A method for judging whether the prediction result of element grade based on LIBS is abnormal

Abstract

A method for judging whether the element grade prediction result is abnormal based on LIBS, comprising the following steps: S1: obtaining the average LIBS spectrum data of the to-be-tested ore pulp sample; S2: inputting the average LIBS spectrum data of the to-be-tested ore pulp sample into the element grade prediction model of each selected element respectively, obtaining the element grade prediction result corresponding to each selected element; inputting the average LIBS spectrum data of the to-be-tested ore pulp sample into the moisture content prediction model, obtaining the moisture content prediction result of the to-be-tested ore pulp sample; S3: judging whether the moisture content prediction result of the to-be-tested ore pulp sample is within the preset range, if not, determining that the element grade prediction result corresponding to each selected element is abnormal. The present application can automatically, quickly and accurately judge whether the element grade prediction result is abnormal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ore pulp element grade analysis, and more particularly to a method for judging whether element grade prediction results are abnormal based on LIBS. BACKGROUND

[0002] At present, ore pulp element grade analysis is mostly offline analysis, that is, manually intercepting shift samples at designated sampling points, and then obtaining ore pulp element grade through chemical titration or X-ray fluorescence spectrum analysis after a series of pretreatment processes such as laboratory filtration, drying and sample preparation. The entire analysis process takes about one hour, and if the number of shift samples is large, the analysis time will be longer, which leads to serious lag in obtaining ore pulp element grade, and has no real-time guiding significance for the adjustment of the front-end flotation process and the addition of flotation reagents, resulting in poor stability of concentrate product quality in the production process, and causing great impact on production cost and product quality of the enterprise.

[0003] Laser-induced breakdown spectroscopy (LIBS) is a kind of plasma emission spectrum analysis technology, which has the advantages of simultaneous measurement of multiple elements, simple sample pretreatment, fast analysis speed, and small sample damage, and is very suitable for full-element detection. At present, this technology has been researched and applied in the fields of mineral industry process monitoring, agriculture, food, environment and medical treatment.

[0004] One of the difficulties in analyzing ore pulp element grade by using LIBS technology is that the intensity and stability of LIBS spectrum signal are easily affected by the water content in ore pulp, which further affects the accuracy of ore pulp element grade analysis.

[0005] Therefore, how to provide a method for automatically, quickly and accurately judging whether element grade prediction results are abnormal is a problem that those skilled in the art need to solve. SUMMARY

[0006] Therefore, the present application provides a method for judging whether element grade prediction results are abnormal based on LIBS, and the technical scheme is as follows:

[0007] A method for judging whether element grade prediction results are abnormal based on LIBS, comprising the following steps:

[0008] S1: obtaining average LIBS spectrum data of a to-be-tested ore pulp sample;

[0009] S2: inputting the average LIBS spectrum data of the to-be-tested ore pulp sample into an element grade prediction model of each selected element respectively to obtain element grade prediction results corresponding to each selected element;

[0010] inputting the average LIBS spectrum data of the to-be-tested ore pulp sample into a water content prediction model to obtain water content prediction results of the to-be-tested ore pulp sample;

[0011] S3: determining whether the water content prediction result of the ore pulp sample is within a preset range; if the water content prediction result is within the preset range, determining that the element grade prediction result is normal; if the water content prediction result is not within the preset range, determining that the element grade prediction result is abnormal;

[0012] The preset range is 65-80%.

[0013] The element grade prediction model obtaining method is:

[0014] S21: obtaining average LIBS spectrum data of the ore pulp sample;

[0015] obtaining a measured element grade of a selected element in the ore pulp sample;

[0016] S22: taking the average LIBS spectrum data in S21 as an input of the partial least squares method; taking the measured element grade in S21 as an output of the partial least squares method; and obtaining an element grade prediction model of the selected element.

[0017] The water content prediction model obtaining method is:

[0018] S21': obtaining average LIBS spectrum data of the ore pulp sample;

[0019] obtaining a measured water content of the ore pulp sample;

[0020] S22': taking the average LIBS spectrum data in S21' as an input of the partial least squares method; taking the measured water content in S21' as an output of the partial least squares method; and obtaining the water content prediction model.

[0021] The average LIBS spectrum data obtaining method is:

[0022] (1) intercepting ore pulp in a main ore pulp pipe by using an ore pulp sampler;

[0023] (2) dividing the intercepted ore pulp by using a multi-channel divider;

[0024] (3) inputting the divided ore pulp into a LIBS analyzer to obtain average LIBS spectrum data.

[0025] The measured element grade obtaining method is:

[0026] intercepting an ore pulp sample from a sample port of the multi-channel divider; performing suction filtration and drying on the intercepted ore pulp sample to prepare a mineral powder; and measuring a measured element grade of a selected element in the ore pulp sample by a chemical titration method or an X-ray fluorescence spectrum analysis method.

[0027] The measured water content obtaining method is:

[0028] The slurry sample is intercepted from the sample port of the multi-path splitter; the intercepted slurry sample is weighed to obtain a weighing value A; the intercepted slurry sample is filtered and dried to prepare a mineral powder sample, and the mineral powder sample is weighed to obtain a weighing value B; and the measured water content of the slurry sample is calculated according to the weighing values, and the measured water content = (A-B) / A.

[0029] The average LIBS spectral data include the characteristic spectral line intensity of all elements in the to-be-detected slurry sample.

[0030] The present application has the following beneficial effects:

[0031] The present application provides a method for judging whether the element grade prediction result is abnormal based on LIBS, by establishing an element grade prediction model and a water content prediction model, obtaining the element grade prediction result and the water content prediction result according to the LIBS spectral data, and determining whether the element grade prediction result is abnormal according to the water content prediction result, the element grade prediction result can be automatically, quickly and accurately judged. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The flowchart of the present application;

[0033] Figure 2 The relationship diagram between the training data, the verification data and the prediction data of the phosphorus element in the slurry sample. DETAILED DESCRIPTION

[0034] The embodiments of the present application will be described in detail below with reference to the embodiments, and the following embodiments are only used to illustrate the present application, and should not be regarded as limiting the scope of the present application.

[0035] Example 1

[0036] Prediction model establishment:

[0037] 1) Select magnesium concentrate slurry with a water content of 65% to 80%, and intercept the slurry every two hours using a magnesium concentrate sampler;

[0038] Specifically, 560 magnesium concentrate slurry samples are obtained in the present application;

[0039] 2) The slurry sample intercepted each time is input into the multi-path splitter (the multi-path splitter realizes that a part of the slurry is input into the LIBS analyzer) through a pipeline, and then flows into the LIBS analyzer for spectral measurement after being divided, and the average LIBS spectral data of the slurry sample are obtained and saved;

[0040] Specifically: in actual measurement, 1000 pulses are excited for each ore pulp sample to obtain 1000 LIBS spectrum data, the first 200 LIBS spectrum data with large fluctuation are removed, and then the average of the remaining 800 LIBS spectrum data is calculated; the average is the average LIBS spectrum data of the ore pulp sample; the reason for measuring multiple times and taking the average is that the particles in the ore pulp are not uniform, resulting in large fluctuation of the obtained spectrum; only by measuring multiple times and taking the average can part of the measurement error be eliminated, so that the spectrum data is more representative.

[0041] At the same time of measurement by the LIBS analyzer, the ore pulp sample is obtained at the sample port of the multi-path splitter;

[0042] The measured element grades of the four selected elements of phosphorus, magnesium, iron and aluminum are measured by a chemical titration method or an X-ray fluorescence spectrum analysis method;

[0043] Then, the ore powder before filtration and after drying is weighed, the difference between the weight B of the ore powder after drying and the weight A of the ore powder before filtration is calculated, and the measured water content of the ore pulp sample is obtained = (A-B) / A;

[0044] Based on this, the 560 average LIBS spectrum data, 560 measured element grades of phosphorus, 560 measured element grades of magnesium, 560 measured element grades of iron, 560 measured element grades of aluminum and 560 measured water contents are obtained;

[0045] 3) The first 500 average LIBS spectrum data, 500 measured element grades of phosphorus, 500 measured element grades of magnesium, 500 measured element grades of iron, 500 measured element grades of aluminum and 500 measured water contents are used as training data.

[0046] The first 500 average LIBS spectrum data are used as the input of the partial least squares method, and the corresponding first 500 measured element grades of phosphorus are used as the output of the partial least squares method to obtain an element grade prediction model of phosphorus;

[0047] The first 500 average LIBS spectrum data are used as the input of the partial least squares method, and the corresponding first 500 measured element grades of magnesium are used as the output of the partial least squares method to obtain an element grade prediction model of magnesium;

[0048] The first 500 average LIBS spectrum data are used as the input of the partial least squares method, and the corresponding first 500 measured element grades of iron are used as the output of the partial least squares method to obtain an element grade prediction model of iron;

[0049] The first 500 average LIBS spectrum data are taken as the input of the partial least squares method, and the corresponding first 500 measured element grade of aluminum element is taken as the output of the partial least squares method, to obtain an element grade prediction model of the aluminum element;

[0050] The first 500 average LIBS spectrum data are taken as the input of the partial least squares method, and the corresponding first 500 measured moisture content is taken as the output of the partial least squares method, to obtain a moisture content prediction model.

[0051] Example 2

[0052] Prediction model verification:

[0053] On the basis of example 1, the remaining 60 average LIBS spectrum data, the remaining 60 measured element grade of magnesium element, the remaining 60 measured element grade of iron element, the remaining 60 measured element grade of iron element and the remaining 60 measured moisture content are taken as verification data and prediction data to verify and predict the above-mentioned five models;

[0054] Specifically: the remaining 60 average LIBS spectrum data are input into the above-mentioned five prediction models to obtain prediction data; the remaining 60 measured element grade of phosphorus element, the remaining 60 measured element grade of magnesium element, the remaining 60 measured element grade of iron element, the remaining 60 measured element grade of aluminum element and the remaining 60 measured moisture content are taken as verification data.

[0055] Figure 2 The R of the element grade prediction model of the phosphorus element is 0.984, and the RMSEP is 0.465. 2 The R of the element grade prediction model of the phosphorus element is 0.984, and the RMSEP is 0.465.

[0056] In addition, the R of the moisture content prediction model is 0.988, the R of the element grade prediction model of the magnesium element is 0.983, the R of the element grade prediction model of the iron element is 0.982, the R of the element grade prediction model of the aluminum element is 0.983. 2 2 2 2

[0057] Example 3

[0058] Formal production application:

[0059] 1) The five prediction models of example 1 are imported into the LIBS analyzer;

[0060] ​​​​2) Obtain the sample of the magnesium concentrate slurry to be tested by the magnesium concentrate sampler, and input into the element grade prediction model of phosphorus element, the element grade prediction model of magnesium element, the element grade prediction model of iron element, the element grade prediction model of aluminum element and the moisture content prediction model respectively, to obtain the element grade of phosphorus element, the element grade of magnesium element, the element grade of iron element, the element grade of aluminum element and the moisture content of the magnesium concentrate slurry;

[0061] 3) Determine whether the moisture content of the magnesium concentrate slurry is within 65% to 80%:

[0062] If the moisture content is not within 65% to 80%, it is determined that the obtained prediction results of the element grade of phosphorus element, the element grade of magnesium element, the element grade of iron element and the element grade of aluminum element are abnormal, and the abnormal information is uploaded to the data center to prompt the flotation workshop to pay attention to the slurry concentration and timely adjust the process.

[0063] If the moisture content is within 65% to 80%, it is determined that the obtained prediction results of the element grade of phosphorus element, the element grade of magnesium element, the element grade of iron element and the element grade of aluminum element are normal.

Claims

1. A method for determining whether an element grade prediction result is abnormal based on LIBS, characterized in that: The following steps are involved: S1: Obtain the average LIBS spectrum data of the slurry sample to be tested; S2: inputting the average LIBS spectrum data of the slurry sample to be tested into the element grade prediction model of each selected element to obtain the element grade prediction result corresponding to each selected element; Inputting the average LIBS spectrum data of the slurry sample to be tested into the moisture content prediction model to obtain a moisture content prediction result of the slurry sample to be tested; S3: Determine whether the moisture content prediction result of the slurry sample to be tested is within a preset range: if the moisture content prediction result is within the preset range, then determine that the element grade prediction result is normal; If the water content prediction result is not within the preset range, it is determined that the element grade prediction result is abnormal; The preset range is 65~80%.

2. The method for determining whether an element grade prediction result is abnormal based on LIBS according to claim 1, characterized in that: The method for obtaining the element grade prediction model is: S21: Obtain the average LIBS spectrum data of the slurry sample; Obtain measured elemental grades of selected elements in slurry samples; S22: Using the average LIBS spectral data in S21 as the input of the partial least squares method; using the measured element grade in S21 as the output of the partial least squares method; and obtaining an element grade prediction model for the selected element.

3. The method for determining whether the element grade prediction result is abnormal based on LIBS according to claim 2, characterized in that: The method for obtaining the moisture content prediction model is: S21': obtain the average LIBS spectrum data of the pulp sample; Obtain the measured moisture content of the slurry sample; S22': using the average LIBS spectrum data in S21' as the input of the partial least squares method; using the measured moisture content in S21' as the output of the partial least squares method; and obtaining the moisture content prediction model.

4. The method for determining whether an element grade prediction result is abnormal based on LIBS according to claim 3, characterized in that: The method for obtaining the average LIBS spectrum data is: (1) Use the slurry sampler to intercept the slurry in the main slurry pipe; (2) Using a multi-channel reducer to reduce the intercepted slurry; (3) The slurry after reduction is input into the LIBS analyzer to obtain the average LIBS spectrum data.

5. The method for determining whether the element grade prediction result is abnormal based on LIBS according to claim 4, characterized in that: The method for obtaining the measured element grade is: A pulp sample is intercepted from the sample port of the multi-channel reducer; the intercepted pulp sample is filtered and dried to prepare a sample into a mineral powder; and the measured element grade of the selected element in the pulp sample is obtained by chemical titration or X-ray fluorescence spectrometry.

6. The method for determining whether an element grade prediction result is abnormal based on LIBS according to claim 5, characterized in that: The method for obtaining the measured moisture content is: A slurry sample is intercepted from the sample port of the multi-channel reducer; the intercepted slurry sample is weighed to obtain a weight value A; the sample is filtered and dried to prepare a mineral powder, and the mineral powder is weighed to obtain a weight value B; based on the weighed value, the measured moisture content of the slurry sample is calculated, and the measured moisture content = (AB) / A.

7. The method for determining whether an element grade prediction result is abnormal based on LIBS according to claim 4, characterized in that: The average LIBS spectrum data includes the characteristic spectral line intensities of all elements in the ore pulp sample to be tested.

Images