovel is placed into the first bucket or sealed bag;

[0069] Time: 00:00:30, the second shovel is placed into the second bucket or sealed bag;

[0070] Time: 00:00:45, the third shovel is placed into the third bucket or sealed bag;

[0071] Time: 00:01:00, the fourth shovel is placed in the fourth bucket or sealed bag;

[0072] Time: 00:01:10, the fifth shovel is placed into the first bucket or sealed bag;

[0073] Time: 00:01:30, the sixth shovel is placed into the second bucket or sealed bag;

[0074] Time: 00:01:45, the seventh shovel is placed into the third bucket or sealed bag;

[0075] Time: 00:02:00, the eighth shovel is placed into the fourth bucket or sealed bag;

[0076] The cycle is repeated until sampling is completed.

[0077] The specific contents of the original sampling operation are as follows, and the on-site sampling diagram is as follows: Figure 1 As shown:

[0078] 1) Sampling location: Sampling is carried out at the ore yard;

[0079] 2) Sample quantity: The sampling quantity is 3, each sample is about 10kg;

[0080] 3) Sampling range: Sampling is carried out according to the different particle sizes, colors, hardness and content of different indicators of the ore to ensure full coverage of the model data;

[0081] 4) Sampling requirements: Each sample is sampled for at least 20 shovels, each shovel is about 500g, and one sample is 10kg;

[0082] Step 2: Scaling operation;

[0083] The specific reduction operations of this application are as follows:

[0084] The bauxite obtained in the sampling stage is placed in a container as sample 1, sample 1 is evenly divided into sample 2 and sample 3, sample 2 is evenly divided into sample 4 and sample 5, and sample 5 is combined with sample 3; this operation is repeated 5 times, and the samples after the first round of division are tested and analyzed;

[0085] Next, a second round of subdivision operation is performed on the remaining samples, and a second test and analysis is performed on the subdivided samples; the second round of subdivision operation is the same as the first round of subdivision operation;

[0086] Finally, the remaining samples that have not been reduced are double-packed in sealed bags and the sample labels are sealed for processing;

[0087] The specific process is as follows Figure 2 As shown;

[0088] The original reduction operation uses the traditional four-part reduction method. The specific steps are as follows:

[0089] 1) Cone stacking: stack the sample into a cone shape to ensure that the sample forms a regular particle size distribution from top to bottom and from center to outer edge. When stacking the cone, you can stand on both sides and operate alternately, and stack at least 3 times.

[0090] 2) Flat pile: insert a thin plate into the top of the cone to spread the sample along the circumference and finally form a disc shape. Ensure that the sample is evenly distributed.

[0091] 3) Sample division: Divide the disc-shaped sample into four quadrants, combine the two diagonal parts, discard one part and use the other part as the sample.

[0092] 4) Mixing: Mixing is required before each reduction. The mixing by the cone method and the reduction by quartering method are often carried out together.

[0093] 5) Repeat: Repeat the above steps until the required amount of sample is obtained.

[0094] Step 3: Detection operation;

[0095] The detection operation used in this application is the NIR technology online detection intelligent analysis system. The online detection site is shown in the figure below. Figure 3 As shown, the original detection operation is the laboratory XRF method, following YS / T 575.23-2021 "Chemical Analysis Methods for Bauxite Ore Part 23: Determination of Element Content by X-ray Fluorescence Spectrometry". The analysis content is the main component indicators of bauxite and the test component data values.

[0096] Step 4: Data assessment;

[0097] Data accuracy requirements during modeling and calibration: System measurement of main component indicators and assay component data R 2 >0.9.

[0098] Deviation requirement: The committed standard deviation between the system online detection value and the laboratory XRF analysis value.

[0099] The assessment data in the following table is established with reference to the national standard for alumina:

[0100] Table 1 Standard table of chemical composition of alumina

[0101]

[0102] According to the central limit theorem, the online prediction value is Gaussian distributed. Therefore, SEP / σ is equivalent to the standard deviation, which promises that 63% of the standard deviations are within ±1σ and 95% of the standard deviations are within ±2σ.

[0103] Step 5: Data comparison;

[0104] Ore composition data commitment testing in bauxite is carried out by sampling while the belt is running, with the aim of comparing the online detection analysis values ​​with the corresponding laboratory XRF analysis values ​​over a period of time.

[0105] The laboratory XRF analysis value has a sampling representative error (ER) and an assay analysis error (EA), which will be determined during performance testing and deducted from the difference between the laboratory XRF analysis value (VL) and the online detection analysis value (VS).

[0106] Absolute error of ore composition data in bauxite, evaluation method:

[0107] Absolute error = (VL-VS)-(ER+EA) < STDEV value defined in the assessment table;

[0108] VL = laboratory XRF analysis value or analysis average;

[0109] VS = average value of online detection analysis value during the corresponding sampling time;

[0110] The sampling representative error (ER) is determined based on the dynamic sampling procedure.

[0111] The sampling representative error (ER) is calculated as ER = (Xmax-Xmin) / 2;

[0112] Xmax is the maximum measurement value, Xmin is the minimum measurement value

[0113] To determine sample preparation and laboratory analytical error (EA), samples were mixed and each sample was analyzed twice.

[0114] The analytical error EA is calculated as EA = (Xmax - Xmin) / 2;.

[0115] Xmax is the maximum measured value, and Xmin is the minimum measured value.

[0116] The following are the data results of the technical method of this application and the original technical method. The steps of the two methods are as follows: Figure 4 As shown, the data lasting nearly 3 months, a total of 84 sets of data, were obtained and analyzed for the ore composition data in the bauxite.

[0117] Table 2

[0118]

[0119]

[0120]

[0121]

[0122] The original method is used to statistically analyze the composition data of the ore. The specific data are as follows:

[0123] Table 3

[0124]

[0125]

[0126]

[0127]

[0128] The data in Table 2 and Table 3 show that when the analysis method of the present application is compared with the original method, the deviation rates of total aluminum and effective aluminum are both less than 0.4%, which is far lower than the allowable error of laboratory analysis and can fully guide the ore matching.

[0129] After collating the data for three months and analyzing them from the perspectives of month and overall, the data comparison between the methods is shown in the following table.

[0130] Table 4

[0131] By comparing the method of this patent application with the original method, the data volatility of the method of this application is lower, and the results show that

[0132]

[0133] The data obtained by the method of the present application is consistent with the data obtained by the original method, with a maximum deviation rate of only 2.90%. In addition, the data stability of the method of the present application is better, and it can objectively reflect the indicators of the overall ore composition grade.

[0134] The above description is a detailed description of the preferred feasible embodiments of the present invention, but the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modified changes completed under the technical spirit suggested by the present invention should fall within the patent scope covered by the present invention.

Claims

1. A method for improving the accuracy of bauxite ore composition detection data, characterized in that: The steps include: (1) Sampling: dynamic sampling method is used for sampling; (2) Sample reduction operation: the bauxite obtained in the sampling stage is placed in a container as sample 1, sample 1 is evenly reduced and divided into sample 2 and sample 3, sample 2 is evenly divided into sample 4 and sample 5, and sample 5 is merged with sample 3; this operation is repeated 5 times, and the samples after the first round of reduction are tested and analyzed; Next, a second round of subdivision operation is performed on the remaining samples, and a second test and analysis is performed on the subdivided samples; the second round of subdivision operation is the same as the first round of subdivision operation; Finally, the remaining samples that have not been reduced are double-packed in sealed bags and the sample labels are sealed for processing; (3) Detection operation: Use NIR technology to detect the intelligent analysis system online to detect and analyze the main component indicators and test component data values ​​of bauxite; (4) Data assessment: The main component index and chemical component data of bauxite detected and analyzed in the previous step R 2 >0.9; The standard deviation of the main component index and the test component data of bauxite detected and analyzed in the previous step and the main component index and the test component data of bauxite analyzed by laboratory XRF is within ±2σ; The assessment data in the following table is established with reference to the national standard for alumina: (5) Data comparison: The main component indicators of bauxite obtained by the online detection and analysis of the system described in the previous step and the analysis using the existing technical method are compared with the test component data results.

2. A method for improving the accuracy of bauxite ore component detection data according to claim 1, characterized in that: The specific requirements for dynamic sampling in step (1) are as follows: 1) Sampling location: above the belt transporting aluminum ore; 2) Sample quantity: The sampling quantity is 12, each sample is about 5kg; 3) Sampling cycle: 4 samples are taken each time, once every 8 hours; 4) Sampling requirements: Use a sampling shovel for sampling. A single sample is completed by sampling more than 10 times. Each shovel samples about 500g. After sampling, each sample is about 5kg. The steps are as follows: Time: 00:00:15, the first shovel is placed into the first bucket or sealed bag; Time: 00:00:30, the second shovel is placed into the second bucket or sealed bag; Time: 00:00:45, the third shovel is placed into the third bucket or sealed bag; Time: 00:01:00, the fourth shovel is placed in the fourth bucket or sealed bag; Time: 00:01:10, the fifth shovel is placed into the first bucket or sealed bag; Time: 00:01:30, the sixth shovel is placed into the second bucket or sealed bag; Time: 00:01:45, the seventh shovel is placed into the third bucket or sealed bag; Time: 00:02:00, the eighth shovel is placed into the fourth bucket or sealed bag; The cycle is repeated until sampling is completed.

3. A method for improving the accuracy of bauxite ore component detection data according to claim 1, characterized in that: In step (3), the main component indicators and the assay components are: activated alumina, reactive silicon, total alumina, silicon dioxide, ferric oxide, total carbon, sodium oxide and water.

4. A method for improving the accuracy of bauxite ore component detection data according to claim 1, characterized in that: In step (3), the standard deviation is SEP / σ, where SEP represents standard error prediction and σ represents standard deviation.

5. The method for improving the accuracy of bauxite ore component detection data according to claim 1, characterized in that: In step (5), the main component index and the assay component data of bauxite obtained by the analysis of the prior art method are as follows: S1. Sampling: 1) Sampling location: Sampling is carried out at the ore yard; 2) Sample quantity: The sampling quantity is 3, each sample is about 10kg; 3) Sampling range: Sampling is carried out according to the different particle sizes, colors, hardness and content of different indicators of the ore to ensure full coverage of the model data; 4) Sampling requirements: Each sample is sampled for at least 20 shovels, each shovel is about 500g, and one sample is 10kg; S2, the original reduction operation used the traditional quartering method to reduce the sample; S3. Testing operation uses laboratory XRF method to test and analyze the main component indicators and chemical components of bauxite.

6. A method for improving the accuracy of bauxite ore component detection data according to claim 1, characterized in that: In step (5), the data standards for comparison are as follows: ① The laboratory XRF analysis value has sampling representative error (ER) and assay analysis error (EA), which will be determined during the performance test and deducted from the difference between the laboratory XRF analysis value (VL) and the online detection analysis value (VS); ② The absolute error between the main component index of bauxite and the tested component, evaluation method: Absolute error = (laboratory XRF analysis value or analysis average - average value of online detection analysis value during the corresponding sampling time) - (sampling representative error + laboratory XRF analysis error) < STDEV value defined in the assessment table in step (4).

7. A method for improving the accuracy of bauxite ore component detection data according to claim 6, characterized in that: The sampling representative error (ER) is determined according to a dynamic sampling step; The sampling representative error (ER) is calculated as ER = (Xmax-Xmin) / 2; Xmax is the maximum measured value, and Xmin is the minimum measured value.

8. A method for improving the accuracy of bauxite ore component detection data according to claim 6, characterized in that: The laboratory analysis error (EA) is that the samples are mixed and each sample is analyzed twice; The analytical error EA is calculated as EA = (Xmax - Xmin) / 2; Xmax is the maximum measured value, and Xmin is the minimum measured value.

Images