Search

CN-121994780-A - Mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy

CN121994780ACN 121994780 ACN121994780 ACN 121994780ACN-121994780-A

Abstract

The invention relates to the technical field of spectrum online detection, and particularly discloses a mining full-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy, which comprises the following steps: the method comprises the steps of collecting mining online mineral flows, initial laser parameters, real-time environment parameters and mineral preprocessing information, conducting pulse coding laser induced plasma optimization, collecting self-adaptive spectrums, conducting background suppression on mining scenes, obtaining net characteristic spectrums after background removal, constructing a spectrum band attention-element category attention double-branch network, outputting high-dimensional compact characteristic vectors, characteristic attention weight graphs and element primary identification results, conducting characteristic weighting-plasma parameter correction fusion analysis, conducting mining full-element detection correction, and outputting corrected final element content and mineral identification results. The invention solves the problems of complex sample pretreatment, long time consumption, low light element detection precision and incapability of online detection in the traditional detection technology.

Inventors

  • ZHANG XIAHUI
  • YANG LONGFENG
  • ZHANG YU

Assignees

  • 机数仪器(浙江)有限公司

Dates

Publication Date
20260508
Application Date
20260127

Claims (10)

  1. 1. The mining full-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy is characterized by comprising the following steps of: Acquiring mining online mineral flow, initial laser parameters, real-time environment parameters and mineral preprocessing information, performing pulse coding laser-induced plasma optimization, and outputting a plasma time resolution spectrum and plasma key parameters; performing self-adaptive spectrum acquisition based on the plasma time resolution spectrum and plasma key parameters, and performing background suppression on mining scenes to obtain a net characteristic spectrum after background removal; based on the net characteristic spectrum and the mineral aggregate matrix type, constructing a spectrum band attention-element category attention double-branch network, and outputting a high-dimensional compact characteristic vector, a characteristic attention weight graph and an element primary identification result; performing feature weighting-plasma parameter correction fusion analysis on the high-dimensional compact feature vector, the feature attention weight graph, the element primary identification result and the plasma key parameters to obtain the initial value of the quantitative content of each element and the confidence coefficient of the quantitative result; And carrying out mining full-element detection correction by combining the quantitative result confidence, the high-dimensional compact feature vector and the net feature spectrum, and outputting corrected final element content and mining species identification result.
  2. 2. The mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy of claim 1 is characterized in that the method is used for collecting mining online mineral flow, initial laser parameters, real-time environment parameters and mineral pretreatment information, and comprises the following specific analysis processes: collecting mineral online mineral aggregate flow, initial laser parameters, real-time environment parameters and mineral aggregate pretreatment information: The mineral online mineral material flow specifically comprises mineral material belt conveying speed, mineral material granularity and mineral material water content; The initial laser parameters comprise a fundamental frequency, a pulse energy range, a pulse width and a repetition frequency; The real-time environment parameters specifically comprise real-time environment temperature T, real-time environment humidity and real-time environment dust concentration; The mineral aggregate pretreatment information specifically comprises the primary screening granularity of mineral aggregate and the surface roughness of the mineral aggregate.
  3. 3. The mining all-element intelligent online detection and analysis method based on the laser-induced breakdown spectroscopy of claim 2 is characterized in that the pulse-code laser-induced plasma optimization is carried out, the plasma time-resolved spectroscopy and plasma key parameters are output, and the specific analysis process is as follows: The mining online mineral flow, the initial laser parameters, the real-time environment parameters and mineral preprocessing information are converted into a specific three-pulse coding sequence by a genetic algorithm, and the optimized coding sequence is obtained by taking the optimization of signal-to-noise ratio maximization and the optimal stability of plasma excitation temperature generated by laser induction as targets: Shooting the surface of the mineral aggregate in real time through a high-speed linear array camera, calculating the fluctuation height of the surface of the mineral aggregate by combining a laser ranging module, and monitoring the offset of the mineral aggregate flow on a belt through visual positioning; the piezoelectric ceramic drives the dynamic lens group, and according to the surface relief height, the lens position which corresponds to the surface relief height stored in the database and meets the requirement that the laser focus always falls on the surface of the mineral aggregate is determined, and the lens position is adjusted; According to the optimized coding sequence, the laser is controlled to emit three pulses, the preheating pulse bombards the surface of the mineral aggregate, the core of the main excitation pulse breaks down the mineral aggregate to form high-temperature plasma, and the energy supplementing pulse supplements energy at the initial stage of plasma expansion to inhibit rapid cooling; the spectrometer synchronously collects plasma continuous radiation spectrums, two spectral lines of which the iron elements are 259.94nm and 260.07nm are selected, and the excitation temperature of the plasma induced by laser is calculated through a Boltzmann graph method, so that the target range is 8000-15000K; Measurement of hydrogen atom Baric System The Stark stretching degree of the wire is combined with the Stark stretching coefficient to calculate the electron density of the plasma generated by laser induction; the plasma excitation temperature and the plasma electron density are recorded as key parameters of the plasma; and outputting a plasma time resolution spectrum and plasma key parameters.
  4. 4. The mining all-element intelligent online detection and analysis method based on the laser-induced breakdown spectroscopy of claim 1 is characterized in that the self-adaptive spectrum acquisition is carried out based on the plasma time-resolved spectroscopy and plasma key parameters, and the background suppression is carried out on a mining scene, so that a net characteristic spectrum after background removal is obtained, and the specific analysis process is as follows: Obtaining a mineral aggregate matrix type; determining an optimal spectrum acquisition window based on the plasma excitation temperature; acquiring original spectrum intensity based on a plasma time resolution spectrum, and acquiring an instrument noise spectrum; and constructing a dynamic gradient background suppression formula containing self-adaptive weights, and performing background suppression on the mining scene to obtain a net characteristic spectrum after background removal.
  5. 5. The mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy of claim 4, wherein the construction of the dynamic gradient background suppression formula containing the self-adaptive weight is characterized in that background suppression is carried out on a mining scene to obtain a net characteristic spectrum after background removal, and the specific analysis process is as follows: constructing a dynamic gradient background suppression formula containing self-adaptive weights, wherein the specific calculation formula is as follows: ; Wherein, the In order to be a net characteristic spectrum of light, For the intensity of the original spectrum of light, For background spectra based on non-negative matrix factorization, Is a first order gradient of the background spectrum, For the spectrum of the instrument noise, The weight is suppressed for the background of the substrate, For the gradient background suppression weights, Is noise suppression weight; ; ; ; Wherein, the For the plasma excitation temperature, For the real-time ambient dust concentration, For the reference dust concentration stored in the database, Is the real-time environmental humidity; and outputting the net characteristic spectrum after background removal.
  6. 6. The mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy of claim 1, wherein the method is characterized in that a spectrum band attention-element category attention double-branch network is constructed based on a net characteristic spectrum and a mineral aggregate matrix type, and a high-dimensional compact characteristic vector, a characteristic attention weight graph and an element preliminary identification result are output, and the specific analysis process is as follows: Acquiring a marked mineral sample spectrum data set; Building a spectrum band attention-element category attention double-branch network: inputting a net characteristic spectrum, normalizing the input net characteristic spectrum, including mean subtraction and variance normalization, and smoothing; Learning the weight of each spectrum band through squeeze-extraction module, mapping the preprocessed net characteristic spectrum into a characteristic map, compressing the net characteristic spectrum into global characteristics through global average pooling, learning the weight of the band through a full-connection layer, and outputting the attention weight of each band through a sigmoid function; setting 4 convolution blocks, wherein each convolution block comprises 1 convolution layer, 1 BN layer, 1 ReLU activation function and 1 maximum pooling layer, the convolution layer kernel sizes are respectively 3×3, 5×5, 3×3 and 5×5, the output channel numbers are respectively 32, 64, 128 and 256, the maximum pooling layer kernel sizes are 2×2, the step length is 2, and the local peak shape and intensity of the spectrum are extracted; loading a corresponding element category attention matrix based on the mineral aggregate matrix type; After the convolution feature and the attention weight are fused, the dimension is reduced through a full connection layer, a feature vector F is output and recorded as a high-dimension compact feature vector, and a feature attention weight map and an element primary recognition result are output.
  7. 7. The mining all-element intelligent online detection analysis method based on the laser-induced breakdown spectroscopy of claim 1, wherein the characteristic weighting-plasma parameter correction fusion analysis is performed on the high-dimensional compact characteristic vector, the characteristic attention weight graph, the element primary identification result and the plasma key parameter to obtain the initial value of the quantitative content of each element and the confidence of the quantitative result, and the specific analysis process is as follows: Performing feature weighting-plasma parameter correction fusion analysis on the high-dimensional compact feature vector, the feature attention weight graph, the element primary identification result and the plasma key parameters, constructing a feature weighting-plasma parameter correction quantitative model, and calculating an element quantitative content initial value; Calculating the similarity between the net characteristic spectrum and the standard spectrum template of the same mineral species in the training set by adopting cosine similarity, and marking the similarity as confidence similarity; Calculating plasma key parameter stability based on a weighted average of the plasma key parameter relative standard deviation; carrying out weighted summation on the confidence similarity and the stability of the plasma key parameters to obtain a quantitative result confidence factor; acquiring a quantitative result confidence factor-quantitative result confidence coefficient mapping table stored in a database, and determining matched quantitative result confidence coefficient based on the current quantitative result confidence factor; outputting the initial value of the quantitative content of each element and the confidence of the quantitative result.
  8. 8. The mining all-element intelligent online detection and analysis method based on the laser-induced breakdown spectroscopy of claim 7 is characterized in that the characteristic weighting-plasma parameter correction quantitative model is constructed, the initial value of the quantitative element content is calculated, and the specific analysis process is as follows: ; Wherein, the Is the initial value of the quantitative content of the element, Weighting the regression coefficient matrix for the features stored in the database, For a high-dimensional compact feature vector, For the bias vector stored in the database, For the difference between the current plasma ignition temperature and the reference plasma ignition temperature stored in the database, As the difference between the current plasma electron density and the reference plasma electron density stored in the database, For the plasma excitation temperature correction coefficients stored in the database, Correcting coefficients for plasma electron density stored in a database; Outputting the initial quantitative content of the element.
  9. 9. The mining all-element intelligent online detection and analysis method based on the laser-induced breakdown spectroscopy of claim 1 is characterized in that the combination of quantitative result confidence, high-dimensional compact feature vector and net feature spectrum is used for carrying out mining all-element detection and correction, and outputting corrected final element content and mining species identification result, wherein the specific analysis process is as follows: Constructing a mineral seed fingerprint library, wherein the mineral seed fingerprint library specifically comprises spectrum fingerprints and element proportion fingerprints, and carrying out mineral seed automatic identification and mineral mixing proportion calculation: fingerprint matching and feature vector classification double-verification strategies are adopted: performing correlation matching on the net characteristic spectrum and spectrum fingerprints in a fingerprint library, if the pearson correlation coefficient is not lower than 0.85, and the matching is successful, inputting a high-dimensional compact characteristic vector into a trained SVM mineral seed classifier to obtain a classification result, and outputting a mineral seed name if the net characteristic spectrum and the spectrum fingerprints are consistent; if the net characteristic spectrum is matched with a plurality of ore types represented by spectral fingerprints in an ore type fingerprint library, calculating the ore mixing proportion based on the element proportion fingerprints; and outputting corrected final element content and ore species identification results.
  10. 10. The mining all-element intelligent online detection and analysis method based on the laser-induced breakdown spectroscopy of claim 9 is characterized in that the construction of the mineral fingerprinting library comprises the following specific analysis processes: establishing a standardized fingerprint library, collecting spectral fingerprints and element proportion fingerprints of mineral seeds detected in five years in history, and recording the spectral fingerprints and the element proportion fingerprints as a mineral seed fingerprint library; The spectral fingerprint is a characteristic spectral template for detecting mineral seeds in five years, the characteristic peak position and the intensity ratio of key elements are contained, and the element proportion fingerprint is a typical element content proportion for detecting mineral seeds in five years.

Description

Mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy Technical Field The invention relates to the technical field of spectrum online detection, in particular to a mining full-element intelligent online detection analysis method based on laser-induced breakdown spectroscopy. Background Currently, modern mining industry is developed to be refined and efficient, core indexes such as ore grade are required to be mastered in real time, dynamic adjustment of processes such as exploitation and ore dressing is supported, and resource utilization rate and economic benefit are improved. Real-time and comprehensive element data are needed for intelligent construction of mining industry, support is provided for intelligent control and digital twinning, and automation and few humanization of production processes are promoted. At present, research on intelligent online detection and analysis of all elements in mining industry has some defects, and the problems that the traditional detection technology has inherent short plates such as complex sample pretreatment, long time consumption, low light element detection precision, incapability of online detection and the like are embodied, so that the requirements of the industry on quick, nondestructive, all-element and low-pollution detection are difficult to meet. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a mining all-element intelligent online detection and analysis method based on laser-induced breakdown spectroscopy, which can effectively solve the problems related to the background art. The mining full-element intelligent online detection analysis method based on the laser-induced breakdown spectroscopy comprises the following steps of collecting mining online mineral flow, initial laser parameters, real-time environment parameters and mineral preprocessing information, conducting pulse coding laser-induced plasma optimization, outputting plasma time resolution spectroscopy and plasma key parameters, conducting self-adaptive spectroscopy collection based on the plasma time resolution spectroscopy and the plasma key parameters, conducting background suppression on a mining scene to obtain a net characteristic spectrum after background removal, constructing a spectral band attention-element category attention double-branch network based on the net characteristic spectrum and the mineral substrate type, outputting a high-dimensional compact feature vector, a feature attention weight map and an element primary recognition result, conducting feature weighting-plasma parameter correction fusion analysis on the high-dimensional compact feature vector, the feature attention weight map and the element primary recognition result and the plasma key parameters, obtaining quantitative content initial values and quantitative confidence of all elements, combining the quantitative confidence level, the high-dimensional compact feature vector and the net feature spectrum, conducting mining full-element detection correction, and outputting final element content and mineral seed recognition result after correction. The method comprises the steps of collecting mining online mineral flows, initial laser parameters, real-time environment parameters and mineral preprocessing information, wherein the mining online mineral flows comprise mineral belt conveying speed, mineral granularity and mineral water content, the initial laser parameters comprise fundamental frequency, pulse energy range, pulse width and repetition frequency, the real-time environment parameters comprise real-time environment temperature T, real-time environment humidity and real-time environment dust concentration, and the mineral preprocessing information comprises mineral primary screening granularity and mineral surface roughness. The method comprises the steps of carrying out pulse coding laser induced plasma optimization, outputting a plasma time resolution spectrum and plasma key parameters, wherein the specific analysis process comprises the steps of carrying out genetic algorithm on mineral online mineral flow, initial laser parameters, real-time environment parameters and mineral aggregate pretreatment information, taking the maximization of signal-to-noise ratio and the optimal stability of plasma excitation temperature generated by laser induction as targets, converting the targets into a specific three-pulse coding sequence, obtaining an optimized coding sequence, carrying out real-time shooting on the mineral aggregate surface by a high-speed linear array camera, combining a laser ranging module, calculating the fluctuation height of the mineral aggregate surface, and monitoring the offset of the mineral aggregate flow on a belt by visual positioning; the piezoelectric ceramic driving dynamic lens group is used for determining the lens position which is stored in a database and corresponds to the surface re