CN-122023353-A - Electromagnetic tomography-based real-time monitoring method and system for components in metallurgical mineral conveying process

CN122023353ACN 122023353 ACN122023353 ACN 122023353ACN-122023353-A

Abstract

The invention provides a method and a system for monitoring components in a metallurgical mineral conveying process in real time based on electromagnetic tomography. The method comprises the steps of adopting an EMT sensor array which is excited by multiple frequencies and arranged eccentrically to carry out non-contact scanning next to a conveyor belt, combining an electromagnetic chromatography image reconstruction algorithm and a deep learning method to construct a mineral-electromagnetic feature mapping database, carrying out real-time mineral pixel level identification through a trained machine learning model, dividing an identification result according to grid areas, combining pre-calibrated mineral density and area weight factors corrected by belt curvature, calculating real-time mass ratio data of each mineral on a material section, and carrying out signal drift compensation by combining high-precision temperature sensor data. Finally, the real-time mass ratio data is converted into control signals, and downstream equipment such as a sorting mechanical arm and a batching valve are driven, so that the on-line identification, accurate quantification and intelligent regulation and control of mineral components are realized, and the resource utilization efficiency and the automation level in the metallurgical process are remarkably improved.

Inventors

Xiao qingtai
ZHANG XIAOXUE
XU JIANXIN
YANG KAI
WANG YUFENG
WANG HUA

Assignees

昆明理工大学

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. A method for monitoring components of a metallurgical mineral conveying process in real time based on electromagnetic tomography, which is characterized by comprising the following steps: respectively carrying out multi-frequency EMT scanning on mineral types related to a metallurgical process to obtain original data of electromagnetic field disturbance; Converting the original data of electromagnetic field disturbance into electromagnetic characteristic space distribution images of each mineral by using an electromagnetic chromatographic image reconstruction algorithm, extracting corresponding electromagnetic characteristic space-time variation feature vectors from the electromagnetic characteristic space distribution images of each mineral by using a deep learning method, and constructing a mineral-electromagnetic feature mapping database; Based on a multi-frequency eccentric type EMT sensor array embedded with a temperature sensor, acquiring electromagnetic field disturbance data and temperature data of a mixed material on a mineral conveyor belt in real time, and acquiring electromagnetic characteristic space distribution images and electromagnetic characteristic space-time variation feature vectors of the mixed material through an electromagnetic chromatographic image reconstruction algorithm and a deep learning method which are optimized in real time; Matching the electromagnetic characteristic space distribution image and the electromagnetic characteristic space-time variation feature vector of the mixed material with a database through a trained machine learning model, and identifying mineral types of each space position of the mixed material based on pixel-level feature comparison; Dividing an electromagnetic characteristic space distribution image of the mixed material into grid areas based on a mineral type identification result, calling a pre-calibrated mineral density for each grid area, and calculating and outputting real-time mass ratio data of each mineral on a cross section of the mixed material; And converting the real-time mass ratio data into control signals to drive the sorting mechanical arm and the dispensing valve executing equipment to regulate and control the metallurgical process in real time.
2. The method of claim 1, wherein the method of converting the raw data of the electromagnetic field disturbance into the spatially distributed image of the electromagnetic properties of each mineral using an electromagnetic tomography image reconstruction algorithm comprises: ; Wherein, the As a function of the object to be processed, For the electromagnetic property profile to be sought, In order to measure the vector quantity, For forward operators solved by discretization of the finite element method or boundary element method, In order for the parameters to be regularized, In order to regularize the matrix, The L 2 norm of the vector is represented.
3. The method of claim 1, wherein the multi-frequency eccentric EMT sensor array comprises: 8 to 16 groups of exciting coils and 32 to 64 detection electrodes, wherein the exciting coils and the detection electrodes are arranged in an arc shape, and the curvature radius of the exciting coils and the detection electrodes is as follows The method meets the following conditions: ; In the formula, For the width of the mineral conveyor belt, Is an arc wrap angle, the shape of the wrap angle is an arc wrap angle, ; The circle centers of the arc-shaped arrangement are offset by 10% of the width of the mineral conveyor belt relative to the central line of the mineral conveyor belt; The working frequency of the exciting coil covers the frequency range from 10kHz to 1 MHz; The temperature sensors are embedded in the multi-frequency eccentric type EMT sensor array structure in a distributed mode or are close to the multi-frequency eccentric type EMT sensor array structure in a distributed mode and are used for monitoring the temperature of the sensor body structure and the temperature of the adjacent material area in real time.
4. The method according to claim 1, wherein the method for acquiring electromagnetic field disturbance data and temperature data of the mixed material on the mineral conveyor belt in real time and acquiring electromagnetic characteristic space distribution images and electromagnetic characteristic space-time variation feature vectors of the mixed material in real time through an electromagnetic tomography image reconstruction algorithm and a deep learning method which are optimized in real time comprises the following steps: ; Wherein, the For the voltage after the signal drift compensation, For the original measured voltage, the voltage is measured, As the current temperature is set to be the current temperature, In order to calibrate the temperature of the environment, For calibrating the system temperature compensation coefficient obtained by the experiment, Is a linear drift compensation coefficient.
5. The method of claim 1, wherein matching the electromagnetic property spatial distribution image and the electromagnetic property spatiotemporal variation feature vector of the mixture obtained in real time with the database, and identifying the mineral category to which each spatial position of the mixture belongs based on the pixel-level feature comparison comprises: extracting a local space feature vector and a global context vector of a target pixel, and fusing the local space feature vector and the global context vector into a comprehensive feature vector; the comprehensive feature vector is input into two paths in parallel by adopting a dual-path collaborative decision, wherein the first probability distribution is obtained through a classifier, and the second probability distribution is obtained through similarity matching with a database; weighting and fusing the first probability distribution and the second probability distribution to obtain final probability; Determining a mineral type label of the target pixel according to the final probability; after traversing all pixels, outputting a final mineral type label graph to realize the identification of the mineral types of each spatial position of the mixed material.
6. The method of claim 1, wherein the step of calculating real-time mass fraction data for each mineral in the cross section of the mixture comprises: ; Wherein, the For real-time mass-to-duty data, And Is a grid area Middle minerals And Is used for the number of pixels of a picture, And Is mineral And Is used for the density of the (c) in the (c), Is a grid area Is used for the area weight factor of (a), For the total number of mineral species, N is the grid area M is the grid area Is a column number of columns.
7. A metallurgical mineral conveying process component real-time monitoring system based on electromagnetic tomography, which is used for realizing the method of any one of claims 1-6, and is characterized by comprising a multi-frequency eccentric type EMT sensor array module, a multi-channel data acquisition and preprocessing module, an image reconstruction and feature extraction server module, a real-time mineral identification and component quantization module and a control instruction generation and execution interface module; The multi-frequency eccentric type EMT sensor array module is arranged on the detection side of the mineral conveyor belt or the material transmission channel, is closely adjacent to the belt surface in a non-contact mode and is used for outputting electromagnetic field disturbance data and temperature data of the mixed material; the multi-channel data acquisition and preprocessing module is used for driving the multi-frequency eccentric type EMT sensor array module to work and conditioning, acquiring and preprocessing electromagnetic field disturbance data and temperature data of the mixed materials; The image reconstruction and feature extraction server module is used for receiving the preprocessed data, performing image reconstruction and feature extraction operations, and obtaining electromagnetic characteristic space distribution images and electromagnetic characteristic space-time variation feature vectors of the mixed materials; The real-time mineral identification and component quantization module is used for receiving electromagnetic characteristic space distribution images and electromagnetic characteristic space-time variation feature vectors of the mixed materials, identifying mineral types, dividing the electromagnetic characteristic space distribution images of the received mixed materials into grid areas according to identification results, calling pre-calibrated mineral density, and calculating real-time mass ratio data of each mineral on the cross section of the mixed materials; The control instruction generation and execution interface module is used for converting the mineral identification result and the real-time mass ratio data into control signals and driving downstream execution equipment to realize real-time regulation and control of the metallurgical process.
8. The system of claim 7, wherein the multi-channel data acquisition and preprocessing module comprises a controllable multi-frequency signal source, an integrated signal conditioner, a high-precision multi-channel synchronous data acquisition card and a preprocessing unit; the controllable multi-frequency signal source is used for outputting adjustable excitation current so as to drive the multi-frequency eccentric EMT sensor array module to work; The integrated signal conditioner comprises a low-noise amplifier, an analog filter and a phase-locked amplifier, and is used for amplifying, filtering and demodulating the induction signals output by the multi-frequency eccentric EMT sensor array module; The high-precision multichannel synchronous data acquisition card is used for acquiring electromagnetic field disturbance data and temperature data after demodulation processing in real time; And the preprocessing unit is used for carrying out noise reduction, normalization and format conversion operation on the electromagnetic field disturbance data and the temperature data.
9. The system of claim 7, wherein the image reconstruction and feature extraction server module comprises an electromagnetic tomography reconstruction engine, a feature extraction unit, and a temperature drift compensation unit; the electromagnetic tomography reconstruction engine is used for receiving the preprocessed data, running a real-time optimized electromagnetic tomography reconstruction algorithm and obtaining an electromagnetic characteristic space distribution image of the mixed material; The feature extraction unit is used for extracting features of the electromagnetic property space distribution image of the mixed material by adopting a deep learning method to generate electromagnetic property space-time variation feature vectors of the mixed material; And the temperature drift compensation unit is integrated with an electromagnetic characteristic temperature drift compensation model based on temperature data and is used for correcting electromagnetic characteristic space distribution images or electromagnetic characteristic space-time change characteristic vectors of the mixed materials.
10. The system of claim 7, wherein the real-time mineral identification and component quantification module comprises a mineral identification unit and a component quantification unit; The mineral identification unit is used for comparing the electromagnetic characteristic space distribution image and the electromagnetic characteristic space-time variation characteristic vector of the mixed material with a pre-stored database based on a trained machine learning model, and identifying the mineral type corresponding to each space position in the electromagnetic characteristic space distribution image of the mixed material; And the component quantization unit is used for dividing the electromagnetic characteristic space distribution image of the mixture into grid areas, calling the pre-calibrated mineral density and calculating and outputting real-time mass ratio data of the cross section of the mixture.

Description

Electromagnetic tomography-based real-time monitoring method and system for components in metallurgical mineral conveying process Technical Field The invention belongs to the technical field of metallurgical mineral processing, and particularly relates to a method and a system for monitoring components in a metallurgical mineral conveying process in real time based on electromagnetic tomography, which are suitable for component analysis in a continuous conveying process of non-magnetite materials such as copper mine, iron mine, coke and the like. Background In the metallurgical, mining and resource processing processes, accurate acquisition of the spatial distribution and mass ratio of various mineral components in materials is a key for realizing intelligent ore matching and automatic separation. The traditional method relies on manual sampling and laboratory analysis, and has the problems of long detection period, low efficiency, poor real-time performance and the like. In order to meet the demands of on-line detection in industrial sites, there is a need to develop a mineral composition monitoring technology with high spatial-temporal resolution, non-contact and dynamic imaging. Electromagnetic tomography (Electromagnetic Tomography, EMT) is used as an imaging technology for reconstructing the distribution of the conductivity, the permittivity and the permeability in the object by utilizing the response of a multi-frequency electromagnetic field, has the characteristics of non-invasiveness, high sensitivity and multi-physical parameter sensing, and is particularly suitable for on-line monitoring of complex multiphase media such as mineral mixtures and the like. Disclosure of Invention The invention provides a method and a system for monitoring components in a metallurgical mineral conveying process in real time based on electromagnetic tomography, aiming at realizing the identification of various mineral types in mixed materials and the real-time calculation of mass ratio and linking downstream equipment to realize intelligent sorting and regulation. In order to achieve the above object, the present invention provides the following solutions: A method for real-time monitoring of metallurgical mineral delivery process components based on electromagnetic tomography, the method comprising: respectively carrying out multi-frequency EMT scanning on mineral types related to a metallurgical process to obtain original data of electromagnetic field disturbance; Converting the original data of electromagnetic field disturbance into electromagnetic characteristic space distribution images of each mineral by using an electromagnetic chromatographic image reconstruction algorithm, extracting corresponding electromagnetic characteristic space-time variation feature vectors from the electromagnetic characteristic space distribution images of each mineral by using a deep learning method, and constructing a mineral-electromagnetic feature mapping database; Based on a multi-frequency eccentric type EMT sensor array embedded with a temperature sensor, acquiring electromagnetic field disturbance data and temperature data of a mixed material on a mineral conveyor belt in real time, and acquiring electromagnetic characteristic space distribution images and electromagnetic characteristic space-time variation feature vectors of the mixed material through an electromagnetic chromatographic image reconstruction algorithm and a deep learning method which are optimized in real time; Matching the electromagnetic characteristic space distribution image and the electromagnetic characteristic space-time variation feature vector of the mixed material with a database through a trained machine learning model, and identifying mineral types of each space position of the mixed material based on pixel-level feature comparison; Dividing an electromagnetic characteristic space distribution image of the mixed material into grid areas based on a mineral type identification result, calling a pre-calibrated mineral density for each grid area, and calculating and outputting real-time mass ratio data of each mineral on a cross section of the mixed material; And converting the real-time mass ratio data into control signals to drive the sorting mechanical arm and the dispensing valve executing equipment to regulate and control the metallurgical process in real time. Preferably, the method for converting the raw data of the electromagnetic field disturbance into the electromagnetic property spatial distribution image of each mineral by using an electromagnetic tomographic image reconstruction algorithm comprises: ; Wherein, the As a function of the object to be processed,For the electromagnetic property profile to be sought,In order to measure the vector quantity,For forward operators solved by discretization of the finite element method or boundary element method,In order for the parameters to be regularized,In order to regularize the matrix,The L 2 n