Detection method based on lithium ore content

Abstract

The invention discloses a lithium ore content-based detection method, which relates to the field of mineral product detection, and comprises the following steps: 1, collecting geological data in different stratums of a mining area, and extracting lithium ore samples in the data to simulate ore particle distribution; according to the method, the high-probability region of the lithium content in the ore can be identified, the maximum information amount provided by each sampling point is ensured by maximizing the information gain, the condition of redundant sampling is reduced, the sampling cost is greatly reduced, the sampling time is greatly shortened, the data quality is effectively improved, and the necessity of manual intervention is reduced; a higher degree of automation can be realized in the exploration process; the lithium content in the ore sample can be better reflected by the model, the robustness and the adaptability of the model are enhanced, the calculation efficiency is improved, and the consumption of calculation resources is reduced, so that the time cost and the hardware overhead of the whole detection process are reduced, and a more comprehensive optimization solution can be provided for practical application.

Description

Technical Field

[0001] The present invention relates to the field of mineral detection, and in particular to a detection method based on lithium ore content. Background Art

[0002] The distribution of lithium ore is highly complex, and the lithium content in the ore often has spatial heterogeneity, which makes the traditional lithium ore exploration methods face challenges in accuracy and efficiency. Therefore, developing a detection method based on lithium ore content has become the key to improving the accuracy of lithium mining, reducing environmental impact and improving resource utilization; The existing detection methods based on lithium ore content cannot identify high-probability areas of lithium content in ores, there are many cases of redundant sampling, and the cost and time of required sampling are high, increasing the necessity of manual intervention; in addition, the existing detection methods based on lithium ore content cannot better reflect the lithium content in ore samples, reduce the robustness and adaptability of the model, and increase the consumption of computing resources. For this reason, we propose a detection method based on lithium ore content. Summary of the invention

[0003] The purpose of the present invention is to solve the defects in the prior art and provide a detection method based on lithium ore content.

[0004] The present invention proposes a detection method based on lithium ore content, and the technical solution adopted by the present invention to solve the technical problem is: Step 1: Collect geological data from different strata in the mining area and extract lithium ore samples from the data to simulate the distribution of ore particles; Step 2: Obtain the spectrum data of the sample, extract the characteristics of each group of spectrum data, and identify the characteristic peak of lithium element; Step 3: Distribute and store the spectrum data and sample information of multiple mining areas, and build a detection model to predict the lithium content of unknown samples based on the existing sample data and its corresponding lithium content; Step 4: Based on the prediction results, optimize the detection model parameters, perform real-time lithium content detection on the newly collected samples, and dynamically optimize the detection results.

[0005] As a further solution of the present invention, the specific steps of simulating the ore particle distribution in step 1 are as follows: S101: Sort the collected geological data from small to large, divide each group of sorted geological data into 4 parts according to the ratio of 25% using the quartile algorithm, and calculate the first quartile of the data and the third quartile ; S102: Based on the first quartile and the third quartile , calculate the interquartile range of various geological data , and then Set to the upper limit of the normal data range, Set it as the lower limit of the normal data range, identify all data outside the normal data range in the geological data, regard them as outliers and remove them, and after removing the outliers, re-examine the distribution of the data; S103: Standardize and process the screened geological data, extract lithium ore samples from the geological data, establish a three-dimensional simulation model of ore particle distribution based on each group of geological data and lithium ore samples, and use the constructed three-dimensional simulation model to generate particle distribution probability maps at different depths and locations in the mining area, and simultaneously obtain high probability areas of lithium ore particles, wherein the specific expression of the high probability areas of lithium ore particles is as follows:

[0006]

[0007] In the formula, Representative position The particle density function at ; Represents the diameter distribution of particles; Represents the density of the ore; The randomness parameter representing the particle distribution is adjusted by It can simulate the distribution of ore particles under different geological conditions; Representative The probability of particle distribution within a sampling area; Representative The volume of the sampling area; represents the particle density function; S104: The obtained high-probability areas of each lithium ore particle are used as candidate sampling points, and the potential information gain of each candidate sampling point for lithium content detection is calculated. The number of sampling points is optimized according to the information gain distribution of the sampling points and the sampling cost.

[0008] As a further solution of the present invention, the specific steps of optimizing the number of sampling points in S104 are as follows: S201: Establishing a sampling point location set based on the information gain distribution of the sampling points , with the goal of maximizing the total information gain, and setting the objective function, then determining the search range of the sampling points according to the two-dimensional and three-dimensional areas of the mining area, and randomly generating multiple groups of particle positions within the specified search range, where each particle represents a sampling point layout scheme; S202: Initialize the initial velocity of each particle and set the particle velocity range. Then calculate the information gain corresponding to the position of each particle according to the objective function, and record the calculation result as the fitness value of the corresponding particle. At the beginning of each iteration, each particle compares the current fitness value with its corresponding historical fitness value, and takes the fitness peak value as the historical individual merit. Then compare the current fitness values ​​of each particle, and take the position of the fitness peak particle as the global optimal position. S203: updating the speed of each particle according to the individual merit of each particle and the global optimal position of the group, and then updating the position of the particle according to the updated speed. After the update is completed, detecting whether the updated position and speed of each group of particles are within the preset search range and speed range, if not, correcting the position and speed of the particle to the boundary value of that range; S204: Repeat the selection of historical individual merits, the selection of the global optimal point, the updating of particle speeds, and the updating of particle positions until the fitness value change of the global optimal point converges to the set threshold, stop the iteration, and output the particle position of the global optimal point as the sampling point layout plan for information gain peaking.

[0009] As a further solution of the present invention, the specific steps of extracting the features of each group of spectrum data and identifying the characteristic peak value of lithium element in step 2 are as follows: S301: Acquire spectrum data of lithium ore samples through XRF and LIBS instruments, perform noise removal, baseline correction and standardization preprocessing on the acquired spectrum data, and then divide the preprocessed spectrum data into windows of fixed length to construct corresponding input data; S302: construct a CNN network, input the input data into the CNN network, and forward propagate the input data through the CNN network, use the CNN network convolution layer to perform a convolution operation on the input data to extract feature data from the spectrogram data, and then perform feature dimensionality reduction on the extracted feature data through a pooling layer; S303: Repeat convolution and pooling operations through the CNN network until the tail pooling layer is reached and the final feature data after dimensionality reduction is output. Each wavelength point in the extracted feature data is classified through the fully connected layer, and the specific position of the lithium element characteristic peak is output. Then, the specific wavelength position of the lithium characteristic peak is located by maximizing the prediction probability.

[0010] As a further solution of the present invention, the specific steps of predicting the lithium content of the unknown sample in step 3 are as follows: S401: Collect spectrogram data from different mining areas, remove noise signals from each group of spectrogram data, scale the data to the range of [0, 1] through normalization, divide the preprocessed sample data into 75% training set, 15% validation set and 15% test set, and build a CNN-LSTM hybrid model as a detection model; S402: Input the training set into the detection model in batches. The CNN module in the model receives the training set data, and performs convolution operation on the spectrogram data through the convolution layer in the CNN module, and extracts local features in each group of spectrogram data. Then, the pooling layer is used to average pool the extracted local features to reduce the local feature dimension. After multiple convolutional layers and pooling layers, the processed local features are modeled through the ReLU activation function in the fully connected layer. S403: The CNN module uses the modeling result as input data and transmits it to the LSTM module. The LSTM module calculates the output of the input gate, the forget gate, and the output gate according to the current input data and the hidden state at the previous moment, and updates the memory unit at the current moment. Then, according to the value of the output gate, the hidden state at the current moment is calculated and output as the final prediction result of the lithium ore content. S404: Calculate the loss value between the prediction result and the actual lithium ore content data through the mean square error loss function, then input the calculated loss value from the output layer of the detection model, and transfer it layer by layer based on the chain rule, and update the parameters of each layer by gradient descent. After each round of training, evaluate the performance of the model through the validation set, and adjust the detection model learning rate, batch size, number of CNN convolutional layers, and number of LSTM layers based on the evaluation results. S405: Perform multiple rounds of training and verification on the detection model until the change in the loss value of the model on the verification set converges to the set threshold. After the training and verification process is completed, the trained model is finally evaluated using the test set, and the difference between the predicted value and the actual value is calculated using the mean square error and mean absolute error functions to evaluate the accuracy of the model. If the model meets the preset performance, it is deployed to the actual detection platform, otherwise the model is retrained and verified; S406: The actually collected mining area sample data is input into the detection model. The input mining area sample data is forward propagated and processed by the CNN and LSTM modules in the detection model to output the lithium ore content prediction results of each mining area.

[0011] As a further solution of the present invention, the specific steps of optimizing the detection model parameters in step 4 are as follows: S501: Initialize a population containing multiple sets of detection model parameter combinations, where each individual represents a parameter combination, calculate the prediction accuracy of the current lithium ore content of each individual, traverse the prediction accuracy of each individual in the initial population, and select individuals with high prediction accuracy as qualified solutions for the current population; S502: Set a set of coefficient vectors, and generate a random number between 0 and 1 in each iteration. If the generated random number is greater than the preset probability parameter, the coefficient for controlling the position update is calculated based on the coefficient vector. and step length , according to the size of the direction, determine whether the remaining individuals are contracted around the qualified solution; S503: If <1, it means that the remaining individuals are surrounded by the qualified solution, and the positions of the remaining individuals are updated according to the qualified solution, that is, the detection model parameter values ​​of each individual are adjusted. ≥1, it means that the remaining individuals are far away from the qualified solution. Through random search, random positions are selected in the population space, and the positions of the remaining individuals are updated. After each iteration, the coefficient vector is updated based on the linear decreasing rule. S504: If the generated random number is less than or equal to the preset probability parameter, the distance between the positions of the remaining individuals in the population and the position of the solution that meets the condition is calculated, and the motion law of each individual along the spiral trajectory approaching the solution that meets the condition is simulated through the spiral motion formula to update the individual position; S505: Repeatedly select solutions that meet the conditions and iteratively update the positions until the fitness value converges to a preset threshold range, then compare the prediction accuracy of individuals in each group, and output a detection model parameter combination with high prediction accuracy, and adjust the current detection model parameters based on the detection model parameter combination.

[0012] As a further solution of the present invention, the specific steps of dynamically optimizing the detection results in step 4 are as follows: S601: Obtain the data of the detection platform through historical data and expert experience, and calculate the prior distribution of the current detection platform based on the collected data. At the same time, collect the lithium ore detection sample data of each mining area in real time, and express the new lithium ore content detection results as ,in Indicates The lithium content of the samples was measured. ; S602: Update the likelihood function corresponding to the detection platform according to the newly acquired lithium ore content detection result, and then calculate the posterior distribution of the mining area detection platform according to the corresponding prior distribution and the likelihood function to obtain the probability distribution of the model parameters under the new data, visualize the posterior distribution results of the detection model, and select the peak values ​​in the posterior distribution to select the required control scheme; S603: Repeat the update of the posterior distribution and the adjustment of the scheme until the subsequent control scheme prediction accuracy change value converges to the set range, and adjust each detection platform according to the obtained optimal control scheme. After that, each time a new sample arrives, the detection platform updates the posterior distribution according to the new data and optimizes the control scheme. After that, the detection platform updates the detection results according to the optimized control scheme, and visualizes the detection results based on the simulation model.

[0013] Beneficial effects of the present invention: 1. Compared with the traditional sampling method based on experience or rules, which may lead to too many or too few sampling points and waste time and resources, the present invention simulates the distribution of ore particles to conduct a more detailed analysis of the mining area, predict the changes in lithium content in different areas, and can identify high-probability areas of lithium content in the ore. By maximizing information gain, it ensures that the amount of information provided by each sampling point is maximized, reduces redundant sampling, greatly reduces the cost and time of sampling, effectively improves the quality of data, reduces the necessity of manual intervention, and enables the exploration process to achieve a higher degree of automation.

[0014] 2. When faced with different types of mining area data, the present invention conducts a global search in a high-dimensional parameter space, searches for solutions that meet the conditions in a wide range of parameter spaces, and dynamically adjusts the hyperparameters of the model. This ensures that the model can better reflect the lithium content in ore samples, enhances the robustness and adaptability of the model, improves computational efficiency, and reduces the consumption of computing resources, thereby reducing the time cost and hardware overhead of the entire detection process, and can provide a more comprehensive optimization solution for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings.

[0016] Figure 1 This is a framework diagram of a detection method based on lithium ore content. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0018] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0019] Example 1 The present invention provides a method for detecting lithium ore content. Figure 1 , Figure 1 A framework diagram of a lithium ore content detection method provided by an embodiment of the present invention. The system includes the following steps Collect geological data from different strata in the mining area and extract lithium ore samples from the data to simulate the distribution of ore particles.

[0020] Specifically, the collected geological data are sorted from small to large, and the sorted geological data are divided into 4 parts according to the ratio of 25% using the quartile algorithm, and the first quartile of the data is calculated. and the third quartile , based on the first quartile and the third quartile , calculate the interquartile range of various geological data , and then Set to the upper limit of the normal data range, It is set as the lower limit of the normal data range, and all data outside the normal data range in the geological data are identified, regarded as outliers and eliminated. After eliminating the outliers, the data distribution is re-checked, the screened geological data are standardized, and the lithium ore samples in the geological data are extracted. According to each group of geological data and lithium ore samples, a three-dimensional simulation model of ore particle distribution is established, and the constructed three-dimensional simulation model is used to generate particle distribution probability maps at different depths and locations in the mining area. At the same time, the high probability areas of lithium ore particles are obtained, and the obtained high probability areas of lithium ore particles are used as candidate sampling points. The potential information gain of each candidate sampling point for lithium content detection is calculated, and the number of sampling points is optimized according to the information gain distribution of the sampling points and the sampling cost.

[0021] It should be further explained that the sampling point location set is established according to the information gain distribution of the sampling points , with the goal of maximizing the total information gain, and setting the objective function, then determining the search range of the sampling point according to the two-dimensional and three-dimensional areas of the mining area, and randomly generating multiple groups of particle positions within the specified search range, where each particle represents a sampling point layout scheme, initializing the initial velocity of each particle, and setting the particle velocity range, and then calculating the information gain corresponding to the position of each particle according to the objective function, and recording the calculation result as the fitness value of the corresponding particle. At the beginning of each round of iteration, each particle compares the current fitness value with its corresponding historical fitness value, and takes the fitness peak as the historical individual merit, then compares the current fitness value of each particle, and takes the fitness peak particle as the historical individual merit. The position is taken as the global optimal position, and the speed of each particle is updated according to the individual merit of each particle and the global optimal position of the group. Then the position of the particle is updated according to the updated speed. After the update is completed, the position and speed of each group of particles are detected to see if they are within the preset search range and speed range. If not, the position and speed of the particle are corrected to the boundary value of that range. The historical individual merit selection, global optimal position selection, particle speed update and particle position update are repeated until the fitness value change of the global optimal position converges to the set threshold, the iteration is stopped, and the particle position of the global optimal position is output as the sampling point layout plan for information gain peaking.

[0022] In this embodiment, the specific expression of the high probability area of ​​lithium ore particles is as follows:

[0023]

[0024] In the formula, Representative position The particle density function at ; Represents the diameter distribution of particles; Represents the density of the ore; The randomness parameter representing the particle distribution is adjusted by It can simulate the distribution of ore particles under different geological conditions; Representative The probability of particle distribution within a sampling area; Representative The volume of the sampling area; Represents the particle density function.

[0025] The spectrum data of the sample is obtained, the features of each group of spectrum data are extracted, and the characteristic peaks of the lithium element are identified.

[0026] Specifically, the spectrum data of the lithium ore sample is obtained by XRF and LIBS instruments, and the collected spectrum data is preprocessed by noise removal, baseline correction and standardization. The preprocessed spectrum data is then divided into windows of fixed length to construct corresponding input data, and a CNN network is constructed. The input data is input into the CNN network, and the input data is forward propagated through the CNN network. The input data is convolved by the convolution layer of the CNN network to extract feature data from the spectrum data, and then the extracted feature data is subjected to feature dimensionality reduction through the pooling layer. The convolution and pooling operations are repeated through the CNN network until the final dimensionality reduction feature data is output after reaching the tail pooling layer. Each wavelength point in the extracted feature data is classified through the fully connected layer, and the specific position of the lithium element characteristic peak is output. The specific wavelength position of the lithium characteristic peak is located by maximizing the prediction probability.

[0027] Example 2 The present invention provides a method for detecting lithium ore content. Figure 1 , Figure 1 A framework diagram of a detection method based on lithium ore content provided by an embodiment of the present invention. The system includes: Distributed storage of spectral data and sample information of multiple mining areas, and construction of a detection model to predict the lithium content of unknown samples based on the existing sample data and its corresponding lithium content.

[0028] Specifically, spectrum data from different mining areas are collected, and noise signals in each group of spectrum data are removed. The data are scaled to the range of [0, 1] through normalization processing. The preprocessed sample data is divided into 75% training set, 15% validation set and 15% test set. A CNN-LSTM hybrid model is constructed as a detection model, and the training set is input into the detection model in batches. The CNN module in the model receives the training set data, and performs convolution operation on the spectrum data through the convolution layer in the CNN module, and extracts local features in each group of spectrum data. The pooling layer is then used to average pool the extracted local features to reduce the local feature dimension. After multiple convolutional layers and pooling layers, the processed local features are modeled through the ReLU activation function in the fully connected layer. The CNN module uses the modeling result as input data and transmits it to the LSTM module. The LSTM module calculates the output of the input gate, the forget gate and the output gate according to the current input data and the hidden state at the previous moment, and updates the memory unit at the current moment. Then, according to the value of the output gate, the hidden state at the current moment is calculated and used as the lithium ore. The final prediction result of the content is output, and the loss value between the prediction result and the actual lithium ore content data is calculated through the mean square error loss function. The calculated loss value is then input from the output layer of the detection model and transmitted layer by layer based on the chain rule. At the same time, the parameters of each layer are updated by gradient descent. After each round of training, the performance of the model is evaluated through the validation set, and the learning rate, batch size, number of CNN convolutional layers, and number of LSTM layers of the detection model are adjusted based on the evaluation results. The detection model is trained and verified for multiple rounds until the loss value change of the model on the validation set converges to the set threshold. After the training and verification process is completed, the trained model is finally evaluated using the test set, and the mean square error and mean absolute error functions are used to calculate the difference between the predicted value and the actual value to evaluate the accuracy of the model. If the model meets the preset performance, it will be deployed to the actual detection platform, otherwise the model will be retrained and verified, and the actual collected mining area sample data will be input into the detection model. The input mining area sample data is forward propagated and processed by the CNN and LSTM modules in the detection model to output the lithium ore content prediction results of each mining area.

[0029] Based on the prediction results, the detection model parameters are optimized, and then the newly collected samples are tested for real-time lithium content, and the test results are dynamically optimized.

[0030] Specifically, a population containing multiple sets of detection model parameter combinations is initialized, in which each individual represents a parameter combination, and the prediction accuracy of the current lithium ore content of each individual is calculated. The prediction accuracy of each individual in the initial population is traversed, and the individuals with high prediction accuracy are selected as the qualified solution of the current population. A set of coefficient vectors is set, and at the same time, a random number between 0 and 1 is generated in each round of iteration. If the generated random number is greater than the preset probability parameter, the coefficient of controlling the position update is calculated according to the coefficient vector. and step length , according to the size of the direction, determine whether the remaining individuals are shrinking around the qualified solution. If <1, it means that the remaining individuals are surrounded by the qualified solution, and the positions of the remaining individuals are updated according to the qualified solution, that is, the detection model parameter values ​​of each individual are adjusted. ≥1, it means that the remaining individuals are far away from the qualified solution. Through random search, random positions are selected in the population space, and the positions of the remaining individuals are updated. After each iteration, the coefficient vector is updated based on the linear decreasing rule. If the generated random number is less than or equal to the preset probability parameter, the distance between the positions of the remaining individuals in the population and the qualified solution position is calculated, and the spiral motion formula is used to simulate the movement law of each individual approaching the qualified solution along the spiral trajectory, update the individual position, repeatedly select the qualified solution and iteratively update the position until the fitness value converges to the preset threshold range, then compare the prediction accuracy of each group of individuals, output the detection model parameter combination with high prediction accuracy, and adjust the current detection model parameters based on the detection model parameter combination.

[0031] Specifically, the data of the detection platform are obtained through historical data and expert experience, and the prior distribution of the current detection platform is calculated based on the collected data. At the same time, the lithium ore detection sample data of each mining area is collected in real time, and the new lithium ore content detection results are expressed as ,in Indicates The lithium content of the samples was measured. , update the likelihood function corresponding to the detection platform according to the newly obtained lithium ore content detection results, and then calculate the posterior distribution of the mining area detection platform according to the corresponding prior distribution and likelihood function to obtain the probability distribution of the model parameters under the new data, visualize the posterior distribution results of the detection model, and select the peaks in the posterior distribution to select the required control scheme, repeat the posterior distribution update and scheme adjustment until the subsequent control scheme prediction accuracy change value converges to the set range, and adjust each detection platform according to the obtained optimal control scheme, and then each new sample arrives, the detection platform updates the posterior distribution according to the new data and optimizes the control scheme, and then the detection platform updates the detection results according to the optimized control scheme, and visualizes the detection results based on the simulation model.

[0032] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A detection method based on lithium ore content, characterized in that: The following steps are involved: Step 1: Collect geological data from different strata in the mining area and extract lithium ore samples from the data to simulate the distribution of ore particles; Step 2: Obtain the spectrum data of the sample, extract the characteristics of each group of spectrum data, and identify the characteristic peak of lithium element; Step 3: Distribute and store the spectrum data and sample information of multiple mining areas, and build a detection model to predict the lithium content of unknown samples based on the existing sample data and its corresponding lithium content; Step 4: Based on the prediction results, optimize the detection model parameters, perform real-time lithium content detection on the newly collected samples, and dynamically optimize the detection results.

2. A detection method based on lithium ore content according to claim 1, characterized in that: The specific steps of simulating the ore particle distribution in step 1 are as follows: S101: Sort the collected geological data from small to large, divide each group of sorted geological data into 4 parts according to the ratio of 25% using the quartile algorithm, and calculate the first quartile of the data and the third quartile ; S102: Based on the first quartile and the third quartile , calculate the interquartile range of various geological data , and then Set to the upper limit of the normal data range, Set it as the lower limit of the normal data range, identify all data outside the normal data range in the geological data, regard them as outliers and remove them, and after removing the outliers, re-examine the distribution of the data; S103: Standardize and process the screened geological data, extract lithium ore samples from the geological data, establish a three-dimensional simulation model of ore particle distribution based on each group of geological data and lithium ore samples, and use the constructed three-dimensional simulation model to generate particle distribution probability maps at different depths and locations in the mining area, and simultaneously obtain high probability areas of lithium ore particles, wherein the specific expression of the high probability areas of lithium ore particles is as follows: ； ； In the formula, Representative position The particle density function at represents the diameter distribution of the particles, Represents the density of the ore, The randomness parameter representing the particle distribution is adjusted by It can simulate the distribution of ore particles under different geological conditions. Representative The probability of particle distribution in a sampling area is Representative The volume of the sampling area, represents the particle density function; S104: The obtained high-probability areas of each lithium ore particle are used as candidate sampling points, and the potential information gain of each candidate sampling point for lithium content detection is calculated. The number of sampling points is optimized according to the information gain distribution of the sampling points and the sampling cost.

3. A detection method based on lithium ore content according to claim 2, characterized in that: The specific steps of optimizing the number of sampling points in S104 are as follows: S201: Establishing a sampling point location set based on the information gain distribution of the sampling points , with the goal of maximizing the total information gain, and setting the objective function, then determining the search range of the sampling points according to the two-dimensional and three-dimensional areas of the mining area, and randomly generating multiple groups of particle positions within the specified search range, where each particle represents a sampling point layout scheme; S202: Initialize the initial velocity of each particle and set the particle velocity range. Then calculate the information gain corresponding to the position of each particle according to the objective function, and record the calculation result as the fitness value of the corresponding particle. At the beginning of each iteration, each particle compares the current fitness value with its corresponding historical fitness value, and takes the fitness peak value as the historical individual merit. Then compare the current fitness values ​​of each particle, and take the position of the fitness peak particle as the global optimal position. S203: updating the speed of each particle according to the individual merit of each particle and the global optimal position of the group, and then updating the position of the particle according to the updated speed. After the update is completed, detecting whether the updated position and speed of each group of particles are within the preset search range and speed range, if not, correcting the position and speed of the particle to the boundary value of that range; S204: Repeat the selection of historical individual merits, the selection of the global optimal point, the updating of particle speeds, and the updating of particle positions until the fitness value change of the global optimal point converges to the set threshold, stop the iteration, and output the particle position of the global optimal point as the sampling point layout plan for information gain peaking.

4. A detection method based on lithium ore content according to claim 2, characterized in that: The specific steps of extracting the features of each group of spectrum data and identifying the characteristic peaks of lithium elements in step 2 are as follows: S301: Acquire spectrum data of lithium ore samples through XRF and LIBS instruments, perform noise removal, baseline correction and standardization preprocessing on the acquired spectrum data, and then divide the preprocessed spectrum data into windows of fixed length to construct corresponding input data; S302: construct a CNN network, input the input data into the CNN network, and forward propagate the input data through the CNN network, use the CNN network convolution layer to perform a convolution operation on the input data to extract feature data from the spectrogram data, and then perform feature dimensionality reduction on the extracted feature data through a pooling layer; S303: Repeat convolution and pooling operations through the CNN network until the tail pooling layer is reached and the final feature data after dimensionality reduction is output. Each wavelength point in the extracted feature data is classified through the fully connected layer, and the specific position of the lithium element characteristic peak is output. Then, the specific wavelength position of the lithium characteristic peak is located by maximizing the prediction probability.

5. A detection method based on lithium ore content according to claim 3, characterized in that: The specific steps of predicting the lithium content of the unknown sample in step 3 are as follows: S401: Collect spectrogram data from different mining areas, remove noise signals from each group of spectrogram data, scale the data to the range of [0, 1] through normalization, divide the preprocessed sample data into 75% training set, 15% validation set and 15% test set, and build a CNN-LSTM hybrid model as a detection model; S402: Input the training set into the detection model in batches. The CNN module in the model receives the training set data, and performs convolution operation on the spectrogram data through the convolution layer in the CNN module, and extracts local features in each group of spectrogram data. Then, the pooling layer is used to average pool the extracted local features to reduce the local feature dimension. After multiple convolutional layers and pooling layers, the processed local features are modeled through the ReLU activation function in the fully connected layer. S403: The CNN module uses the modeling result as input data and transmits it to the LSTM module. The LSTM module calculates the output of the input gate, the forget gate, and the output gate according to the current input data and the hidden state at the previous moment, and updates the memory unit at the current moment. Then, according to the value of the output gate, the hidden state at the current moment is calculated and output as the final prediction result of the lithium ore content. S404: Calculate the loss value between the prediction result and the actual lithium ore content data through the mean square error loss function, then input the calculated loss value from the output layer of the detection model, and transfer it layer by layer based on the chain rule, and update the parameters of each layer by gradient descent. After each round of training, evaluate the performance of the model through the validation set, and adjust the detection model learning rate, batch size, number of CNN convolutional layers, and number of LSTM layers based on the evaluation results. S405: Perform multiple rounds of training and verification on the detection model until the change in the loss value of the model on the verification set converges to the set threshold. After the training and verification process is completed, the trained model is finally evaluated using the test set, and the difference between the predicted value and the actual value is calculated using the mean square error and mean absolute error functions to evaluate the accuracy of the model. If the model meets the preset performance, it is deployed to the actual detection platform, otherwise the model is retrained and verified; S406: The actually collected mining area sample data is input into the detection model. The input mining area sample data is forward propagated and processed by the CNN and LSTM modules in the detection model to output the lithium ore content prediction results of each mining area.

6. A detection method based on lithium ore content according to claim 5, characterized in that: The specific steps for optimizing the detection model parameters described in step 4 are as follows: S501: Initialize a population containing multiple sets of detection model parameter combinations, where each individual represents a parameter combination, calculate the prediction accuracy of the current lithium ore content of each individual, traverse the prediction accuracy of each individual in the initial population, and select individuals with high prediction accuracy as qualified solutions for the current population; S502: Set a set of coefficient vectors, and generate a random number between 0 and 1 in each iteration. If the generated random number is greater than the preset probability parameter, the coefficient for controlling the position update is calculated based on the coefficient vector. and step length , according to the size of the direction, determine whether the remaining individuals are contracted around the qualified solution; S503: If <1, it means that the remaining individuals are surrounded by the qualified solution, and the positions of the remaining individuals are updated according to the qualified solution, that is, the detection model parameter values ​​of each individual are adjusted. ≥1, it means that the remaining individuals are far away from the qualified solution. Through random search, random positions are selected in the population space, and the positions of the remaining individuals are updated. After each iteration, the coefficient vector is updated based on the linear decreasing rule. S504: If the generated random number is less than or equal to the preset probability parameter, the distance between the positions of the remaining individuals in the population and the position of the solution that meets the condition is calculated, and the motion law of each individual along the spiral trajectory approaching the solution that meets the condition is simulated through the spiral motion formula to update the individual position; S505: Repeatedly select solutions that meet the conditions and iteratively update the positions until the fitness value converges to a preset threshold range, then compare the prediction accuracy of individuals in each group, and output a detection model parameter combination with high prediction accuracy, and adjust the current detection model parameters based on the detection model parameter combination.

7. A detection method based on lithium ore content according to claim 1, characterized in that: The specific steps of dynamically optimizing the detection results described in step 4 are as follows: S601: Obtain the data of the detection platform through historical data and expert experience, and calculate the prior distribution of the current detection platform based on the collected data. At the same time, collect the lithium ore detection sample data of each mining area in real time, and express the new lithium ore content detection results as ,in Indicates The lithium content of the samples was measured. ; S602: Update the likelihood function corresponding to the detection platform according to the newly acquired lithium ore content detection result, and then calculate the posterior distribution of the mining area detection platform according to the corresponding prior distribution and the likelihood function to obtain the probability distribution of the model parameters under the new data, visualize the posterior distribution results of the detection model, and select the peak values ​​in the posterior distribution to select the required control scheme; S603: Repeat the update of the posterior distribution and the adjustment of the scheme until the subsequent control scheme prediction accuracy change value converges to the set range, and adjust each detection platform according to the obtained optimal control scheme. After that, each time a new sample arrives, the detection platform updates the posterior distribution according to the new data and optimizes the control scheme. After that, the detection platform updates the detection results according to the optimized control scheme, and visualizes the detection results based on the simulation model.

Images