Element detection method and device based on model fusion migration, medium and equipment

Abstract

The invention provides an element detection method and device based on model fusion migration, a medium and equipment, and the method comprises the steps: carrying out the operation of a to-be-detected mineral spectrum data set preprocessed on an application site through various regression models in a laboratory, and outputting the current prediction contents of a plurality of elements corresponding to the spectrum data set through the various regression models; combining the current predicted contents through the optimized fusion weight matrix to obtain the fusion predicted contents of the plurality of elements in the to-be-detected mineral; therefore, due to the fact that different types of regression models have preference differences on different elements, the regression models with different characteristics are used for calculating the spectrum, the predicted contents of all the models are combined, and optimization adaptation of different types of regression models in a laboratory to the to-be-detected minerals on the application site is achieved; the defect that a single model cannot be optimized and generalized is effectively overcome, the detection accuracy of the content of multiple elements of the complex minerals is improved, and the in-situ and online detection analysis requirements of the complex minerals are met.

Description

Technical Field

[0001] The present application relates to the technical field of analyzing the element content of minerals using laser-induced breakdown spectroscopy data, and in particular to an element detection method, device, medium and equipment based on model fusion migration. Background Art

[0002] In-situ, online detection and analysis of complex substances (including those with complex chemical compositions and physical forms, such as mineral gravel) is of great significance in numerous fields, including mineral mining and processing, customs inspection of imported minerals, and industrial process optimization and control. Generally speaking, multi-element and multi-index detection and analysis of minerals are required to provide in-situ, online, real-time, and accurate analytical data.

[0003] In related technologies, a single machine learning algorithm is generally used to analyze Laser-Induced Breakdown Spectroscopy (LIBS) data, thereby detecting the content of each element in the mineral in real time online. However, in practical applications, for minerals from different sources, the spectral data of the minerals may be significantly affected by chemical matrix effects; the spectral data may also be affected by significant physical matrix effects, such as the particle size distribution and surface morphology changes of mineral gravel; the spectral data may also be affected by a combination of matrix effects, which cannot be accurately described and corrected by physical models. The above factors will affect the final detection accuracy of the mineral element content. Specifically, a single model trained in the laboratory cannot optimize the generalization performance for practical applications due to problems with the selection of the model algorithm or the model's adaptation preference for the spectral characteristics of each element.

[0004] Therefore, a new method for detecting mineral element content is currently needed to improve the detection accuracy of mineral element content and meet the requirements of in-situ, online, and multi-element detection and analysis of complex minerals. Summary of the Invention

[0005] In response to the problems existing in the prior art, the embodiments of the present application provide an element detection method, device, medium and equipment based on model fusion migration to solve or partially solve the technical problem in the prior art that the detection accuracy cannot be ensured when analyzing mineral spectral data to detect the content of complex mineral elements.

[0006] In a first aspect of the present application, a method for element detection based on model fusion migration is provided, the method comprising: Collecting laser-induced breakdown spectra of training mineral samples in a laboratory setting to obtain a training spectral dataset; collecting laser-induced breakdown spectra of internal test mineral samples in the laboratory setting to obtain an internal test spectral dataset; and using the training spectral dataset and the internal test spectral dataset to train and test multiple machine learning models, respectively, to obtain corresponding multiple regression models; Collecting laser-induced breakdown spectra of external test mineral samples for each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and optimizing a pre-constructed fusion weight matrix using the external test spectrum data set for each actual application scenario and the multiple regression models to obtain an optimized fusion weight matrix for each actual application scenario; Collecting a spectrum of a mineral to be tested in a target practical application scenario to obtain a spectrum dataset of the mineral to be tested, and preprocessing the spectrum dataset of the mineral to be tested to obtain a preprocessed spectrum dataset of the mineral to be tested; Inputting the pre-processed mineral spectral dataset to be tested into the multiple regression models respectively, and obtaining the current predicted content of several elements corresponding to the pre-processed mineral spectral dataset for each regression model; By combining the current predicted contents of several elements corresponding to the pre-processed mineral spectral data set to be tested using the optimized fusion weight matrix of the target actual application scenario, the fusion predicted contents of several elements in the mineral to be tested are obtained.

[0007] In the above scheme, the training spectral dataset and the internal test spectral dataset are used to train and test multiple machine learning models respectively to obtain corresponding multiple regression models, including: Obtaining the actual contents of several elements of the training mineral samples as the label element contents of the training mineral sample set; obtaining the actual contents of the internal test mineral samples as the label element contents of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as an input variable, various machine learning models are trained under the supervision of the elemental content of the training mineral samples, and cross-validation and model iteration are performed during the training process until a training termination condition is reached, thereby obtaining multiple regression models; For each regression model, the performance of the regression model is tested respectively according to the label element content of the internal test mineral sample using the preprocessed internal test spectral data set. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

[0008] In the above solution, the optimization of the pre-constructed fusion weight matrix using the external test spectral dataset and the multiple regression models includes: Preprocessing the external test spectral dataset to obtain a preprocessed external test spectral dataset; obtaining actual contents of several elements of the external test mineral sample as label element contents of the external test mineral sample; Based on a preset optimization algorithm, the predicted contents of several elements in the external test mineral samples by the multiple regression models are used as input variables, and the label element contents of the external test mineral samples are used as targets. The fusion weight matrix is ​​iteratively optimized until the preset requirements are met, and the optimized fusion weight matrix is ​​output; wherein, The predicted contents of several elements in the external test mineral sample by the multiple regression models are obtained by respectively calculating the pre-processed external test spectral data set by the multiple regression models.

[0009] In the above solution, before optimizing the pre-constructed fusion weight matrix using the external test spectral dataset and the multiple regression models, the method further includes: The dimension of the fusion weight matrix is ​​determined according to the number of the elements and the number of the regression models; the dimension of the fusion weight matrix is The fusion weight matrix includes the weight and prediction deviation of each regression model when predicting each element when calculating the fusion prediction content of each element; M is the number of the elements, N is the number of regression models; In the fusion weight matrix, for each element, the weight contributed by each regression model when predicting each element and the prediction deviation are respectively assigned preset initial values.

[0010] In the above scheme, the preset optimization algorithm is based on multiple regression models, and the predicted content of several elements in the external test mineral sample is used as input variables. The content of the label elements in the external test mineral sample is used as the target, and the fusion weight matrix is ​​iteratively optimized, including: Based on the fusion weight matrix corresponding to each iteration, the predicted contents of the plurality of elements are combined and added with the predicted deviations of the corresponding elements in the fusion weight matrix to obtain the fusion predicted contents of the plurality of elements in the external test mineral sample; With the goal of reducing the residual value between the fused predicted content and the label element content of several elements in the external test mineral sample, the matrix elements of the fusion weight matrix are used as parameters to be optimized, and the fusion weight matrix is ​​iteratively optimized using the optimization algorithm until the optimization termination condition is reached; The residual value is calculated by a pre-constructed residual function, and is used to characterize the error between the fusion predicted content of several elements in the external test mineral sample and the content of each label element.

[0011] In the above scheme, the current predicted contents of several elements corresponding to the pre-processed mineral spectral dataset to be tested are combined by the optimized fusion weight matrix of the target actual application scenario to obtain the fused predicted contents of several elements in the mineral to be tested, including: For each element, performing a weighted combination of the weight contributed by each regression model in the optimized fusion weight matrix when predicting each element and the current predicted content of each element by each regression model to obtain a weighted combined predicted content of each element; The weighted combined prediction content of each element is added to the prediction deviation of the corresponding element in the optimized fusion weight matrix to obtain the fused prediction content of each element.

[0012] In a second aspect of the present application, a device for detecting elements based on model fusion migration is provided, the device comprising: A training unit is configured to collect laser-induced breakdown spectra of training mineral samples in a laboratory setting to obtain a training spectral dataset; collect laser-induced breakdown spectra of internal test mineral samples in the laboratory setting to obtain an internal test spectral dataset; and use the training spectral dataset and the internal test spectral dataset to train and test a plurality of machine learning models, respectively, to obtain a plurality of corresponding regression models; an optimization unit, configured to collect laser-induced breakdown spectra of external test mineral samples of each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and optimize a pre-constructed fusion weight matrix using the external test spectrum data set of each actual application scenario and the multiple regression models to obtain an optimized fusion weight matrix for each actual application scenario; An acquisition unit is used to acquire a spectrum of a mineral to be tested in a target actual application scenario to obtain a spectrum data set of the mineral to be tested, and preprocess the spectrum data set of the mineral to be tested to obtain a preprocessed spectrum data set of the mineral to be tested; A prediction unit, configured to input the pre-processed mineral spectrum dataset to be tested into the multiple regression models respectively, and obtain a current predicted content of several elements corresponding to the pre-processed mineral spectrum dataset for each regression model; A fusion unit is used to combine the current predicted contents of several elements corresponding to the preprocessed mineral spectral data set by the multiple regression models through the optimized fusion weight matrix of the target actual application scenario to obtain a fused predicted content of several elements in the mineral to be tested.

[0013] In the above solution, the training unit is specifically used to: Obtaining the actual contents of several elements of the training mineral samples as the label element contents of the training mineral sample set; obtaining the actual contents of the internal test mineral samples as the label element contents of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as an input variable, various machine learning models are trained under the supervision of the elemental content of the training mineral samples, and cross-validation and model iteration are performed during the training process until a training termination condition is reached, thereby obtaining multiple regression models; For each regression model, the performance of the regression model is tested respectively according to the label element content of the internal test mineral sample using the preprocessed internal test spectral data set. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

[0014] According to a third aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0015] In a fourth aspect of the present application, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods described in the first aspect when executing the program.

[0016] The present application provides an element detection method, device and equipment based on model fusion migration, the method comprising: collecting laser-induced breakdown spectra of training mineral samples in a laboratory scenario to obtain a training spectrum data set; collecting laser-induced breakdown spectra of internal test mineral samples in the laboratory scenario to obtain an internal test spectrum data set; using the training spectrum data set and the internal test spectrum data set to train and test a plurality of machine learning models respectively to obtain a plurality of corresponding regression models; collecting laser-induced breakdown spectra of external test mineral samples of each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, using the external test spectrum data set of each actual application scenario and the plurality of regression models to optimize a pre-constructed fusion weight matrix to obtain an optimized fusion weight matrix for each actual application scenario; collecting the spectrum of the mineral to be tested in the target actual application scenario to obtain the mineral spectrum data set to be tested, The data set is preprocessed to obtain a preprocessed mineral spectral data set to be tested; the preprocessed mineral spectral data set to be tested is respectively input into the multiple regression models to obtain the current predicted content of several elements corresponding to the preprocessed mineral spectral data set to be tested by each regression model; the current predicted contents of several elements corresponding to the preprocessed mineral spectral data set to be tested by the multiple regression models are combined through the optimized fusion weight matrix of the target actual application scenario to obtain the fused predicted content of several elements in the mineral to be tested; in this way, since different types of regression models have different preferences for different elements, the spectra are calculated using regression models with different characteristics, and the predicted contents of each model are combined to achieve the optimized adaptation of different types of laboratory regression models to the minerals to be tested in the application field, effectively overcoming the defect that a single model cannot optimize and generalize, improving the detection accuracy of multiple element contents in complex minerals, and meeting the requirements of in-situ and online detection and analysis of complex minerals. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present application. The same reference symbols are used throughout the drawings to represent the same components. In the drawings: Figure 1 A schematic flow chart of a method for detecting mineral element content based on model fusion migration according to an embodiment of the present application is shown; Figure 2 A schematic diagram of the process of optimizing the fusion weight matrix and generating the final optimized fusion prediction content according to one embodiment of the present application is shown; Figure 3A comparison chart of the element contents and label element contents predicted for an external bauxite crushed rock sample using a model fusion and an independent regression model according to an embodiment of the present application is shown; Figure 4 A comparison chart showing the content of each element and the content of the label element predicted for the bauxite crushed stone sample to be tested by the model fusion and independent regression models according to one embodiment of the present application is shown; Figure 5 A schematic structural diagram of a mineral element content detection device based on model fusion migration according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0018] This application provides an element detection method based on model fusion migration, such as Figure 1 As shown, the method mainly includes the following steps: S110, collecting laser-induced breakdown spectra of training mineral samples in a laboratory scene to obtain a training spectrum data set; collecting laser-induced breakdown spectra of internal test mineral samples in the laboratory scene to obtain an internal test spectrum data set; using the training spectrum data set and the internal test spectrum data set to train and test multiple machine learning models respectively, to obtain corresponding multiple regression models.

[0019] In this application, the regression model refers to a model obtained by training and testing a pre-built machine learning model in a laboratory scenario. The machine learning model may include: shallow neural network, XGBoost, convolutional neural network, etc.

[0020] Then, laser-induced breakdown spectroscopy data of training mineral samples can be collected in a laboratory scenario to obtain a training spectral dataset; laser-induced breakdown spectroscopy data of internal test mineral samples can be collected in a laboratory scenario to obtain an internal test spectral dataset; and the training spectral dataset and the internal test spectral dataset are used to train and test multiple machine learning models to obtain corresponding multiple regression models.

[0021] It should be noted that, in order to facilitate training, multiple training mineral samples (pre-collected or pre-prepared) are generally prepared in laboratory scenarios. Each training mineral sample can be about 5 kg. Therefore, the laser-induced breakdown spectroscopy data of multiple training mineral samples can be directly collected in the laboratory scenario to obtain a training spectral dataset.

[0022] To prevent overfitting during training, which could affect the prediction accuracy of the regression model, the internal test mineral samples in this application were separated from laboratory samples at the same time as the training mineral samples. The separation ratio was 8:2 (training mineral sample: internal test mineral sample).

[0023] Then, the laser-induced breakdown spectrum of the internal test mineral sample can be collected in a laboratory scenario to obtain an internal test spectrum data set.

[0024] In one embodiment, a plurality of machine learning models are trained and tested using a training spectral dataset and an internal test spectral dataset, respectively, to obtain corresponding plurality of regression models, including: Obtain the actual content of several elements of the training mineral samples as the label element content of the training mineral sample set; obtain the actual content of the internal test mineral samples as the label element content of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as input variables, various machine learning models are trained under the supervision of the elemental content of the training mineral samples. Cross-validation and model iteration are performed during the training process until the training termination condition is met, resulting in multiple regression models. For each regression model, the performance of the regression model is tested based on the label element content of the internal test mineral samples using the preprocessed internal test spectral dataset. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

[0025] Specifically, please refer to Figure 2 First, multiple training mineral samples are collected or prepared in a laboratory setting, and multiple internal test mineral samples are collected or prepared in a laboratory setting. The number of samples in the training mineral sample set Greater than the number of samples in the internally tested mineral sample set Generally speaking, the training mineral sample set can contain hundreds of mineral samples, and the internal test mineral sample set can contain dozens of mineral samples.

[0026] Assign corresponding label element content to several elements of the training mineral samples in the training mineral sample set , ; Assign corresponding label element content to several elements of the internal test mineral samples of the internal test mineral sample set , ;in, m is any one of several elements, m =1…… M , M is the number of elements.

[0027] For each training mineral sample, laser-induced breakdown spectroscopy data are collected for the training mineral sample and the internal test mineral sample under laboratory scene conditions and configurations; for each training mineral sample, several repeated laser-induced breakdown spectroscopy data are collected for each training mineral sample under statistically equivalent experimental conditions to form an independent training spectral data set; the number of repeated spectra ranges from dozens to thousands.

[0028] Similarly, for each internal test mineral sample, under statistically equivalent experimental conditions, several repeated laser-induced breakdown spectroscopy data of each internal test mineral sample are collected to form an independent internal test spectral data set; the number of repeated spectra ranges from dozens to thousands.

[0029] Then, the spectral data in the training spectral dataset and the internal test spectral dataset need to be preprocessed. For example, the dataset can be preprocessed for common characteristics first, and then the dataset can be preprocessed for individual characteristics that match the algorithm based on the category of the machine learning model.

[0030] Common preprocessing is a processing method shared by different categories of machine learning models, which can include the removal of abnormal spectral data, baseline removal, and other processing.

[0031] Personalized preprocessing can include normalization (normalization of spectral total intensity, unit vector, and internal standard elements, etc.), averaging, regularization, etc.

[0032] The training spectral dataset is then used to train various machine learning models independently. During the training process, the preprocessed training spectral dataset is used as the input variable. Under the supervision of the elemental content of the training mineral sample labels, various machine learning models are trained. Cross-validation and model iteration are performed during the training process until the training termination condition (such as the number of training cycles) is met, resulting in multiple regression models.

[0033] In order to evaluate the generalization performance of multiple regression models, it is also necessary to use the internal test spectral dataset to test each type of regression model separately. For any regression model, if the generalization performance is determined to meet the requirements based on the test results, the regression model can be used directly.

[0034] If it is determined that the generalization performance of the regression model cannot meet the requirements (for example, overfitting occurs), it is necessary to adjust the training strategy (for example, by adjusting the model complexity, adjusting the hyperparameters, adjusting the optimizer type, etc.) and retrain the regression model until the performance of the regression model meets the requirements.

[0035] S111, collecting laser-induced breakdown spectra of external test mineral samples of each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and using the external test spectrum data set of each actual application scenario and the multiple regression models to optimize the pre-constructed fusion weight matrix to obtain an optimized fusion weight matrix for each actual application scenario.

[0036] In order to improve the generalization performance of various models trained and tested in the laboratory in practical applications, the external test mineral samples of this application are collected from various practical application scenarios, that is, the external test mineral samples are transported from various practical application scenarios (such as alumina plants) to the laboratory. Each practical application scenario can correspond to multiple external test mineral samples, and each external test mineral sample is approximately 50 kg.

[0037] Then, the laser-induced breakdown spectrum of the external test mineral sample can be collected in the simulated application scenario to obtain an external test spectrum data set.

[0038] After the external test mineral samples are transported to the laboratory, laser-induced breakdown spectroscopy data can be collected for each external test mineral sample in a simulated application scenario, thereby generating an external test spectral dataset. The simulated application scenario can be a dynamic scenario in the laboratory where the external test mineral samples are moving on a conveyor belt, simulating the minerals in the application site.

[0039] This application takes into account that different types of regression models have different characteristics, and have fitting efficiency biases for different elements in minerals, and data characteristic adaptation cannot be optimized at the same time. For example, some regression models are more suitable for predicting aluminum oxide Al2O3, some regression models are more suitable for predicting silicon dioxide SiO2, and some regression models are more suitable for predicting iron oxide Fe2O3, etc. Different types of regression models have fitting efficiency biases for elements in minerals. In order to improve the prediction accuracy of each element in the mineral to be tested and make the final predicted content of each element closer to the actual content, this application needs to optimize the pre-constructed fusion weight matrix. Finally, when predicting the content of each element, the predicted content of each regression model is weighted and combined using the optimized fusion weight matrix.

[0040] Because minerals in different practical application scenarios have their own unique mineral properties, it is necessary to collect laser-induced breakdown spectroscopy data of external test mineral samples for different practical application scenarios in the simulated application scenarios to obtain the external test spectral dataset corresponding to each practical application scenario. The pre-constructed fusion weight matrix is ​​optimized using the external test spectral dataset corresponding to each practical application scenario and multiple regression models to obtain the optimized fusion weight matrix corresponding to each practical application scenario. There is a certain correspondence between each practical application scenario and the corresponding optimized fusion weight matrix, which facilitates the subsequent fusion process to find the corresponding optimized fusion weight matrix based on the practical application scenario.

[0041] The pre-built fusion weight matrix includes the weights and prediction biases for each regression model when predicting each current element. The initial values ​​for each element in the fusion weight matrix are pre-set, but these initial values ​​are not necessarily the optimal weights. Therefore, the fusion weight matrix needs to be optimized.

[0042] In one embodiment, the pre-built fusion weight matrix is ​​optimized using an external test spectral dataset of each actual application scenario and multiple regression models. Before obtaining the optimized fusion weight matrix for each actual application scenario, it is necessary to pre-build the fusion weight matrix, including: The dimension of the fusion weight matrix is ​​determined according to the number of elements and the number of regression models; the dimension of the fusion weight matrix is The fusion weight matrix includes the weight and prediction bias of each regression model when calculating the fusion prediction content of each element; M is the number of elements, N is the number of regression models; In the fusion weight matrix, for each element, the weight contributed by each regression model when predicting each element and the prediction deviation are respectively assigned preset initial values ​​to complete the construction of the fusion weight matrix.

[0043] In one embodiment, the pre-built fusion weight matrix is ​​optimized using external test spectral datasets of various actual application scenarios and multiple regression models, including: preprocessing the external test spectral dataset to obtain a preprocessed external test spectral dataset; Obtaining actual contents of several elements of an externally tested mineral sample as label element contents of the externally tested mineral sample; Based on the preset optimization algorithm, the predicted content of several elements in the external test mineral samples using multiple regression models is used as input variables, and the label element content of the external test mineral samples is used as the target. The fusion weight matrix is ​​iteratively optimized until the preset requirements are met, and the optimized fusion weight matrix is ​​output; wherein, The predicted contents of several elements in the external test mineral samples by multiple regression models are obtained by operating the pre-processed external test spectral data sets separately by multiple regression models.

[0044] In one embodiment, based on a preset optimization algorithm, the predicted contents of several elements in the external test mineral samples using multiple regression models are used as input variables, and the label element contents of the external test mineral samples are used as targets. The fusion weight matrix is ​​iteratively optimized, including: Based on the fusion weight matrix corresponding to each iteration, the predicted contents of several elements are combined and added with the predicted deviations of the corresponding elements in the fusion weight matrix to obtain the fusion predicted contents of several elements in the external test mineral sample; With the goal of reducing the residual value between the fusion prediction content of several elements in the external test mineral samples and the label element content, the matrix elements of the fusion weight matrix are used as the parameters to be optimized, and the optimization algorithm is used to iteratively optimize the fusion weight matrix until the optimization termination condition is reached; Among them, the residual value is calculated by a pre-built residual function, and the residual value is used to characterize the error between the fusion predicted content of several elements in the external test mineral sample and the content of their respective label elements.

[0045] Specifically, we first need to obtain an external test spectral dataset. The method for obtaining an external test spectral dataset is as follows: In order to improve the optimization accuracy of the fusion weight matrix, the external test mineral samples of this application are transported from each application site to the laboratory. Each actual application scenario uses multiple external test mineral samples, and each external test mineral sample is approximately 50 kg.

[0046] Multiple external test mineral samples form an external test mineral sample set, and the number of samples in the external test mineral sample set is ,In order to better fit the actual application scenario and improve the fusion accuracy of the fusion weight matrix in ,the actual application scenario, when collecting the spectrum of the ,external test mineral samples in the simulated application scenario, the spectrum of the ,external test mineral samples is collected when they are in a dynamic state (such as ,simulating conveyor belt mineral transportation), and the external test ,spectral data set is obtained.

[0047] Specifically, for each external test mineral sample in the external test mineral sample set, several repeated spectra are collected; the number of repeated spectra ranges from tens to thousands, and an external test spectrum data set is obtained. The spectral data in the external test spectrum data set is preprocessed, and the external test spectrum data set is first subjected to common preprocessing, and then the external test spectrum data set is subjected to individual preprocessing according to the type of machine learning model to obtain a preprocessed external test spectrum data set. Among them, the methods of common preprocessing and individual preprocessing can be referred to as described above, so they will not be repeated here. Among them, several elements of the external test mineral sample also need to predetermine the corresponding label element content , .

[0048] Then, it is necessary to construct the fusion weight matrix, the fusion weight matrix The dimension is , m is any one of several elements, m =1…… M ; n is any of the regression models, n =1…… N , i is the number of iterations.

[0049] Assume that there are three regression models A, B, and C (model A is the first regression model, model B is the second regression model, and model C is the third regression model), and the number of elements that need to be predicted includes 3, namely: Al2O3, SiO2, and Fe2O3 (Al2O3 is the first element, SiO2 is the second element, and Fe2O3 is the third element). Then the dimension of the fusion weight matrix is ​​12 dimensions, as shown in Table 1.

[0050] Table 1 As can be seen from Table 1, the fusion weight matrix contains the weight parameters of the current prediction content of each regression model for each element and the contribution to the fusion prediction content, as well as the prediction deviation corresponding to each element.

[0051] Taking Al2O3 as an example, the weight parameter of the current predicted content of model A to the fused predicted content is a1, the weight parameter of the current predicted content of model B to the fused predicted content is b1, the weight parameter of the current predicted content of model C to the fused predicted content is c1, and the prediction deviation is d1.

[0052] Taking SiO2 as an example, the weight parameter of the current predicted content of model A to the fused predicted content is a2, the weight parameter of the current predicted content of model B to the fused predicted content is b2, the weight parameter of the current predicted content of model C to the fused predicted content is c2, and the prediction deviation is d2.

[0053] Taking Fe2O3 as an example, the weight parameter of the current predicted content of model A to the fused predicted content is a3, the weight parameter of the current predicted content of model B to the fused predicted content is b3, the weight parameter of the current predicted content of model C to the fused predicted content is c3, and the prediction deviation is d3.

[0054] Since the initial values ​​of the parameters in the fusion weight matrix are pre-set, generally speaking, , , that is, when i =0 (initial state), at this time, for each element, the weight contributed by each regression model to the prediction of the current element is 0.3333, and the initial value of the prediction deviation is Both are 0.0000.

[0055] However, the initial values ​​in the fusion weight matrix are not necessarily optimal. Therefore, it is also necessary to iteratively optimize the initial fusion weight matrix based on a preset optimization algorithm, using the predicted content of several elements in the external test mineral samples of multiple regression models as input variables and the label element content of the external test mineral samples as the target, until the preset requirements are met and the optimized fusion weight matrix is ​​output.

[0056] Among them, the optimization algorithm can be a genetic algorithm. In the optimization process of the genetic algorithm, it is necessary to first determine the objective function, population size, mutation rate, and number of iterations. The objective function in this application is the error function between the fusion prediction content of several elements contained in the external test mineral sample and the content of several element labels in the pre-processed external test spectral data set. RMSE is used to characterize the residual. The residual function is shown in formula (1): (1) In formula (1), For the i The first iteration m The fusion prediction content of the elements and the m The residual between the label element contents of the elements; For the number of samples in the external test mineral sample, For any one of the external test mineral samples, For the i The first iteration m The fusion prediction content of the elements, For the External testing of mineral samples m The tag element content of each element. i =0, When the initial state m The fusion prediction content of the elements and the m The residuals between the label element contents of the elements.

[0057] Among them, i At the first iteration, External testing of mineral samples m Fusion prediction content of elements It can be determined according to formula (2): (2) In formula (2), n For any regression model, N is the number of regression models, For the i At the first iteration, n The regression model for m The weight of the contribution of each element in the prediction, For the n The regression model for m The current predicted content of the output element, For the i At the first iteration, the fusion weight matrix m The prediction deviation of each element.

[0058] Finally, with the goal of reducing the residual value between the fusion prediction content and the label element content of several elements in all external test mineral samples, the matrix element of the fusion weight matrix is ​​used as the parameter to be optimized, and the optimization algorithm is used to iteratively optimize the fusion weight matrix until the optimization termination condition is reached, and finally the optimized fusion weight matrix is ​​output. , I The optimization termination condition can be that the preset maximum number of iterations is reached or the residual value meets the expected requirements.

[0059] In this way, the optimization weight matrix corresponding to each application scenario can be determined according to the above method.

[0060] S112 , collecting spectral data of the mineral to be tested in the target actual application scenario to obtain a spectral dataset of the mineral to be tested, and preprocessing the spectral dataset of the mineral to be tested to obtain a preprocessed spectral dataset of the mineral to be tested.

[0061] The spectral data described in this application is laser-induced breakdown spectroscopy (LIBS) data. In a target application, a LIBS instrument can be used to emit high-energy laser pulses to a mineral to be tested in a practical application scenario, generating a plasma. The emitted light emitted during the cooling process of the plasma is then collected to obtain spectral data of the mineral to be tested. The target application scenario is any one of all practical application scenarios.

[0062] After collecting spectral data for all the minerals to be tested, a spectral dataset of the minerals to be tested is obtained. This spectral dataset is then preprocessed to obtain a preprocessed spectral dataset of the minerals to be tested. The preprocessing method for the spectral dataset of the minerals to be tested can be referenced above for the preprocessing method for the training spectral dataset, so it will not be repeated here.

[0063] S113 , inputting the pre-processed mineral spectral dataset to be tested into the multiple regression models respectively, and obtaining the current predicted content of several elements corresponding to the pre-processed mineral spectral dataset for each regression model.

[0064] As mentioned above, this application takes into account that different types of regression models have different characteristics, have fitting efficiency biases for different elements in minerals, and cannot simultaneously optimize data characteristics. For example, some regression models are more suitable for predicting Al2O3, some regression models are more suitable for predicting SiO2, and some regression models are more suitable for predicting Fe2O3, etc. If a single regression model is used to predict the content of multiple elements in the mineral to be tested, the prediction accuracy cannot be optimized generalization performance.

[0065] Based on this, after obtaining multiple trained regression models, this application inputs the preprocessed mineral spectral dataset into each of these regression models. Using these pretrained regression models, the preprocessed mineral spectral dataset is calculated, resulting in the current predicted content of several elements corresponding to each regression model. These predicted content of several elements are then fused to ultimately improve the prediction accuracy of multiple element contents. These elements include, but are not limited to, Al2O3, SiO2, and Fe2O3.

[0066] Each regression model outputs a set of current predicted contents for several elements.

[0067] S114, combining the current predicted contents of several elements corresponding to the pre-processed mineral spectral dataset of the multiple regression models through the optimized fusion weight matrix of the target actual application scenario to obtain the fused predicted contents of several elements in the mineral to be tested.

[0068] As mentioned above, different types of regression models have fitting efficiency biases for elements in minerals. In order to improve the prediction accuracy of each element in the mineral to be tested and make the final predicted content closer to the actual content, this application also needs to combine the current predicted contents of several elements corresponding to the spectral data of multiple regression models based on the optimized fusion weight matrix, so as to obtain the fused predicted content of several elements in the mineral to be tested.

[0069] Since each practical application scenario has a corresponding optimization fusion weight matrix, it is necessary to first determine the corresponding optimization fusion weight matrix based on the target practical application scenario of the mineral to be tested, including: Obtain the scenario identifier of the target application scenario; According to the scenario identifier of the target application scenario, the corresponding optimized fusion weight matrix is ​​searched in the corresponding relationship between the actual application scenario and the optimized fusion weight matrix.

[0070] After finding the optimized fusion weight matrix, the current predicted contents of several elements corresponding to the pre-processed mineral spectral dataset of the target mineral can be combined by the optimized fusion weight matrix of the target actual application scenario to obtain the fused predicted contents of several elements in the target mineral, including: For each element, the weight contributed by each regression model in the optimized fusion weight matrix when predicting each element, as well as the current predicted content of each regression model for each element, are weighted and combined to obtain the weighted combined predicted content of each element; The weighted combined predicted content of each element is added to the predicted deviation of the corresponding element in the optimized fusion weight matrix to obtain the fused predicted content of each element.

[0071] Specifically, a variety of regression models are used to calculate the spectral data in the pre-processed mineral spectral data set. Each type of regression model will output a set of current predicted contents corresponding to several elements. In order to improve the prediction accuracy of the content of each element, it is necessary to combine the optimized fusion weight matrix and the current predicted contents corresponding to several elements output by each type of regression model to determine the fused predicted content of each element.

[0072] Continuing with the Al2O3 element in Table 1 above, for example, assuming that Table 1 is the fusion weight matrix after optimization, the Al2O3 element content predicted by model A is f1, the Al2O3 element content predicted by model B is f2, and the Al2O3 element content predicted by model C is f3, then the fusion predicted content of the Al2O3 element is: .

[0073] In this way, since different types of regression models have different preferences for different elements, regression models with different characteristics are used to calculate the spectrum, and the predicted contents of each model are combined to achieve optimized adaptation of different types of laboratory regression models to the minerals to be tested in the application field, effectively overcoming the defect that a single model cannot optimize and generalize, improving the accuracy of detection of multiple element contents in complex minerals, and meeting the requirements of in-situ and online detection and analysis of complex minerals.

[0074] Furthermore, to verify the predictive effectiveness of analyzing spectral data using regression models with different characteristics and fusing the predicted elemental content of each model, this application also collected laser-induced breakdown spectroscopy data from 70 bauxite gravel samples in a laboratory setting to obtain a training spectral dataset; collected laser-induced breakdown spectroscopy data from another 20 bauxite gravel samples to obtain an internal test spectral dataset; and collected laser-induced breakdown spectroscopy data from four independent bauxite gravel samples to obtain an external test spectral dataset. Three machine learning algorithms were selected: a back-propagation shallow neural network, XGBoost, and a convolutional neural network. These three algorithms were used to build models, resulting in Machine Learning Model 1, Machine Learning Model 2, and Machine Learning Model 3.

[0075] According to the training strategy mentioned above, machine learning model 1, machine learning model 2, and machine learning model 3 are trained respectively to obtain regression model 1, regression model 2, and regression model 3.

[0076] Then, the optimization strategy mentioned above is used to optimize the fusion weight matrix to obtain the optimized fusion weight matrix. The optimized fusion weight matrix is ​​shown in Table 2: Table 2 Regression model 1, regression model 2, and regression model 3 were used to predict the contents of Al2O3, SiO2, and Fe2O3 in the external bauxite crushed stone test samples in the simulated application scenario, and the fusion weight matrix in Table 2 was used to fuse the contents predicted by each model. The contents predicted by the three regression models for the three elements, the fusion predicted contents of the three elements, and the label element contents of the three elements are shown in Figure 2. Figure 3 In Table 2, “wt” refers to weight and wt% refers to mass percentage.

[0077] Figure 3 is the content of each element predicted by three regression models. Figure 3 middle," " is the label element content of each element, " is the predicted content of model 1," " is the predicted content of model 2," " is the predicted content of model 3," ” is the fusion prediction content.

[0078] from Figure 3 It can be seen that the fusion prediction content is closer to the label element content as a whole than the content predicted by each model separately, indicating that the fusion prediction content is more globally optimized.

[0079] Furthermore, the performance indicators of the above-mentioned independent regression models 1, 2 and 3 were analyzed, and the performance indicators of the fusion prediction content were analyzed, and the results are shown in Table 3.

[0080] Table 3 In Table 3, RE is the relative error and RMSE is the root mean square error. It can be seen that for the three elements Al2O3, SiO2 and Fe2O3, the relative errors of the fused predicted content obtained in this application are smaller than the relative errors of the predicted amounts of the independent regression models, and the root mean square errors of the fused predicted contents obtained in this application are smaller than or equal to the root mean square errors of the predicted amounts of the independent regression models, indicating that the generalized prediction accuracy of this application can be ensured.

[0081] In addition, the present application also uses regression model 1, regression model 2 and regression model 3 to predict the Al2O3, SiO2 and Fe2O3 contents of the bauxite crushed stone samples to be tested in actual application scenarios (the bauxite crushed stone samples to be tested and the external test bauxite crushed stone samples come from the same alumina plant), and uses the optimized fusion weight matrix in Table 2 to fuse the contents predicted by each model separately. The contents predicted by the three regression models for the three elements, the fusion predicted contents of the three elements and the label element contents of the three elements are shown in Figure 2. Figure 4 shown. Figure 4 It is the content of each element predicted by three regression models.

[0082] exist Figure 4 middle," " is the label element content of each element, " is the predicted content of model 1," " is the predicted content of model 2," " is the predicted content of model 3," ” is the fusion prediction content.

[0083] from Figure 4 It can be seen that the fusion prediction content of each element is closer to the label element content as a whole than the content predicted by each model individually, indicating that the fusion prediction content is more globally optimized.

[0084] Furthermore, the performance indicators of the above-mentioned independent regression models 1, 2 and 3 were analyzed, and the performance indicators of the fusion prediction content were analyzed, and the results are shown in Table 4.

[0085] Table 4 In Table 4, RE is the relative error and RMSE is the root mean square error. It can be seen that for the three elements Al2O3, SiO2 and Fe2O3, the relative errors of the fused predicted content obtained in this application are smaller than the relative errors of the predicted amounts of the independent regression models, and the root mean square errors of the fused predicted contents obtained in this application are smaller than the root mean square errors of the predicted amounts of the independent regression models, indicating that the generalized prediction accuracy of this application can be ensured.

[0086] Based on the same inventive concept as in the above embodiment, the present application also provides an element detection device for model fusion migration, such as Figure 5 As shown, the device includes: A training unit 51 is configured to collect laser-induced breakdown spectra of training mineral samples in a laboratory setting to obtain a training spectral dataset; collect laser-induced breakdown spectra of internal test mineral samples in the laboratory setting to obtain an internal test spectral dataset; and train and test a plurality of machine learning models using the training spectral dataset and the internal test spectral dataset to obtain corresponding plurality of regression models. an optimization unit 52 for collecting laser-induced breakdown spectra of external test mineral samples for each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and optimizing a pre-constructed fusion weight matrix using the external test spectrum data set for each actual application scenario and the multiple regression models to obtain an optimized fusion weight matrix for each actual application scenario; The acquisition unit 53 is used to acquire the spectrum of the mineral to be tested in the target actual application scenario to obtain a spectrum data set of the mineral to be tested, and preprocess the spectrum data set of the mineral to be tested to obtain a preprocessed spectrum data set of the mineral to be tested; The prediction unit 54 is used to input the pre-processed mineral spectrum data set to the multiple regression models respectively, and obtain the current predicted content of the elements corresponding to the pre-processed mineral spectrum data set for each regression model; The fusion unit 55 is used to combine the current predicted contents of several elements corresponding to the pre-processed mineral spectral data set by the multiple regression models through the optimized fusion weight matrix of the target actual application scenario to obtain the fused predicted contents of several elements in the mineral to be tested.

[0087] The training unit 51 is specifically used for: Obtaining the actual contents of several elements of the training mineral samples as the label element contents of the training mineral sample set; obtaining the actual contents of the internal test mineral samples as the label element contents of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as an input variable, various machine learning models are trained under the supervision of the elemental content of the training mineral samples, and cross-validation and model iteration are performed during the training process until a training termination condition is reached, thereby obtaining multiple regression models; For each regression model, the performance of the regression model is tested respectively according to the label element content of the internal test mineral sample using the preprocessed internal test spectral data set. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

[0088] The optimization unit 52 is specifically used for: Preprocessing the external test spectral dataset to obtain a preprocessed external test spectral dataset; obtaining actual contents of several elements of the external test mineral sample as label element contents of the external test mineral sample; Based on a preset optimization algorithm, the predicted contents of several elements in the external test mineral samples by the multiple regression models are used as input variables, and the label element contents of the external test mineral samples are used as targets. The fusion weight matrix is ​​iteratively optimized until the preset requirements are met, and the optimized fusion weight matrix is ​​output; wherein, The predicted contents of several elements in the external test mineral sample by the multiple regression models are obtained by respectively calculating the pre-processed external test spectral data set by the multiple regression models.

[0089] The optimization unit 52 is further configured to: The dimension of the fusion weight matrix is ​​determined according to the number of the elements and the number of the regression models; the dimension of the fusion weight matrix is The fusion weight matrix includes the weight and prediction deviation of each regression model when calculating the fusion prediction content of each element; M is the number of the elements, N is the number of regression models; In the fusion weight matrix, for each element, the weight contributed by each regression model when predicting each element and the prediction deviation are respectively assigned preset initial values.

[0090] The optimization unit 52 is specifically used for: Based on the fusion weight matrix corresponding to each iteration, the predicted contents of the plurality of elements are combined and added with the predicted deviations of the corresponding elements in the fusion weight matrix to obtain the fusion predicted contents of the plurality of elements in the external test mineral sample; With the goal of reducing the residual value between the fused predicted content and the label element content of several elements in the external test mineral sample, the matrix elements of the fusion weight matrix are used as parameters to be optimized, and the fusion weight matrix is ​​iteratively optimized using the optimization algorithm until the optimization termination condition is reached; The residual value is calculated by a pre-constructed residual function, and is used to characterize the error between the fusion predicted content of several elements in the external test mineral sample and the content of each label element.

[0091] The fusion unit 55 is specifically used for: For each element, performing a weighted combination of the weight contributed by each regression model in the optimized fusion weight matrix when predicting each element and the current predicted content of each element by each regression model to obtain a weighted combined predicted content of each element; The weighted combined prediction content of each element is added to the prediction deviation of the corresponding element in the optimized fusion weight matrix to obtain the fused prediction content of each element.

[0092] Since the element detection device based on model fusion migration introduced in the embodiments of this application is a device used to implement the element detection method based on model fusion migration in the embodiments of this application, those skilled in the art will be able to understand the specific structure and variations of the device based on the methods introduced in the embodiments of this application, and therefore, no further details will be given here. All devices used in the methods of the embodiments of this application fall within the scope of protection to be provided by this application.

[0093] Based on the same inventive concept, the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any step of the method described above is implemented.

[0094] Based on the same inventive concept, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above-mentioned methods when executed by a processor.

[0095] Through one or more embodiments of the present application, the present application has the following beneficial effects or advantages: This application provides an element detection method, device, medium, and equipment based on model fusion migration, the method comprising: Collect laser-induced breakdown spectra of training mineral samples in a laboratory scenario to obtain a training spectrum data set; collect laser-induced breakdown spectra of internal test mineral samples in the laboratory scenario to obtain an internal test spectrum data set; use the training spectrum data set and the internal test spectrum data set to train and test multiple machine learning models respectively to obtain corresponding multiple regression models; collect laser-induced breakdown spectra of external test mineral samples of each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, use the external test spectrum data set of each actual application scenario and the multiple regression models to optimize a pre-constructed fusion weight matrix to obtain an optimized fusion weight matrix for each actual application scenario; collect the spectrum of the mineral to be tested in the target actual application scenario to obtain a mineral spectrum data set to be tested, pre-process the mineral spectrum data set to be tested to obtain a pre-processed mineral to be tested spectral data set; inputting the pre-processed mineral spectral data set to be tested into the multiple regression models respectively, and obtaining the current predicted content of several elements corresponding to the pre-processed mineral spectral data set to be tested by each regression model; combining the current predicted contents of several elements corresponding to the pre-processed mineral spectral data set to be tested by the multiple regression models through the optimized fusion weight matrix of the target actual application scenario, and obtaining the fusion predicted content of several elements in the mineral to be tested; in this way, since different types of regression models have different preferences for different elements, regression models with different characteristics are used to calculate the spectrum, and the predicted contents of each model are combined to achieve the optimized adaptation of different types of laboratory regression models to the minerals to be tested in the application field, effectively overcoming the defect that a single model cannot optimize and generalize, improving the detection accuracy of multiple element contents in complex minerals, and meeting the in-situ and online detection and analysis requirements of complex minerals.

[0096] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0097] The above is only a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. An element detection method based on model fusion migration, characterized in that: The method comprises: Collecting laser-induced breakdown spectra of training mineral samples in a laboratory setting to obtain a training spectral dataset; collecting laser-induced breakdown spectra of internal test mineral samples in the laboratory setting to obtain an internal test spectral dataset; and using the training spectral dataset and the internal test spectral dataset to train and test multiple machine learning models, respectively, to obtain corresponding multiple regression models; Collecting laser-induced breakdown spectra of external test mineral samples for each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and optimizing a pre-constructed fusion weight matrix using the external test spectrum data set for each actual application scenario and the multiple regression models to obtain an optimized fusion weight matrix for each actual application scenario; Collecting a spectrum of a mineral to be tested in a target practical application scenario to obtain a spectrum dataset of the mineral to be tested, and preprocessing the spectrum dataset of the mineral to be tested to obtain a preprocessed spectrum dataset of the mineral to be tested; Inputting the pre-processed mineral spectral dataset to be tested into the multiple regression models respectively, and obtaining the current predicted content of several elements corresponding to the pre-processed mineral spectral dataset for each regression model; By combining the current predicted contents of several elements corresponding to the pre-processed mineral spectral data set to be tested using the optimized fusion weight matrix of the target actual application scenario, the fusion predicted contents of several elements in the mineral to be tested are obtained.

2. The method according to claim 1, wherein The plurality of machine learning models are trained and tested using the training spectral dataset and the internal test spectral dataset to obtain corresponding plurality of regression models, including: Obtaining the actual contents of several elements of the training mineral samples as the label element contents of the training mineral sample set; obtaining the actual contents of the internal test mineral samples as the label element contents of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as an input variable, various machine learning models are trained under the supervision of the elemental content of the training mineral samples, and cross-validation and model iteration are performed during the training process until a training termination condition is reached, thereby obtaining multiple regression models; For each regression model, the performance of the regression model is tested respectively according to the label element content of the internal test mineral sample using the preprocessed internal test spectral data set. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

3. The method according to claim 1, wherein The optimizing of the pre-built fusion weight matrix using the external test spectral dataset and the multiple regression models includes: Preprocessing the external test spectral dataset to obtain a preprocessed external test spectral dataset; obtaining actual contents of several elements of the external test mineral sample as label element contents of the external test mineral sample; Based on a preset optimization algorithm, the predicted contents of several elements in the external test mineral samples by the multiple regression models are used as input variables, and the label element contents of the external test mineral samples are used as targets. The fusion weight matrix is ​​iteratively optimized until the preset requirements are met, and the optimized fusion weight matrix is ​​output; wherein, The predicted contents of several elements in the external test mineral sample by the multiple regression models are obtained by respectively calculating the pre-processed external test spectral data set by the multiple regression models.

4. The method according to claim 1, wherein Before optimizing the pre-built fusion weight matrix using the external test spectral dataset and the multiple regression models, the method further includes: The dimension of the fusion weight matrix is ​​determined according to the number of the elements and the number of the regression models; the dimension of the fusion weight matrix is The fusion weight matrix includes the weight and prediction deviation of each regression model when calculating the fusion prediction content of each element; M is the number of the elements, N is the number of regression models; In the fusion weight matrix, for each element, the weight contributed by each regression model when predicting each element and the prediction deviation are respectively assigned preset initial values.

5. The method according to claim 3, wherein The preset optimization algorithm is based on multiple regression models, and the predicted content of several elements in the external test mineral sample is used as input variables. The content of the label elements in the external test mineral sample is used as the target, and the fusion weight matrix is ​​iteratively optimized, including: Based on the fusion weight matrix corresponding to each iteration, the predicted contents of the plurality of elements are combined and added with the predicted deviations of the corresponding elements in the fusion weight matrix to obtain the fusion predicted contents of the plurality of elements in the external test mineral sample; With the goal of reducing the residual value between the fused predicted content and the label element content of several elements in the external test mineral sample, the matrix elements of the fusion weight matrix are used as parameters to be optimized, and the fusion weight matrix is ​​iteratively optimized using the optimization algorithm until the optimization termination condition is reached; The residual value is calculated by a pre-constructed residual function, and is used to characterize the error between the fusion predicted content of several elements in the external test mineral sample and the content of each label element.

6. The method according to claim 1, wherein The optimized fusion weight matrix of the target actual application scenario is combined with the current predicted contents of several elements corresponding to the pre-processed mineral spectral dataset of the multiple regression models to obtain the fused predicted contents of several elements in the mineral to be tested, including: For each element, performing a weighted combination of the weight contributed by each regression model in the optimized fusion weight matrix when predicting each element and the current predicted content of each element by each regression model to obtain a weighted combined predicted content of each element; The weighted combined prediction content of each element is added to the prediction deviation of the corresponding element in the optimized fusion weight matrix to obtain the fused prediction content of each element.

7. An element detection device based on model fusion migration, characterized in that: The device comprises: A training unit is configured to collect laser-induced breakdown spectra of training mineral samples in a laboratory setting to obtain a training spectral dataset; collect laser-induced breakdown spectra of internal test mineral samples in the laboratory setting to obtain an internal test spectral dataset; and use the training spectral dataset and the internal test spectral dataset to train and test a plurality of machine learning models, respectively, to obtain a plurality of corresponding regression models; an optimization unit, configured to collect laser-induced breakdown spectra of external test mineral samples of each actual application scenario in a simulated application scenario to obtain an external test spectrum data set, and optimize a pre-constructed fusion weight matrix using the external test spectrum data set of each actual application scenario and the multiple regression models to obtain an optimized fusion weight matrix for each actual application scenario; An acquisition unit is used to acquire a spectrum of a mineral to be tested in a target actual application scenario to obtain a spectrum data set of the mineral to be tested, and preprocess the spectrum data set of the mineral to be tested to obtain a preprocessed spectrum data set of the mineral to be tested; A prediction unit, configured to input the pre-processed mineral spectrum dataset to be tested into the multiple regression models respectively, and obtain a current predicted content of several elements corresponding to the pre-processed mineral spectrum dataset for each regression model; A fusion unit is used to combine the current predicted contents of several elements corresponding to the preprocessed mineral spectral data set by the multiple regression models through the optimized fusion weight matrix of the target actual application scenario to obtain a fused predicted content of several elements in the mineral to be tested.

8. The device according to claim 7, wherein The training unit is specifically used for: Obtaining the actual contents of several elements of the training mineral samples as the label element contents of the training mineral sample set; obtaining the actual contents of the internal test mineral samples as the label element contents of the internal test mineral sample set; Preprocessing the training spectral dataset to obtain a preprocessed training spectral dataset; preprocessing the internal test spectral dataset to obtain a preprocessed internal test spectral dataset; Using the preprocessed training spectral dataset as an input variable, various machine learning models are trained under the supervision of the elemental content of the training mineral samples, and cross-validation and model iteration are performed during the training process until a training termination condition is reached, thereby obtaining multiple regression models; For each regression model, the performance of the regression model is tested respectively according to the label element content of the internal test mineral sample using the preprocessed internal test spectral data set. If the performance of the regression model does not meet the requirements, the regression model is trained again according to the adjusted training strategy until the performance of the regression model meets the requirements.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images