Ore sorting system and method based on deep learning

Abstract

The invention discloses an ore sorting system and method based on deep learning, and the method comprises the steps: carrying out particle size grading and ore washing treatment on raw ore through a plurality of layers of vibrating screens to obtain pretreated graded ore, and collecting image data and hyperspectral data of the pretreated graded ore; respectively extracting surface features of the image data and spectral features of the hyperspectral data by using a first network and a second network constructed based on a deep learning algorithm, fusing the surface features and the spectral features, and inputting the fused features into a classification network to obtain ore types and ore principal components; and air injection parameters are calculated according to the conveyor belt speed, the ore mass and the air density, a concentrate separation instruction is determined, and ore separation is executed according to the air injection parameters and the separation instruction. The invention relates to the technical field of ores, and solves the technical problems of low treatment efficiency and insufficient separation accuracy of an existing ore separation method.

Description

Technical Field

[0001] The present invention belongs to the field of mining and relates to deep learning technology, specifically to an ore sorting system and method based on deep learning. Background Art

[0002] Mineral resources are the basic materials for national economic construction. With the rapid development of the economy, the demand for them is growing rapidly, and efficient mining technology has become increasingly important. Ore sorting is a key link in the mining industry, and its accuracy and efficiency have a significant economic impact on the mining industry.

[0003] Traditional mechanical sorting is mainly based on the separation of ore and gangue (waste rock) based on the differences in physical properties (such as density, magnetism, conductivity, hydrophobicity, etc.). Common methods include gravity separation, flotation, magnetic separation, and electrostatic separation. There are generally problems such as strong dependence on the physical properties of the ore, complex processes, large equipment investment, and lack of environmental friendliness. It is also difficult to adapt to the needs of efficient sorting of low-grade and complex co-existing ores.

[0004] With the development of computer technology and image processing technology, image recognition technology has gradually been applied to the field of ore sorting. For example, by collecting mineral images, using convolutional neural networks to realize mineral recognition and sorting, or using support vector machines to sort single ore. However, such methods usually have strict requirements on the original image data, and can often only sort specific types of ores, with a narrow scope of application. Moreover, since the characteristic information of the samples to be sorted is sometimes difficult to distinguish, the sorting results lack interpretability and have a low accuracy rate.

[0005] At the same time, when determining the processing flow of the concentrate, the traditional method is usually to analyze the components of the selected concentrate after ore sorting, and determine the subsequent processing flow according to the content of various elements in the concentrate, mineral composition and other characteristics, combined with the requirements of the target product. If the concentrate has a high content of a certain valuable metal and few impurities, it may be directly smelted and further processed; if the concentrate contains multiple valuable components or has a high impurity content, it may need to go through multiple complex purification and separation processes, such as chemical leaching, flotation and other methods for further processing to improve the purity and quality of the concentrate and meet the needs of different industrial production. However, when determining the concentrate processing flow, traditional methods often rely on manual experience and conventional analytical methods, lack a comprehensive and accurate grasp of ore characteristics, resulting in the processing flow may not be optimized enough, affecting production efficiency and economic benefits. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an ore sorting system and method based on deep learning, which are used to solve the technical problems of low processing efficiency and insufficient sorting accuracy of existing ore sorting methods.

[0007] To achieve the above objectives, the first aspect of the present invention provides an ore sorting system based on deep learning, comprising: Data acquisition module: used to classify and wash the raw ore through several layers of vibrating screens to obtain pre-treated classified ore, and to collect image data and hyperspectral data of the pre-treated classified ore; Ore classification module: used to extract the surface features of image data and the spectral features of hyperspectral data respectively by using the first network and the second network constructed based on the deep learning algorithm, and then fuse the surface features and the spectral features and input them into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; Sorting execution module: used to calculate the jet parameters according to the conveyor belt speed, ore quality and air density, and determine the sorting instructions for the concentrate, and execute ore sorting according to the jet parameters and sorting instructions.

[0008] It should be noted that the data acquisition module, the ore classification module and the sorting execution module of the present invention are in communication connection.

[0009] Furthermore, the acquisition of pre-processed image data and hyperspectral data of the graded ore includes: The pre-treated graded ore is spread on the conveyor belt by using an electromagnetic vibrating feeder, and the conveyor belt is set to convey at a uniform speed at a preset speed; Industrial cameras and hyperspectral cameras are used to obtain image data and hyperspectral data of graded ores at preset shooting intervals.

[0010] Furthermore, the first network includes a multi-scale residual pyramid module, a dynamic convolutional attention module and a Transformer encoder which are sequentially constructed; wherein, The input of the multi-scale residual pyramid module is image data, which includes several levels of dilated convolution branches composed of dilated convolution and dilated spatial pyramid pooling layers in parallel, and the outputs of several levels of dilated convolution branches are aggregated through bilinear interpolation upsampling channel splicing and 1×1 convolution layer to obtain the aggregated feature F agg ; The input of the dynamic convolution attention module is the aggregated feature, which includes a parameter generation network composed of a global average pooling layer, a 1×1 convolution layer, and a ReLU function, and a dynamic kernel convolution module composed of a depth-separable convolution and a residual connection, to obtain the attention enhancement feature, which is input into the Transformer encoder to obtain the surface feature. .

[0011] Furthermore, the formula of the dynamic convolution attention module is: The convolution kernel parameter K is obtained through the parameter generation network: K=f(1) 1×1 (ReLU(f (0) 1×1 (GAP(F agg ))))); Among them, GAP( ) represents the global average pooling layer, f (0) 1×1 ( ) represents the first 1×1 convolutional layer, f (1) 1×1 ( ) represents the second 1×1 convolutional layer, ReLU( ) represents the ReLU function; The dynamic kernel convolution module is used to perform deep separable convolution and residual linking on the convolution kernel parameters and aggregate features to obtain the attention enhancement feature: Y out =F agg +α×reshape(f g (reshape(F agg ),K)); where reshape( ) represents the dimension reshaping operation, f g ( ) represents a depth-wise separable convolution, and α represents a learnable parameter.

[0012] The first network adopts a multi-scale residual pyramid module, which extracts multi-scale texture features through parallel connection of dilated convolution and void space pyramid pooling layer, and combines bilinear interpolation and 1×1 convolution aggregation features to effectively retain the details of the ore surface; the dynamic convolution attention module uses global average pooling to generate dynamic convolution kernel parameters, combines depthwise separable convolution and residual connection to achieve dynamic focusing on key areas of the ore surface and enhance feature discriminability; the Transformer encoder further captures global context dependency and improves the feature expression capability under complex ore surface conditions.

[0013] Furthermore, the second network includes a spectral band grouping module, a local-global spectral attention module and a three-dimensional convolutional feature compression layer; wherein, The input of the spectral band grouping module is hyperspectral data, including using the K-means algorithm to divide several spectral bands of the spectral data into a preset number of groups to obtain several band subsets, and then inputting the several band subsets into the three-dimensional convolution layer and the pooling layer in turn to perform spectral dimension splicing, and generating intermediate features F through the spectral normalization layer. group ; The local-global spectral attention module includes extracting the local band correlation of the intermediate features using three-dimensional convolution and GELU activation function to obtain the local feature F local, and input the local features into the multi-head self-attention mechanism: , and obtain the global feature F global , where Q, K, and V represent the query matrix, key matrix, and value matrix obtained by the linear projection layer of local features, T represents the transposition operation, and d represents the scaling factor. global The number of channels is determined; The three-dimensional convolution feature compression layer includes processing the global features using two-dimensional convolution and three-dimensional convolution, and then inputting them into the deformable three-dimensional convolution layer and the global maximum pooling layer in sequence to obtain the spectral features. .

[0014] The second network divides the band subsets through K-means clustering, and combines three-dimensional convolution and pooling layers to reduce the dimensional redundancy of the hyperspectral features of the ore. The local-global spectral attention module extracts local band correlations through three-dimensional convolution, and uses the multi-head self-attention mechanism to model cross-band global relationships, which can enhance the semantic expression of spectral features. The three-dimensional convolution feature compression layer uses deformable convolution and global pooling, which can flexibly adapt to spectral response differences and compress feature dimensions, thereby improving the processing efficiency of ore hyperspectral feature extraction.

[0015] Furthermore, the fusion of the surface features and the spectral features and inputting them into the classification network comprises: S1, establishes semantic associations between surface features and spectral features through a bidirectional cross-attention mechanism to obtain joint features ; Among them, the bidirectional cross attention mechanism includes: The surface feature is used as the query vector, the spectral feature is used as the key vector and the value vector, and the first fusion feature is calculated according to the cross attention calculation formula; Using surface features as query vectors , the spectral features are used as key vectors Sum value vector , the first fusion feature is calculated according to the attention calculation formula : ; Use spectral features as query vectors , surface features as key vectors Sum value vector , the second fusion feature is calculated according to the attention calculation formula : ; The first fusion feature and the second fusion feature are concatenated in the channel dimension to obtain the joint feature ; S2, the joint features, surface features, and spectral features are combined using a gating network to generate a gating weight G, and the gating weight is element-wise multiplied with the joint features to obtain the optimized fusion feature ; Among them, the mathematical expression of the gating network is: ,σ( ) represents the Sigmoid function, δ( ) represents the GELU activation function, represents the feature concatenation operation, W1 and W2 represent the preset learnable parameter matrices; S3, inputting the optimized fusion features into the classification network to obtain the main components and ore types of the ore; wherein the classification network includes: Use normalization layers, fully connected layers, and random dropout layers to extract shared features from optimized fusion features , the mathematical expression is: ,LayerNorm( ) represents the normalization layer, Linear( ) represents the fully connected layer, Dropout( ) represents a random dropout layer; The classification features in the shared features are extracted using the gradient reversal layer and the domain adversarial network, and then the classification features are input into the ore type classifier containing the fully connected layer to output the probability distribution P1 of the predicted ore type; the ore types include concentrate, middlings and tailings; The multi-layer perceptron MLP and Softmax function are used to calculate the attention weight of the optimized fusion feature, and element-by-element product is performed with the optimized fusion feature to obtain the enhanced main feature. The enhanced main feature is input into the ore component classifier containing a fully connected layer, and the probability distribution P2 of the ore main component is output; among them, the ore main components include Fe2O3, Fe3O4, SiO2, Al2O3, CuFeS2, FeS2, etc.

[0016] The classification network establishes semantic associations between image and spectral features through a bidirectional cross-attention mechanism to generate joint features of the graph; and uses a gating network to dynamically adjust the weights of joint features, surface features, and spectral features to optimize the fusion results; finally, in the classifier, the data processing efficiency is improved by extracting shared features, and then the domain adversarial network is used to enhance the discriminability of classification features, and the multi-layer perceptron is used to enhance the expression of principal component features through attention weights, respectively processing the ore type and principal component prediction tasks to achieve intelligent ore classification.

[0017] Furthermore, the jet parameters are calculated according to the conveyor belt speed, ore quality and air density, including: Use Kalman filter to predict the movement trajectory of ore and get the time when ore reaches the jet position ; According to the formula Calculate the jet delay time ;in, Indicates the physical distance from the detection position to the injection valve, v b Indicates the conveyor belt speed, Indicates the solenoid valve response time, Indicates the time consumption of the ore classification module; Time for the ore to reach the jet position With jet delay time Subtract and get the jet trigger time ; Set up sorting channels according to ore type and set the sorting channel spacing according to Calculate the target lateral velocity , the calculation formula is: ; Where h represents the falling height and g represents the acceleration due to gravity; According to the formula The jet duration is calculated; where, Indicates the quality of the ore, represents the air density, represents the nozzle cross-sectional area, Indicates the nozzle flow coefficient.

[0018] Furthermore, the determining of the sorting instructions for the concentrate includes: Obtain the physical and chemical property value A of the main component i of the concentrate according to the ore characteristics database i ; chemical properties include magnetism, density, hydrophobicity, conductivity and dielectric constant; Set the economic weight E of the main component i of the concentrate i ; According to the formula CII i =P i ×[β×A i +(1-β)×E i ] Calculate the physical and chemical property influence index CII of the main component i of the ore i ; Among them, P i represents the probability value of the main component i of the ore, which is obtained by the output of the classification network, and β represents the preset influence coefficient; According to the formula Calculate the concentration separation method adaptation score S k ; where k represents the sorting method index, represents the applicability function of sorting method k to ore principal component i, and n represents the total number of ore principal components; The adaptation score S k The method that is greater than the preset threshold is used as the concentrate sorting method to obtain the concentrate sorting instruction.

[0019] Furthermore, the applicability function of the separation method k to the main component i of the ore includes: Magnetic separation suitability function: ; where k1 represents the slope factor, represents the magnetic susceptibility of principal component i, Indicates the preset magnetic susceptibility threshold; Flotation suitability function: ; where γ represents the hydrophobicity gain coefficient, represents the hydrophobic contact angle of principal component i, represents the surface potential of the principal component i, represents the critical value of the surface potential of the principal component i, represents the potential sensitivity factor; Reselect the suitability function: ;in, represents the density difference between the principal component i and the separation medium, Indicates the density distribution dispersion of the ore, represents the density difference gain coefficient; Electroselection suitability function: ;in, represents the conductivity of the main component i, represents the dielectric constant of principal component i, , represent the reference conductivity and reference dielectric constant respectively, and x represents the nonlinear coefficient.

[0020] The Kalman filter is used to predict the trajectory of the ore, and the jet delay time and duration are dynamically calculated in combination with the conveyor belt speed, ore quality and air density, which can ensure the accuracy of the ore sorting action; and in the subsequent sorting process of the concentrate, based on the physicochemical properties and economic weights of the main components of the concentrate, the adaptability score is calculated through the suitability function (such as magnetic separation, flotation, etc.), and the optimal sorting method is selected, which can improve the resource recovery rate and sorting targeting.

[0021] A second aspect of the present invention provides an ore sorting method based on deep learning, comprising: The raw ore is subjected to particle size classification and ore washing treatment through several layers of vibrating screens to obtain pre-treated classified ore, and image data and hyperspectral data of the pre-treated classified ore are collected; The first network and the second network constructed based on the deep learning algorithm are used to extract the surface features of the image data and the spectral features of the hyperspectral data respectively, and the surface features and the spectral features are fused and input into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; The jet parameters are calculated based on the conveyor belt speed, ore quality and air density, and the sorting instructions for the concentrate are determined. Ore sorting is performed based on the jet parameters and sorting instructions.

[0022] Compared with the prior art, the present invention has the following beneficial effects: The present invention improves the accuracy and efficiency of ore sorting through multimodal data fusion and deep feature extraction technology. First, the image and hyperspectral data of the ore are collected, and the surface texture and spectral component features are extracted respectively by constructing a dual network structure, and cross-modal semantic association is achieved through a two-way cross-attention mechanism to enhance the classification ability of complex ores (such as multi-mineral mixtures); in addition, the dynamic convolutional attention and gated network mechanism dynamically focus on key areas and adjust feature weights to effectively suppress noise interference, especially in scenes with lighting changes or ore surface contamination, it can show stronger robustness; in addition, the multi-scale residual pyramid and spectral band grouping technology can retain key information while reducing computational complexity, so that the ore classification process has the advantages of high precision and low latency in industrial scenarios; In terms of ore sorting strategy, Kalman filtering is used to predict the movement trajectory of the ore, and the jet delay time and duration are accurately calculated based on parameters such as conveyor belt speed and air density to ensure accurate triggering of the sorting action. At the same time, based on the physical and chemical properties of the main components of the ore, a suitability function is constructed to dynamically select the optimal concentrate sorting method, which can not only improve sorting efficiency and reduce sorting costs, but also specifically recover high-value minerals and improve resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A schematic diagram of the technical process of the deep learning-based ore sorting system provided by the present invention; Figure 2 A schematic diagram of the framework of the ore sorting system based on deep learning provided by the present invention; Figure 3 A schematic diagram of the technical process for determining concentrate separation instructions provided by the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] See also Figure 1-Figure 3The first aspect of the present invention provides an ore sorting system based on deep learning, comprising: Data acquisition module: used to classify and wash the raw ore through several layers of vibrating screens to obtain pre-treated classified ore, and to collect image data and hyperspectral data of the pre-treated classified ore; Ore classification module: used to extract the surface features of image data and the spectral features of hyperspectral data respectively by using the first network and the second network constructed based on the deep learning algorithm, and then fuse the surface features and the spectral features and input them into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; Sorting execution module: used to calculate the jet parameters according to the conveyor belt speed, ore quality and air density, and determine the sorting instructions for the concentrate, and execute ore sorting according to the jet parameters and sorting instructions.

[0027] In order to realize the intelligent ore sorting process, the ore must first be graded and pre-processed to eliminate impurities and ensure the uniformity of ore particles. In the data acquisition module, the raw ore needs to be physically screened through multiple layers of vibrating screens. Each layer of screens separates ore particles of different particle size ranges according to the preset aperture. At the same time, high-pressure water flow is used to remove impurities such as dirt and dust attached to the surface, forming clean and graded ore with consistent particle size. Then, the pre-treated graded ore will be spread on the conveyor belt through the electromagnetic vibrating feeder, and the conveyor belt will run stably at a preset uniform speed to ensure that the ore is continuously transported in a single layer without overlapping. The industrial camera and the hyperspectral camera are synchronously triggered to shoot at preset intervals: the industrial camera obtains two-dimensional grayscale or color image data of the ore surface to capture visual features such as texture and shape; the hyperspectral camera collects continuous spectral data covering hundreds of bands to record the spectral response information of the ore composition.

[0028] Then in the ore classification module, the first network (image feature extraction) and the second network (hyperspectral feature extraction) extract the surface features of the image data and the spectral features of the hyperspectral data respectively, and then fuse the surface features and spectral features and input them into the classification network (feature fusion and classification decision) to achieve high-precision identification of ore types and main components; The first network consists of a multi-scale residual pyramid module, a dynamic convolutional attention module, and a Transformer encoder to extract rich surface features from ore images. Specifically: 1. Multi-scale residual pyramid module: The input is preprocessed image data (size is H×W×C, H represents length, W represents width, and C represents the number of channels). By connecting multiple dilated convolution branches in parallel, each branch contains dilated convolution with different dilation rates (dilation rate r=2, 4, 8) and atrous spatial pyramid pooling layer (ASPP), multi-scale texture features are extracted; then the output of each branch is upsampled by bilinear interpolation and spliced ​​along the channel, and then reduced to the aggregate feature F through 1×1 convolution. agg :F agg =Conv1×1(Concat(Upsample(F (1) aspp ),…,Upsample(F (L) aspp ))); Among them, Conv1×1( ) represents 1×1 convolution, Concat( ) indicates channel splicing, Upsample( ) represents the bilinear interpolation upsampling operation, which upsamples the outputs of each branch of ASPP to the same resolution, F (1) aspp represents the first layer of dilated convolution branch, F (L) aspp Represents the L-th layer dilated convolution branch; The combination of dilated convolutions with different dilation rates (r=2, 4, 8) and the multi-level voiding rate of ASPP enables each branch to capture multi-scale features from local texture (small dilation rate) to global structure (large dilation rate). For example, the branch with dilation rate r=2 can detect fine cracks on the surface of ore, while the branch with r=8 can perceive the outline of large minerals; at the same time, traditional downsampling (such as pooling) will reduce image resolution and lose edge details, while the multi-scale residual pyramid module replaces pooling with dilated convolution to expand the receptive field while maintaining spatial resolution; the subsequent bilinear interpolation upsampling restores the output of each branch to the same resolution, which can avoid information loss caused by downsampling, and is particularly suitable for fine analysis of ore surface texture; 2. Dynamic Convolutional Attention Module Taking the aggregate feature F agg As input, the dynamic convolution kernel parameter K=f is generated through global average pooling (GAP) and 1×1 convolution. (1) 1×1 (ReLU(f (0) 1×1 (GAP(F agg ))))); Among them, GAP( ) represents the global average pooling layer, f (0) 1×1 ( ) represents the first 1×1 convolutional layer, f (1)1×1 ( ) represents the second 1×1 convolutional layer, ReLU( ) represents the ReLU function; Then, K and F agg Perform depth-wise separable convolution and combine with F agg Residual connection generates attention-enhanced features: Y out =F agg +α×reshape(f g (reshape(F agg ),K)); where reshape( ) indicates the operation of reshaping the data dimension, f g ( ) represents depth-wise separable convolution, and α represents a learnable parameter; By dynamically adjusting the convolution kernel parameters, the convolution operation can adaptively focus on key areas such as ore boundaries and texture differences. For example, when there is metallic luster on the ore surface, the dynamic kernel will strengthen the response to the highlight area; then the dynamic convolution result is added to the original feature, which can strengthen the feature response of the important area and suppress background noise; 3. Transformer Encoder Input the attention-enhanced features into Transformer, capture the global context dependency through the multi-head self-attention mechanism, and output the surface features , which strengthens the global characteristic expression of ore under complex background.

[0029] The second network consists of a spectral band grouping module, a local-global spectral attention module, and a three-dimensional convolutional feature compression layer, which is used to extract the spectral features of hyperspectral data; the specific network structure is as follows: 1. Spectral band grouping module The input hyperspectral data bands are divided into G groups by K-means algorithm; each group of bands is spliced ​​along the spectral dimension after three-dimensional convolution (kernel size is 3×3×1) and pooling processing to generate intermediate features F group : F group =Concat(Pool(Conv3D(Group1)),…,Pool(Conv3D(Group K ))); Among them, Pool( ) represents the pooling layer, Group i Represents the i-th group of bands, Conv3D( ) represents three-dimensional convolution; Hyperspectral data usually contains hundreds of bands, and direct processing will lead to dimensionality disaster. K-means is used to divide the bands into G groups, and the number of bands in each group is reduced to B / G (B represents the number of hyperspectral bands), which can significantly reduce the computational complexity of subsequent three-dimensional convolutions. After each group of three-dimensional convolutions, it is downsampled through the pooling layer and then spliced ​​along the spectral dimension to generate intermediate features. It can integrate the features of different groups in spectral order, such as splicing the features of metal oxide sensitive bands and silicate sensitive bands, promote cross-group feature interaction, and improve the effective use of band correlation. 2. Local-Global Spectral Attention Module First, the intermediate feature F is obtained by three-dimensional convolution and GELU activation function. group Extract local band correlation and obtain local feature F local : F local =GELU(Conv3D(F gruop )); The local features are then input into the multi-head self-attention mechanism to generate global features: , and obtain the global feature F global , where Q, K, and V represent the query matrix, key matrix, and value matrix obtained by the linear projection layer of local features, T represents the transposition operation, and d represents the scaling factor. global The number of channels is determined by ; through the collaboration of three-dimensional convolution and self-attention, the local spectral details and the global component association are taken into account, and the semantic expression of spectral features is enhanced; 3. 3D convolution feature compression layer After two-dimensional and three-dimensional convolution processing of global features, the convolution kernel sampling position is dynamically adjusted through deformable three-dimensional convolution to adapt to the spectral features, and the dimension is compressed through global maximum pooling to obtain the spectral features. .

[0030] The classification network realizes feature fusion and decision-making through a bidirectional cross-attention mechanism, a gating network, and a multi-task classifier; the specific processing flow is as follows: S1, establish the semantic association between surface features and spectral features through the bidirectional cross attention mechanism to obtain the joint feature F joint ; Among them, the bidirectional cross attention mechanism includes: Using surface features as query vectors , the spectral features are used as key vectors Sum value vector , the first fusion feature is calculated according to the attention calculation formula: ; Use spectral features as query vectors , surface features as key vectors Sum value vector , the second fusion feature is calculated according to the attention calculation formula: ; The first fusion feature and the second fusion feature are concatenated in the channel dimension to obtain the joint feature F joint ; S2, the joint features, surface features, and spectral features are combined using a gating network to generate a gating weight G, and the gating weight is element-wise multiplied with the joint features to obtain the optimized fusion feature F opt ; Among them, the mathematical expression of the gating network is: ,σ( ) represents the Sigmoid function, δ( ) represents the GELU activation function, represents the feature concatenation operation, W1 and W2 represent the preset learnable parameter matrices; S3, inputting the optimized fusion features into the classification network to obtain the main components and ore types of the ore; wherein the classification network includes: The shared features are extracted from the optimized fusion features using the normalization layer, the fully connected layer, and the random dropout layer. The mathematical expression is: shared =Dropout(δ(Linear(LayerNorm(F final )))), LayerNorm( ) represents the normalization layer, Linear( ) represents the fully connected layer, Dropout( ) represents a random dropout layer; The classification features in the shared features are extracted using the gradient reversal layer and the domain adversarial network, and then the classification features are input into the ore type classifier containing the fully connected layer to output the probability distribution P1 of the predicted ore type; the ore types include concentrate, middlings and tailings; The multi-layer perceptron MLP and Softmax function are used to calculate the attention weight of the optimized fusion feature, and element-by-element product is performed with the optimized fusion feature to obtain the enhanced main feature. The enhanced main feature is input into the ore component classifier containing a fully connected layer, and the probability distribution P2 of the ore main component is output; among them, the ore main components include Fe2O3, Fe3O4, SiO2, Al2O3, CuFeS2, FeS2, etc.

[0031] Through the first network (image feature extraction) and the second network (hyperspectral feature extraction) built based on the deep learning algorithm in the ore classification module, the system can efficiently integrate the multimodal information of ore surface texture and chemical composition, and output the probability distribution of ore types (concentrates, middlings, tailings) and main components (such as Fe2O3, CuFeS2, etc.).

[0032] It should be noted that the first network, the second network and the classification network constructed based on the deep learning algorithm can only be applied to this system after training and verification; and the training data of the first network are several ore images, and the labels include concentrates, middlings and tailings; the training data of the second network are the hyperspectral data of several ores, and the labels are predefined ore main component categories (such as Fe2O3, Fe3O4, etc.), covering the generalized categories of known main components and unknown components; the classification network is a downstream task of multimodal fusion, and the input is the fusion features of the image surface features extracted by the first network and the spectral features extracted by the second network, and the output includes the predicted labels of ore type and ore main components.

[0033] Then, in the sorting execution module, by calculating the jet parameters of different types of ores, using the Kalman filter to predict the ore movement trajectory, combined with the conveyor belt speed, the distance from the detection position to the jet valve and other information, the jet delay time, trigger time and jet duration are accurately determined, so that the ore is blown to the corresponding channel to achieve the initial separation of different types of ores. The concentrate is further analyzed, and the physical and chemical property influence index is calculated according to the physical and chemical property values ​​and economic weights of the main components of the concentrate ore. Combined with the sorting method adaptation score, a reference for the subsequent sorting method of the concentrate is given to achieve refined sorting of the concentrate, improve resource recovery efficiency and sorting accuracy, and make the ore sorting process more scientific and efficient.

[0034] Specifically, in one embodiment, the jet parameters are calculated based on the conveyor belt speed, ore mass and air density, which may include the following steps: The sorting execution module is responsible for the key task of converting the classification results into actual sorting operations in the entire ore sorting process. The specific operation steps are as follows: Set up separation channels for concentrate, middlings and tailings according to ore types; The Kalman filter is used to predict the motion trajectory in combination with the current motion state of the ore, so as to obtain the time when the ore reaches the jet position. ; According to the formula Calculate the jet delay time ;in, Indicates the physical distance from the detection position to the injection valve, v b Indicates the conveyor belt speed, Indicates the solenoid valve response time, Indicates the time consumption of the ore classification module; Then the ore reaches the jet position time With jet delay time Subtract and get the jet trigger time , ensuring that the jet device can jet the ore at the best time; According to the sorting channel spacing Calculate the target lateral velocity , the calculation formula is: ; Where h represents the falling height and g represents the acceleration due to gravity; According to the formula The jet duration is calculated; where, Indicates the quality of the ore, represents the air density, represents the nozzle cross-sectional area, It indicates the nozzle flow coefficient, which enables the duration of the jet to meet the requirements of blowing the ore to the specified channel.

[0035] In one embodiment, determining the sorting instructions for the concentrate may include the following steps: Obtain the physical and chemical property value A of the main component i of the concentrate according to the ore characteristics database i ; chemical properties include magnetism, density, hydrophobicity, conductivity and dielectric constant; Set the economic weight E of the main component i of the concentrate i ; According to the formula CII i =P i ×[β×A i +(1-β)×E i ] Calculate the physical and chemical property influence index CII of the main component i of the ore i ; Among them, P i represents the probability value of the main component i of the ore, which is obtained by the output of the classification network, and β represents the preset influence coefficient; According to the formula Calculate the concentration separation method adaptation score S k ; where k represents the sorting method index, The function representing the applicability of sorting method k to the main component i of ore includes: Magnetic separation suitability function: ; where k1 represents the slope factor, represents the magnetic susceptibility of principal component i, Indicates the preset magnetic susceptibility threshold; Flotation suitability function: ; where γ represents the hydrophobicity gain coefficient, represents the hydrophobic contact angle of principal component i, represents the surface potential of the principal component i, represents the critical value of the surface potential of the principal component i, represents the potential sensitivity factor; Reselect the suitability function: ;in, represents the density difference between the principal component i and the separation medium, Indicates the density distribution dispersion of the ore, represents the density difference gain coefficient; Electroselection suitability function: ;in, represents the conductivity of the main component i, represents the dielectric constant of principal component i, , represent the reference conductivity and reference dielectric constant respectively, and x represents the nonlinear coefficient; Finally, the adaptation score S k The method that is greater than the preset threshold is used as the concentrate sorting method to obtain the concentrate sorting instruction, which provides a pre-processing basis for the subsequent processing (such as smelting) of the concentrate; It should be noted that concentrate is the most economically valuable part of the ore, usually rich in target metals (such as Fe2O3, CuFeS2, etc.). After initial separation directly through jet parameters (such as blowing to the concentrate channel), further refined sorting is still required to maximize resource recovery; and, since the concentrate may contain multiple mineral components, it is difficult to completely separate them with a single sorting method, so it is necessary to select a combined sorting strategy based on the physical and chemical properties of the main components to reduce ineffective sorting steps, energy consumption and chemical agent use.

[0036] The second aspect of the present invention provides an ore sorting method based on deep learning, comprising: The raw ore is subjected to particle size classification and ore washing treatment through several layers of vibrating screens to obtain pre-treated classified ore, and image data and hyperspectral data of the pre-treated classified ore are collected; The first network and the second network constructed based on the deep learning algorithm are used to extract the surface features of the image data and the spectral features of the hyperspectral data respectively, and the surface features and the spectral features are fused and input into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; The jet parameters are calculated based on the conveyor belt speed, ore quality and air density, and the sorting instructions for the concentrate are determined. Ore sorting is performed based on the jet parameters and sorting instructions.

[0037] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0038] Working principle of the present invention: First, after the raw ore is graded and washed by a vibrating screen, it is spread on a belt conveyed at a uniform speed by an electromagnetic vibrating feeder, and the image and spectral data are collected synchronously by an industrial camera and a hyperspectral camera. Then, the first network and the second network extract features from the image and hyperspectral data respectively, and then the features are fused through a bidirectional cross-attention mechanism and a gated network. The fused optimized features are input into the classification network, and the ore type and main component probability distribution are output. Then, the jet parameters are calculated based on information such as conveyor belt speed, ore quality, and air density, and different types of ores are sorted using a jet device. The subsequent sorting instructions of the concentrate are further determined in combination with the main component attributes and economic weights of the concentrate.

[0039] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The ore sorting system based on deep learning is characterized by: include: Data acquisition module: used to classify and wash the raw ore through several layers of vibrating screens to obtain pre-treated classified ore, and to collect image data and hyperspectral data of the pre-treated classified ore; Ore classification module: used to extract the surface features of image data and the spectral features of hyperspectral data respectively by using the first network and the second network constructed based on the deep learning algorithm, and then fuse the surface features and the spectral features and input them into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; Sorting execution module: used to calculate the jet parameters according to the conveyor belt speed, ore quality and air density, and determine the sorting instructions for the concentrate, and execute ore sorting according to the jet parameters and sorting instructions.

2. The deep learning-based ore sorting system according to claim 1, characterized in that: The collected pre-processed image data and hyperspectral data of the graded ore include: The pre-treated graded ore is spread on the conveyor belt by using an electromagnetic vibrating feeder, and the conveyor belt is set to convey at a uniform speed at a preset speed; Industrial cameras and hyperspectral cameras are used to obtain image data and hyperspectral data of graded ores at preset shooting intervals.

3. The deep learning-based ore sorting system according to claim 1, characterized in that: The first network includes a multi-scale residual pyramid module, a dynamic convolutional attention module and a Transformer encoder, which are sequentially constructed; wherein, The input of the multi-scale residual pyramid module is image data, which includes several levels of dilated convolution branches composed of dilated convolution and dilated spatial pyramid pooling layers in parallel, and the outputs of several levels of dilated convolution branches are aggregated through bilinear interpolation upsampling channel splicing and 1×1 convolution layer to obtain the aggregated feature F agg ; The input of the dynamic convolution attention module is the aggregated feature, which includes a parameter generation network composed of a global average pooling layer, a 1×1 convolution layer, and a ReLU function, and a dynamic kernel convolution module composed of a depth-separable convolution and a residual connection, to obtain the attention enhancement feature, which is input into the Transformer encoder to obtain the surface feature. .

4. The deep learning-based ore sorting system according to claim 3, characterized in that: The formula of the dynamic convolution attention module is: The convolution kernel parameter K is obtained through the parameter generation network: K=f (1) 1×1 (ReLU(f (0) 1×1 (GAP(F agg ))))); Among them, GAP( ) represents the global average pooling layer, f (0) 1×1 ( ) represents the first 1×1 convolutional layer, f (1) 1×1 ( ) represents the second 1×1 convolutional layer, ReLU( ) represents the ReLU function; The dynamic kernel convolution module is used to perform deep separable convolution and residual linking on the convolution kernel parameters and aggregate features to obtain the attention enhancement feature: Y out =F agg +α×reshape(f g (reshape(F agg ),K)); where reshape( ) represents the dimension reshaping operation, f g ( ) represents a depth-wise separable convolution, and α represents a learnable parameter.

5. The deep learning-based ore sorting system according to claim 4, characterized in that: The second network includes a spectral band grouping module, a local-global spectral attention module and a three-dimensional convolutional feature compression layer; wherein, The input of the spectral band grouping module is hyperspectral data, including using the K-means algorithm to divide several spectral bands of the spectral data into a preset number of groups to obtain several band subsets, and then inputting the several band subsets into the three-dimensional convolution layer and the pooling layer in turn to perform spectral dimension splicing, and generating intermediate features F through the spectral normalization layer. group ; The local-global spectral attention module includes extracting the local band correlation of the intermediate features using three-dimensional convolution and GELU activation function to obtain the local feature F local , and input the local features into the multi-head self-attention mechanism: , and obtain the global feature F global , where Q, K, and V represent the query matrix, key matrix, and value matrix obtained by the linear projection layer of local features, T represents the transposition operation, and d represents the scaling factor. global The number of channels is determined; The three-dimensional convolution feature compression layer includes processing the global features using two-dimensional convolution and three-dimensional convolution, and then inputting them into the deformable three-dimensional convolution layer and the global maximum pooling layer in sequence to obtain the spectral features. .

6. The deep learning-based ore sorting system according to claim 5, characterized in that: The step of fusing the surface features and the spectral features and inputting them into the classification network comprises: S1, establishes semantic associations between surface features and spectral features through a bidirectional cross-attention mechanism to obtain joint features ; Among them, the bidirectional cross attention mechanism includes: Using surface features as query vectors , the spectral features serve as key vectors Sum value vector , the first fusion feature is calculated according to the attention calculation formula : ; Use spectral features as query vectors , surface features as key vectors Sum value vector , the second fusion feature is calculated according to the attention calculation formula : ; The first fusion feature and the second fusion feature are concatenated in the channel dimension to obtain the joint feature ; S2, the joint features, surface features, and spectral features are combined using a gating network to generate a gating weight G, and the gating weight is element-wise multiplied with the joint features to obtain the optimized fusion feature ; Among them, the mathematical expression of the gating network is: ,σ( ) represents the Sigmoid function, δ( ) represents the GELU activation function, represents the feature concatenation operation, W1 and W2 represent the preset learnable parameter matrices; S3, inputting the optimized fusion features into the classification network to obtain the main components and ore types of the ore; wherein the classification network includes: Use normalization layers, fully connected layers, and random dropout layers to extract shared features from optimized fusion features , the mathematical expression is: ,LayerNorm( ) represents the normalization layer, Linear( ) represents the fully connected layer, Dropout( ) represents a random dropout layer; The classification features in the shared features are extracted using the gradient reversal layer and the domain adversarial network, and then the classification features are input into the ore type classifier containing the fully connected layer to output the probability distribution P1 of the predicted ore type; the ore types include concentrate, middlings and tailings; The multi-layer perceptron MLP and Softmax function are used to calculate the attention weight of the optimized fusion feature, and the element-by-element product is performed with the optimized fusion feature to obtain the enhanced main feature. The enhanced main feature is input into the ore component classifier containing a fully connected layer, and the probability distribution P2 of the ore main component is output; among which, the ore main components include Fe2O3, Fe3O4, SiO2, Al2O3, CuFeS2, and FeS2.

7. The deep learning-based ore sorting system according to claim 1, characterized in that: The jet parameters calculated according to the conveyor belt speed, ore mass and air density include: Use Kalman filter to predict the movement trajectory of ore and get the time when ore reaches the jet position ; According to the formula Calculate the jet delay time ;in, Indicates the physical distance from the detection position to the injection valve, v b Indicates the conveyor belt speed, Indicates the solenoid valve response time, Indicates the time consumption of the ore classification module; Time for the ore to reach the jet position With jet delay time Subtract and get the jet trigger time ; Set up sorting channels according to ore type and set the sorting channel spacing according to Calculate the target lateral velocity , the calculation formula is: ; Where h represents the falling height and g represents the acceleration due to gravity; According to the formula The jet duration is calculated; where, Indicates the quality of the ore, represents the air density, represents the nozzle cross-sectional area, Indicates the nozzle flow coefficient.

8. The deep learning-based ore sorting system according to claim 6, characterized in that: The determination of the separation instructions for the concentrate comprises: Obtain the physical and chemical property value A of the main component i of the concentrate according to the ore characteristics database i ; chemical properties include magnetism, density, hydrophobicity, conductivity and dielectric constant; Set the economic weight E of the main component i of the concentrate i ; According to the formula CII i =P i ×[β×A i +(1-β)×E i ] Calculate the physical and chemical property influence index CII of the main component i of the ore i ; Among them, P i represents the probability value of the main component i of the ore, which is obtained by the output of the classification network, and β represents the preset influence coefficient; According to the formula Calculate the concentration separation method adaptation score S k ; where k represents the sorting method index, represents the applicability function of sorting method k to ore principal component i, and n represents the total number of ore principal components; The adaptation score S k The method that is greater than the preset threshold is used as the concentrate sorting method to obtain the concentrate sorting instruction.

9. The deep learning-based ore sorting system according to claim 8, characterized in that: The applicability function of the separation method k to the main component i of the ore includes: Magnetic separation suitability function: ; where k1 represents the slope factor, represents the magnetic susceptibility of principal component i, Indicates the preset magnetic susceptibility threshold; Flotation suitability function: ; where γ represents the hydrophobicity gain coefficient, represents the hydrophobic contact angle of principal component i, represents the surface potential of the principal component i, represents the critical value of the surface potential of the principal component i, represents the potential sensitivity factor; Reselect the suitability function: ;in, represents the density difference between the principal component i and the separation medium, Indicates the density distribution dispersion of the ore, represents the density difference gain coefficient; Electroselection suitability function: ;in, represents the conductivity of the main component i, represents the dielectric constant of principal component i, , represent the reference conductivity and reference dielectric constant respectively, and x represents the nonlinear coefficient.

10. A deep learning-based ore sorting method, applied to a deep learning-based ore sorting system according to any one of claims 1 to 9, characterized in that: include: The raw ore is subjected to particle size classification and ore washing treatment through several layers of vibrating screens to obtain pre-treated classified ore, and image data and hyperspectral data of the pre-treated classified ore are collected; The first network and the second network constructed based on the deep learning algorithm are used to extract the surface features of the image data and the spectral features of the hyperspectral data respectively, and the surface features and the spectral features are fused and input into the classification network to obtain the ore type and the main components of the ore; wherein the ore type includes concentrate, middlings and tailings; The jet parameters are calculated based on the conveyor belt speed, ore quality and air density, and the sorting instructions for the concentrate are determined. Ore sorting is performed based on the jet parameters and sorting instructions.

Images