CN116337911A - Method and device for rapid determination of heavy metal content by portable XRF - Google Patents

Abstract

The invention provides a method and device for quickly measuring heavy metal content by portable XRF, which relates to the technical field of environmental monitoring, including: obtaining pXRF measurement data of various heavy metal elements in a sample to be determined; using a preset data dictionary to perform pXRF measurement data Processing to obtain the characteristic data corresponding to the pXRF measurement data; use the target machine learning model to process the characteristic data to obtain the specified heavy metal element content corresponding to the pXRF measurement data; wherein, the target machine learning model is based on multiple sets of pXRF measurement data and their corresponding A model determined after training on laboratory content measurements of specified heavy metal elements. Compared with methods that use common empirical formulas or linear regression to determine the content of heavy metals, this method has stronger sample adaptability and higher quantitative accuracy, and, because there is no need to wait for the laboratory to prepare a standard curve, it can also guarantee the accuracy of pXRF. quickness.

Description

Method and device for rapid determination of heavy metal content by portable XRF

technical field

The invention relates to the technical field of environmental monitoring, in particular to a method and device for quickly measuring heavy metal content by portable XRF.

Background technique

Portable X-ray fluorescence spectrometer (pXRF) can be used for quantitative analysis of heavy metal elements in soil. Currently, it is widely used in the field of environmental protection. It is superior to other analytical methods in terms of convenient sample preparation, non-destructive, and rapid, but its quantitative accuracy and sample Aspects such as the range of adaptation have been challenged.

Traditional portable X-ray fluorescence spectroscopic analysis uses empirical coefficient method or multiple regression method to establish a mathematical model for the fluorescence intensity and content of each element in the standard sample to obtain a calibration curve for quantitative analysis of the element content. The disadvantage of this method is that a calibration curve needs to be prepared first, which affects the speed of pXRF for elemental analysis to a certain extent. Moreover, even if a series of standard samples are used to establish a calibration curve, the accuracy of quantitative analysis of unknown samples is also affected by the consistency of the standard sample matrix, the content relationship between elements, and whether the sample preparation reaches the physical state of the standard sample.

Contents of the invention

The purpose of the present invention is to provide a method and device for the rapid determination of heavy metal content by portable XRF, so as to alleviate the poor quantitative analysis accuracy, weak sample adaptability and long analysis time in the prior art. question.

In the first aspect, the present invention provides a method for the rapid determination of heavy metal content by portable XRF, comprising: obtaining the pXRF measurement data of various heavy metal elements in the sample to be determined; using a preset data dictionary to process the pXRF measurement data to obtain the The feature data corresponding to the pXRF measurement data; wherein, the preset data dictionary includes: feature name, feature meaning and value type; the target machine learning model is used to process the feature data to obtain the corresponding pXRF measurement data The content of specified heavy metal elements; wherein, the target machine learning model is a model determined after training based on multiple sets of pXRF measurement data and corresponding laboratory content measurement values of specified heavy metal elements.

In an optional embodiment, the method further includes: acquiring multiple sets of sample measurement data; wherein, each set of sample measurement data includes: pXRF measurement data of multiple heavy metal elements and their corresponding designated heavy metal element laboratory Content measurement value; Utilize feature generation tool to process the pXRF measurement data in the multiple groups of sample measurement data, obtain several data features of the multiple groups of sample measurement data; calculate the feature importance score of each described data feature, Retaining target data features with feature importance scores greater than 0; constructing the preset data dictionary based on all target data features.

In an optional embodiment, the method further includes: using the preset data dictionary to process the pXRF measurement data in each set of sample measurement data to obtain characteristic data corresponding to each set of sample measurement data; The initial machine learning model is trained based on multiple sets of feature data corresponding to the sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements to obtain the target machine learning model.

In an optional embodiment, the initial machine learning model is trained based on multiple sets of characteristic data corresponding to the sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements, including: Multiple sets of sample measurement data are divided into a training set, a verification set and a test set; using the training set to train the initial machine learning model, and using the verification set to perform early stopping to obtain the current optimal model; judging the Whether the generalization performance of the current optimal model on the test set meets the preset conditions; if so, use the current optimal model as the target machine learning model; if not, adjust the preset model parameters, And retrain the initial machine learning model until the target machine learning model is obtained.

In an optional embodiment, using the training set to train the initial machine learning model, and using the verification set to perform early stopping, includes: repeatedly performing the following steps for multiple rounds until the heavy metal content of the target round is averaged The absolute error is less than the average absolute error of the heavy metal content of the previous round, and less than the average absolute error of the heavy metal content of the subsequent specified number of rounds; initialize the model parameters to obtain the first model parameters; based on the first model parameters and the training set The initial machine learning model is trained to obtain a first model; and the average absolute error of the heavy metal content of the first model on the verification set is calculated.

In an optional implementation manner, the target machine learning model includes one of the following: an extreme gradient boosting tree model, and a neural network model.

In an optional embodiment, any of the following methods is used to calculate the feature importance score of each of the data features: three-fold cross-validated LASSO regression algorithm, ridge regression algorithm, linear regression algorithm, tree model.

In the second aspect, the present invention provides a portable XRF device for quickly determining the content of heavy metals, including: a first acquisition module, used to acquire pXRF measurement data of various heavy metal elements in the sample to be determined; It is assumed that the data dictionary processes the pXRF measurement data to obtain the characteristic data corresponding to the pXRF measurement data; wherein, the preset data dictionary includes: characteristic name, characteristic meaning and value type; the second processing module is used for The target machine learning model is used to process the characteristic data to obtain the specified heavy metal element content corresponding to the pXRF measurement data; wherein, the target machine learning model is an experiment based on multiple sets of pXRF measurement data and their corresponding specified heavy metal elements A model determined after training on chamber content measurements.

In a third aspect, the present invention provides an electronic device, including a memory and a processor, the memory stores a computer program that can run on the processor, and the processor implements the aforementioned implementation mode when executing the computer program The step of the method for the portable XRF rapid determination heavy metal content described in any one.

In a fourth aspect, the present invention provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the portable XRF rapid Method for the determination of heavy metal content.

The invention provides a portable XRF method for quickly determining the content of heavy metals, using a target machine learning model trained based on multiple sets of pXRF measurement data and corresponding laboratory content measurement values of specified heavy metal elements to process the pXRF of various heavy metal elements Measure the characteristic data corresponding to the data to obtain the specified heavy metal element content corresponding to the pXRF measurement data of various heavy metal elements. Compared with methods that use common empirical formulas or linear regression to determine the content of heavy metals, this method has stronger sample adaptability and higher quantitative accuracy, and, because there is no need to wait for the laboratory to prepare a standard curve, it can also guarantee the accuracy of pXRF. quickness.

Description of drawings

In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings that need to be used in the specific implementation or description of the prior art. Obviously, the accompanying drawings in the following description The drawings show some implementations of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

Fig. 1 is the flow chart of a kind of method for portable XRF rapid determination heavy metal content provided by the embodiment of the present invention;

FIG. 2 is a flow chart of constructing a preset data dictionary provided by an embodiment of the present invention;

Fig. 3 is a functional block diagram of a portable XRF rapid determination of heavy metal content provided by an embodiment of the present invention;

Fig. 4 is a schematic diagram of an electronic device provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

Accordingly, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Some embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. In the case of no conflict, the following embodiments and features in the embodiments can be combined with each other.

Embodiment one

Fig. 1 is a flow chart of a method for the rapid determination of heavy metal content by portable XRF provided by the embodiment of the present invention. As shown in Fig. 1, the method specifically includes the following steps:

Step S102, acquiring pXRF measurement data of various heavy metal elements in the sample to be determined.

Specifically, for any heavy metal-contaminated site, the pXRF measurement values corresponding to various heavy metal elements in the sample to be determined can be measured by using a portable X-ray fluorescence spectrometer (pXRF). , lead, zinc, mercury and nickel. The implementation of the present invention does not specifically limit the quantity and element type of various heavy metal elements, and users can set them according to actual needs.

Step S104, using the preset data dictionary to process the pXRF measurement data to obtain characteristic data corresponding to the pXRF measurement data.

The embodiment of the present invention chooses to use the machine learning model to determine the content of heavy metals. Therefore, in order to improve the performance of the model, after obtaining the pXRF measurement data of various heavy metal elements in the sample to be determined, the embodiment of the present invention further performs feature engineering. The data dictionary is set to process the pXRF measurement data, wherein the preset data dictionary includes: feature name, feature meaning and value type. That is to say, feature data are extracted from the original pXRF measurement data, and the names, meanings and value types of these feature data are all pre-defined. For example, the characteristic data corresponding to the pXRF measurement data can be: the sum of the pXRF measurement value corresponding to the heavy metal element M and the pXRF measurement value corresponding to the heavy metal element N, or the pXRF measurement value corresponding to the heavy metal element M and the pXRF measurement value corresponding to the heavy metal element W The difference in measured values.

Step S106, using the target machine learning model to process the feature data to obtain the specified heavy metal element content corresponding to the pXRF measurement data.

Wherein, the target machine learning model is a model determined after training based on multiple sets of pXRF measurement data and corresponding laboratory content measurement values of specified heavy metal elements.

In the embodiment of the present invention, the input of the target machine learning model is the characteristic data corresponding to the pXRF measurement data, and the output of the model is the specified heavy metal element content corresponding to the pXRF measurement data. The designated heavy metal element means any one of the above-mentioned multiple heavy metal elements. In the embodiment of the present invention, the target machine learning model is a model obtained after training with multiple sets of sample measurement data, each set of sample measurement data is the pXRF measurement data of various heavy metal elements, and the specified heavy metal corresponding to the pXRF measurement data composed of laboratory content measurements of the elements. Therefore, the method provided by the embodiment of the present invention has stronger sample adaptability and higher quantitative precision than the traditional method.

The embodiment of the present invention provides a portable XRF method for quickly determining the content of heavy metals, using a target machine learning model trained based on multiple sets of pXRF measurement data and corresponding laboratory content measurement values of specified heavy metal elements to process various heavy metal elements The characteristic data corresponding to the pXRF measurement data in order to obtain the specified heavy metal element content corresponding to the pXRF measurement data of various heavy metal elements. Compared with methods that use common empirical formulas or linear regression to determine the content of heavy metals, this method has stronger sample adaptability and higher quantitative accuracy, and, because there is no need to wait for the laboratory to prepare a standard curve, it can also guarantee the accuracy of pXRF. quickness.

In an optional embodiment, as shown in Figure 2, the method of the present invention also includes the following steps:

Step S201, acquiring multiple sets of sample measurement data.

Wherein, each set of sample measurement data includes: pXRF measurement data of multiple heavy metal elements and their corresponding laboratory content measurement values of specified heavy metal elements.

Step S202, using a feature generation tool to process the pXRF measurement data in the multiple sets of sample measurement data to obtain several data features of the multiple sets of sample measurement data.

Step S203, calculating the feature importance score of each data feature, so as to retain the target data feature with feature importance score greater than 0.

Step S204, constructing a preset data dictionary based on all target data features.

In the embodiment of the present invention, in order to establish the preset data dictionary required by the target machine learning model, multiple sets of sample measurement data are first obtained, if the multiple metal elements include: arsenic, cadmium, chromium, copper, lead, zinc, mercury and Nickel, then the measurement data of each group of samples are shown in Table 1 below.

Table 1 sample measurement data

serial number name symbol 1 pXRF measurements of arsenic As 2 Cadmium pXRF measurements Cd 3 Chromium pXRF measurements Cr 4 Copper pXRF measurements Cu 5 pXRF measurements of lead Pb 6 Zinc pXRF measurements Zn 7 Mercury pXRF measurements Hg 8 pXRF measurements of nickel Ni 9 Specify laboratory content measurements for heavy metal X X_lab

Next, a feature generation tool is used to process the pXRF measurement data in multiple sets of sample measurement data, so as to quickly derive a large number of data features of the multiple sets of sample measurement data. The embodiment of the present invention does not specifically limit the feature generation tool, and the user can select according to actual needs, for example, use the python package FeatureTools, or realize the extraction of data features through manual programming.

Since there are many data features generated by the feature generation tool, and there are some unimportant data features, in order to reduce the unnecessary data calculation before model training, after obtaining some data features, further data feature screening is carried out. Specifically, the feature importance score of each data feature is calculated, and then the features whose feature importance is equal to 0 are eliminated, and only the target data features with feature importance scores greater than 0 are retained, and finally a preset data dictionary is constructed based on all target data features.

The embodiment of the present invention does not specifically limit the calculation method of the feature importance score. Users can use any of the following methods to calculate the feature importance score of each data feature: three-fold cross-validated LASSO regression algorithm, ridge regression algorithm, linear Regression algorithm, tree model. Among them, the LASSO regression algorithm is a balanced choice between accuracy and speed. A linear regression algorithm would be faster but not as accurate; a tree model would be more accurate but too slow.

The method of how to build the preset data dictionary is described in detail above, and how to obtain the target machine learning model is introduced below. In an optional embodiment, the method of the present invention also includes the following steps:

Step S301, using a preset data dictionary to process the pXRF measurement data in each set of sample measurement data to obtain characteristic data corresponding to each set of sample measurement data.

In step S302, the initial machine learning model is trained based on the feature data corresponding to multiple sets of sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements to obtain a target machine learning model.

In order to improve the quality of the machine learning results, the embodiment of the present invention does not directly use the pXRF measurement data of various heavy metal elements as model input during model training after obtaining multiple sets of sample measurement data, but uses the preset data dictionary The pXRF measurement data of multiple heavy metal elements in each group of sample measurement data are processed separately to obtain the characteristic data corresponding to each group of sample measurement data, and then the obtained characteristic data are used as model input. During the training process, observe the absolute error of the heavy metal content of the specified heavy metal element predicted by the model and the laboratory content measurement value corresponding to the pXRF measurement data of various heavy metal elements in the set of sample measurement data. The absolute error of multiple samples is averaged, which is the mean absolute error MAE. If the mean absolute error is zero, it means that the model predictions are in full agreement with the laboratory data, that is, the model is perfect. However, in the actual training process, the perfect model may not be obtained. Therefore, as long as the preset training end conditions are met, the target machine learning model can be considered to be obtained.

In an optional embodiment, the above step S302 is to train the initial machine learning model based on the characteristic data corresponding to multiple sets of sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements, specifically including the following steps:

Step S3021, divide multiple sets of sample measurement data into training set, verification set and test set according to preset proportions.

Step S3022, use the training set to train the initial machine learning model, and use the verification set to perform early stopping to obtain the current optimal model.

Step S3023, judging whether the generalization performance of the current optimal model on the test set meets the preset condition.

If yes, execute the following step S3024; if not, execute the following step S3025.

Step S3024, using the current optimal model as the target machine learning model;

Step S3025, adjusting the preset model parameters, and retraining the initial machine learning model until the target machine learning model is obtained.

Specifically, when training a machine learning model, only the training set and the verification set are generally divided, but this method does not guarantee the generalization ability of the machine learning model, and overfitting often occurs, resulting in the actual application of the model not as good as expected . In order to solve the above problems, the embodiment of the present invention divides multiple sets of sample measurement data into a training set, a verification set and a test set according to a preset ratio, for example, 80% of the training set, 10% of the verification set, and 10% of the test set. Users can adjust the above ratio according to actual needs.

After the sample is divided, use the training set to train the initial machine learning model, and use the verification set to stop early, that is, observe the mean absolute error on the verification set during training, as long as it is higher than the previous round of training. After the lower MAE (denoted as MAE _t ), if the MAE of the subsequent specified number of rounds is greater than MAE _t , then the model training stops, and the model with the best performance on the verification set is output, that is, the current optimal model . It is known that model training is carried out iteratively, each round of iteration will update the weights of the model, and each set of weights corresponds to a model. In the embodiment of the present invention, which weight is best determined according to the mean absolute error MAE . Therefore, the above current optimal model refers to the machine learning model corresponding to the set of weights with the lowest mean absolute error MAE.

After obtaining the current optimal model, the embodiment of the present invention further verifies the generalization performance of the model on the test set. If the generalization performance meets the preset conditions, use the current optimal model as the target machine learning model used to predict the content of the specified heavy metal element in actual application; if the generalization performance is not good, you need to adjust the preset model parameters and repeat the execution The above step S3022, until a model with better performance and better generalization ability is obtained.

The embodiment of the present invention measures the generalization performance of the model by observing the difference between the MAE on the verification set and the test set. If the MAE of the model on the test set rises too much compared with the MAE on the verification set, it means that the model is overfitting. Generalization performance is not good. Therefore, the above preset condition can be that the difference MAE _δ between MAE _valid (the MAE of the current optimal model on the validation set) and MAE _test (the MAE of the current optimal model on the test set) is less than the preset threshold, or MAE The ratio between _δ and MAE _valid is smaller than a preset ratio.

In an optional implementation, the above step S3022 uses the training set to train the initial machine learning model, and uses the verification set to perform early stopping, specifically including the following:

Repeat the following steps A-C for multiple rounds until the average absolute error of the heavy metal content of the target round is less than the average absolute error of the heavy metal content of the previous round, and smaller than the average absolute error of the heavy metal content of the subsequent specified number of rounds;

Step A, initializing the model parameters to obtain the first model parameters;

Step B, training the initial machine learning model based on the first model parameters and the training set to obtain the first model;

Step C, calculating the mean absolute error of the heavy metal content of the first model on the verification set.

Based on the above content, it can be known that each round of training needs to initialize the model parameters once, and then train the initial machine learning model according to the model parameters and the training set. When the training set ends, the first model can be obtained. The above training set The condition for the end of training may be that the number of training times reaches a specified number of times, or that the MAE reaches a specified condition. After obtaining the first model, the average absolute error of the heavy metal content of the first model on the verification set can be calculated, that is, the average of the absolute errors of all samples in the verification set.

By analogy, calculate the average absolute error of the heavy metal content of the first model obtained in each round of training on the verification set, assuming that the average absolute error of the heavy metal content of the first model in the 100th round on the verification set MAE ₁₀₀ is less than that of the 99th round If the average absolute error of the heavy metal content of a model on the verification set is MAE ₉₉ , then the early-stop observation mechanism can be started to determine whether the average absolute error of the heavy metal content of the specified number of rounds (for example, 100 rounds) after the 100th round is greater than MAE ₁₀₀ , and also That is, MAE ₁₀₁ to MAE ₂₀₀ are all greater than MAE _100. If so, then the 100th round is the target round described above, and the first model corresponding to the 100th round is the above-mentioned current optimal model.

If after starting the early stop observation mechanism, before reaching the specified number of rounds, for example, MAE ₁₂₀ < MAE ₁₀₀ occurs in the 120th round, skip the current early stop observation and start the early stop observation mechanism again at the 120th round .

In an optional embodiment, the target machine learning model includes one of the following: an extreme gradient boosting tree model, and a neural network model.

Taking the target machine learning model as an extreme gradient boosting tree model (XGBoost model) as an example, the parameters used by XGBoost are as follows: XGBRegressor(max_depth=2; learning_rate=0.25; gamma=0.0; min_child_weight=0.0; max_delta_step=0.0; subsample=0.9; colsample_bytree=0.9; colsample_bylevel=1.0; reg_alpha=0.0; reg_lambda=1.0; n_estimators=1000; use_label_encoder=False; nthread=4; scale_pos_weight=1.0; base_score=0.5; seed=1337; random_state=1337).

If the model is overfitted, the adjustment direction of the model parameters can be: reduce the depth of the tree, the number of children, adjust the learning rate, increase regularization, etc., corresponding to learning_rate, n_estimators, min_child_weight and other parameters.

In summary, the portable XRF method for quickly determining the content of heavy metals provided by the embodiment of the present invention uses a large amount of sample measurement data when training the machine learning model. Therefore, the method of the present invention predicts heavy metal elements based on the measured pXRF value It has stronger sample adaptability and higher quantitative accuracy in terms of content. In addition, because this method does not need to wait for the laboratory to prepare a standard curve, it fully maintains the convenience of pXRF.

Embodiment two

The embodiment of the present invention also provides a portable XRF rapid determination of heavy metal content device, the portable XRF rapid determination of heavy metal content device is mainly used to implement the portable XRF rapid determination of heavy metal content method provided in the first embodiment, the following description of the present invention The portable XRF device for rapid determination of heavy metal content provided in the examples is described in detail.

Fig. 3 is a functional module diagram of a portable XRF device for quickly measuring heavy metal content provided by an embodiment of the present invention. As shown in Fig. 3, the device mainly includes: a first acquisition module 10, a first processing module 20, a second processing module Module 30, in which:

The first acquisition module 10 is configured to acquire pXRF measurement data of various heavy metal elements in the sample to be determined.

The first processing module 20 is configured to process the pXRF measurement data by using a preset data dictionary to obtain feature data corresponding to the pXRF measurement data; wherein, the preset data dictionary includes: feature name, feature meaning and value type.

The second processing module 30 is used to process the characteristic data by using the target machine learning model to obtain the specified heavy metal element content corresponding to the pXRF measurement data; wherein, the target machine learning model is based on multiple sets of pXRF measurement data and their corresponding specified heavy metal elements The model was determined after training on laboratory content measurements.

The method performed by the portable XRF rapid determination of heavy metal content provided by the embodiment of the present invention uses the target machine learning model trained based on multiple sets of pXRF measurement data and corresponding laboratory content measurement values of specified heavy metal elements to process multiple The characteristic data corresponding to the pXRF measurement data of various heavy metal elements, in order to obtain the specified heavy metal element content corresponding to the pXRF measurement data of various heavy metal elements. Compared with methods using common empirical formulas or linear regression to determine the content of heavy metals, this device has stronger sample adaptability and higher quantitative accuracy, and, because there is no need to wait for the laboratory to prepare a standard curve, it can also guarantee the accuracy of pXRF. quickness.

Optionally, the device also includes:

The second acquisition module is used to acquire multiple sets of sample measurement data; wherein, each set of sample measurement data includes: pXRF measurement data of various heavy metal elements and corresponding laboratory content measurement values of specified heavy metal elements.

The third processing module is used to process the pXRF measurement data in the multiple sets of sample measurement data by using the feature generation tool to obtain several data features of the multiple sets of sample measurement data.

The calculation and retention module is used to calculate the feature importance score of each data feature, so as to retain target data features with feature importance scores greater than 0.

A building block for building a preset data dictionary based on all target data characteristics.

Optionally, the device also includes:

The fourth processing module is configured to use the preset data dictionary to process the pXRF measurement data in each set of sample measurement data to obtain characteristic data corresponding to each set of sample measurement data.

The training module is used to train the initial machine learning model based on the feature data corresponding to multiple sets of sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements to obtain the target machine learning model.

Optionally, the training modules include:

The division unit is used for dividing multiple sets of sample measurement data into training set, verification set and test set according to preset proportions.

The training unit is used to use the training set to train the initial machine learning model, and use the verification set to perform early stopping to obtain the current optimal model.

The judging unit is used to judge whether the generalization performance of the current optimal model on the test set meets the preset condition.

The determining unit is configured to use the current optimal model as the target machine learning model when it is determined to be consistent.

The adjustment and training unit is used to adjust the preset model parameters and retrain the initial machine learning model until the target machine learning model is obtained when it is determined that the inconsistency is determined.

Optionally, the training unit is specifically used for:

Repeat the following steps for several rounds until the average absolute error of the heavy metal content of the target round is less than the average absolute error of the heavy metal content of the previous round, and smaller than the average absolute error of the heavy metal content of the subsequent specified rounds.

Initialize the model parameters to obtain the first model parameters.

The initial machine learning model is trained based on the first model parameters and the training set to obtain the first model.

Calculate the mean absolute error of the heavy metal content of the first model on the validation set.

Optionally, the target machine learning model includes one of the following: extreme gradient boosting tree model, neural network model.

Optionally, use any of the following methods to calculate the feature importance score of each data feature: three-fold cross-validated LASSO regression algorithm, ridge regression algorithm, linear regression algorithm, tree model.

Embodiment three

Referring to Fig. 4, the embodiment of the present invention provides a kind of electronic equipment, and this electronic equipment comprises: processor 60, memory 61, bus 62 and communication interface 63, described processor 60, communication interface 63 and memory 61 are connected by bus 62 ; The processor 60 is used to execute executable modules stored in the memory 61, such as computer programs.

Wherein, the memory 61 may include a high-speed random access memory (RAM, Random Access Memory), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the Internet, wide area network, local network, metropolitan area network, etc. can be used.

The bus 62 can be an ISA bus, a PCI bus or an EISA bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one double-headed arrow is used in FIG. 4 , but it does not mean that there is only one bus or one type of bus.

Wherein, the memory 61 is used to store a program, and the processor 60 executes the program after receiving an execution instruction, and the method performed by the process-defining device disclosed in any of the above-mentioned embodiments of the present invention can be applied to the processor 60, or implemented by the processor 60.

The processor 60 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 60 or an instruction in the form of software. The above-mentioned processor 60 can be a general-purpose processor, including a central processing unit (Central Processing Unit, referred to as CPU), a network processor (Network Processor, referred to as NP), etc.; it can also be a digital signal processor (Digital Signal Processing, referred to as DSP) , Application Specific Integrated Circuit (ASIC for short), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Various methods, steps and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. The steps of the methods disclosed in the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61, and completes the steps of the above method in combination with its hardware.

The computer program product of a portable XRF rapid determination of heavy metal content method and device provided by an embodiment of the present invention includes a computer-readable storage medium storing a non-volatile program code executable by a processor, and the program code includes The instructions can be used to execute the methods described in the foregoing method embodiments. For specific implementation, refer to the method embodiments, and details are not repeated here.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.

If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes. .

It should be noted that like numerals and letters denote similar items in the following figures, therefore, once an item is defined in one figure, it does not require further definition and explanation in subsequent figures.

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" etc. The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship that is usually placed when the product of the invention is used, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying References to devices or elements must have a particular orientation, be constructed, and operate in a particular orientation and therefore should not be construed as limiting the invention. In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions, and should not be construed as indicating or implying relative importance.

In addition, the terms "horizontal", "vertical", "overhanging" and the like do not mean that the components are absolutely horizontal or overhanging, but may be slightly inclined. For example, "horizontal" only means that its direction is more horizontal than "vertical", and it does not mean that the structure must be completely horizontal, but can be slightly inclined.

In the description of the present invention, it should also be noted that, unless otherwise clearly specified and limited, the terms "installation", "installation", "connection" and "connection" should be understood in a broad sense, for example, it may be a fixed connection, It can also be a detachable connection or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediary, and it can be the internal communication of two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (10)

1. A method for the rapid determination of heavy metal content by portable XRF, characterized in that, comprising: Obtain the pXRF measurement data of various heavy metal elements in the sample to be determined; Using a preset data dictionary to process the pXRF measurement data to obtain feature data corresponding to the pXRF measurement data; wherein, the preset data dictionary includes: feature name, feature meaning and value type; The target machine learning model is used to process the characteristic data to obtain the specified heavy metal element content corresponding to the pXRF measurement data; wherein, the target machine learning model is an experiment based on multiple sets of pXRF measurement data and their corresponding specified heavy metal elements A model determined after training on chamber content measurements. 2. the method for portable XRF rapid determination heavy metal content according to claim 1, is characterized in that, described method also comprises: Obtain multiple sets of sample measurement data; wherein, each set of sample measurement data includes: pXRF measurement data of multiple heavy metal elements and their corresponding laboratory content measurement values of specified heavy metal elements; Using a feature generation tool to process the pXRF measurement data in the multiple sets of sample measurement data to obtain several data features of the multiple sets of sample measurement data; Calculating the feature importance score of each of the data features to retain target data features with feature importance scores greater than 0; The preset data dictionary is constructed based on all the target data features. 3. the method for portable XRF rapid determination heavy metal content according to claim 2, is characterized in that, described method also comprises: Using the preset data dictionary to process the pXRF measurement data in each set of sample measurement data to obtain characteristic data corresponding to each set of sample measurement data; The initial machine learning model is trained based on multiple sets of feature data corresponding to the sample measurement data and the corresponding laboratory content measurement values of specified heavy metal elements to obtain the target machine learning model. 4. The method for portable XRF rapid determination of heavy metal content according to claim 3, characterized in that, based on the characteristic data corresponding to the sample measurement data of multiple groups and the corresponding laboratory content measurement values of the specified heavy metal elements, the initial machine The learning model is trained, including: Divide the plurality of sets of sample measurement data into a training set, a verification set and a test set according to a preset ratio; Using the training set to train the initial machine learning model, and using the verification set to perform early stopping to obtain the current optimal model; Judging whether the generalization performance of the current optimal model on the test set meets a preset condition; If so, using the current optimal model as the target machine learning model; If not, adjust the preset model parameters, and retrain the initial machine learning model until the target machine learning model is obtained. 5. the method for portable XRF fast determination heavy metal content according to claim 4, is characterized in that, utilizes described training set to train described initial machine learning model, and utilizes described validation set to carry out early stop, comprising: Repeat the following steps for multiple rounds until the average absolute error of the heavy metal content of the target round is less than the average absolute error of the heavy metal content of the previous round, and smaller than the average absolute error of the heavy metal content of the subsequent specified number of rounds; Initialize the model parameters to obtain the first model parameters; Train the initial machine learning model based on the first model parameters and the training set to obtain a first model; calculating the mean absolute error of the heavy metal content of the first model on the validation set. 6. The method for rapid determination of heavy metal content by portable XRF according to claim 1, wherein the target machine learning model comprises one of the following: extreme gradient boosting tree model, neural network model. 7. The method for the rapid determination of heavy metal content by portable XRF according to claim 2, characterized in that, adopt any of the following methods to calculate the feature importance score of each described data feature: the LASSO regression algorithm of three-fold cross-validation , ridge regression algorithm, linear regression algorithm, tree model. 8. A device for the rapid determination of heavy metal content by portable XRF, characterized in that it comprises: The first acquisition module is used to acquire the pXRF measurement data of various heavy metal elements in the sample to be determined; The first processing module is configured to use a preset data dictionary to process the pXRF measurement data to obtain characteristic data corresponding to the pXRF measurement data; wherein, the preset data dictionary includes: characteristic names, characteristic meanings and values type; The second processing module is used to process the characteristic data by using the target machine learning model to obtain the specified heavy metal element content corresponding to the pXRF measurement data; wherein, the target machine learning model is based on multiple sets of pXRF measurement data and its A model identified after training on laboratory content measurements corresponding to the specified heavy metal elements. 9. An electronic device, comprising a memory and a processor, the memory is stored with a computer program that can run on the processor, and it is characterized in that the above-mentioned claim 1 is realized when the processor executes the computer program Steps in the method for the portable XRF rapid determination of heavy metal content described in any one of to 7. 10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the portable computer according to any one of claims 1 to 7 is realized. XRF method for rapid determination of heavy metal content.

Images