A method and system for predicting rare earth element component content

Abstract

The invention proposes a method and system for predicting the content of rare earth element components, belonging to the technical field of rare earth element analysis, and comprising the following steps: constructing a historical database; the historical database includes a plurality of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents; obtaining target rare earth solution image color features; selecting from the historical database a plurality of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color features conforms to a preset rule as a learning set; using a least squares support vector machine to establish an element component content prediction model, using the historical rare earth solution image color features in the learning set as the input of the model, and using the corresponding historical rare earth element component contents as the target output of the model; inputting the target rare earth solution image color features into the element component content prediction model to obtain the target rare earth element component content.

Description

Technical Field

[0001] The invention relates to the technical field of rare earth element analysis, and in particular to a method and system for predicting the content of rare earth element components. Background Art

[0002] Rare earths have unique physical and chemical properties, so they are used in all aspects of modern industry and will play an increasingly important role in the future industrial system. They also play an irreplaceable role in high-tech such as new energy, national defense, and emerging industries, and are a national strategic resource. In the rare earth extraction and separation process, real-time detection and understanding of the distribution of rare earth element content is the key to implementing automatic control of rare earth production lines. At present, the prediction of rare earth component content all uses offline detection models. Due to the unstable working conditions at the extraction site and many variable factors, the prediction accuracy of the offline detection model will decrease. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for predicting the content of rare earth element components, thereby improving the accuracy of predicting the content of rare earth element components.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A method for predicting the content of rare earth elements, comprising the following steps:

[0006] Constructing a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents;

[0007] Acquire the color characteristics of the target rare earth solution image;

[0008] Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color feature meets a preset rule as a learning set;

[0009] A least squares support vector machine is used to establish an element component content prediction model, the color features of the historical rare earth solution images in the learning set are used as the input of the model, and the corresponding historical rare earth element component content is used as the target output of the model;

[0010] The target rare earth solution image color feature is input into the element component content prediction model to obtain the target rare earth element component content.

[0011] Optionally, the method further comprises:

[0012] Calculating the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and selecting the maximum value among each root mean square as an update criterion;

[0013] The size of the update criterion and the update threshold is determined. If the update criterion is greater than the update threshold, the target rare earth solution image color feature and the target rare earth element component content are added to the historical database; otherwise, the historical database is kept unchanged.

[0014] Optionally, a plurality of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color feature meets a preset rule are selected from the historical database as a learning set, specifically including:

[0015] Calculating the grey correlation between the target rare earth solution image color feature and each historical rare earth solution image color feature, and determining the similar rare earth solution image color feature with the strongest correlation;

[0016] A plurality of historical rare earth solution image color features whose gray correlation degree with the similar rare earth solution image color features is greater than a preset comparison threshold are selected from the historical database as a learning set.

[0017] Optionally, the grey correlation degree between the color features of the two rare earth solution images is calculated according to the following formula:

[0018]

[0019] Where n represents the dimension of the color feature of the rare earth solution image, r ij,m Represents the color feature x of the rare earth solution image i and rare earth solution image color feature x j The grey relational coefficient of the mth dimension is calculated according to the following formula:

[0020]

[0021] Among them, x i,m Represents the color feature x of the rare earth solution image i The value of the mth dimension, x j,m Represents the color feature x of the rare earth solution image j The value of the mth dimension, ρ is the resolution coefficient.

[0022] Optionally, before selecting from the historical database a plurality of historical rare earth solution image color features having a gray correlation close to the target rare earth solution image color feature as a learning set, the method further comprises:

[0023] Selecting the historical rare earth solution image color features in the historical database as iterative rare earth solution image color features, and the corresponding historical rare earth element component content as iterative rare earth element component content;

[0024] Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the iterative rare earth solution image color feature meets a preset rule as a learning set;

[0025] A least squares support vector machine is used to establish an element component content prediction model, the color features of the historical rare earth solution images in the learning set are used as the input of the model, and the corresponding historical rare earth element component content is used as the target output of the model;

[0026] Inputting the iterative rare earth solution image color feature into the element component content prediction model to obtain an iterative rare earth element component content prediction value;

[0027] The prediction error is obtained by comparing the iterative rare earth element component content prediction value with the iterative rare earth element component content.

[0028] Optionally, the method further comprises: using a genetic algorithm to iteratively optimize the resolution coefficient and the preset comparison threshold to minimize the prediction error.

[0029] Optionally, the iterative optimization of the resolution coefficient and the preset comparison threshold by using a genetic algorithm specifically includes:

[0030] Establish a parent population, each individual in the population has a corresponding fitness value;

[0031] A roulette wheel method is used to select a number of individuals whose fitness values ​​are greater than a set threshold from the parent population to obtain a screening population;

[0032] In the screening population, genetic operations are performed on the population according to the initial crossover probability and mutation probability to obtain an offspring population;

[0033] The offspring population is used as a new parent population, and the steps of "using the roulette method to select a number of individuals with fitness values ​​greater than a set threshold from the parent population to obtain a screening population; in the screening population, performing genetic operations on the population according to the initial crossover probability and mutation probability to obtain an offspring population" are repeated until the preset number of inheritance times is reached.

[0034] Optionally, the method further comprises:

[0035] Determine the change of a single optimal fitness value; the single optimal fitness value is the maximum value of each fitness value in a single genetic operation;

[0036] If the number of times the single optimal fitness value remains unchanged in a number of consecutive genetic operations is greater than a stagnation threshold, modifying the crossover probability and mutation probability of the genetic algorithm; until the single optimal fitness value changes, restoring the number of times, the crossover probability and the mutation probability to initial values;

[0037] If the single optimal fitness value changes, the constant number is reset, and the crossover probability and the mutation probability are kept unchanged.

[0038] Corresponding to the aforementioned rare earth element component content prediction method, the present invention also provides a rare earth element component content prediction system, comprising:

[0039] A historical database construction module is used to construct a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents;

[0040] A grey correlation calculation module is used to calculate the grey correlation between any two rare earth solution image color features;

[0041] A target input acquisition module is used to acquire the color characteristics of the target rare earth solution image;

[0042] A learning set selection module, used to select from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color features meets a preset rule as a learning set;

[0043] A prediction model establishment module, used to establish an element component content prediction model using a least squares support vector machine, using the color features of the historical rare earth solution images in the learning set as the input of the model, and using the corresponding historical rare earth element component content as the target output of the model;

[0044] The target output acquisition module is used to input the color characteristics of the target rare earth solution image into the element component content prediction model to obtain the target rare earth element component content.

[0045] Optionally, the system further comprises:

[0046] An update criterion determination module is used to calculate the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and select the maximum value of each root mean square as the update criterion;

[0047] A comparison module, used for comparing the update criterion and the update threshold; if the update criterion is greater than the update threshold, calling the database update module to update the historical database; otherwise, keeping the historical database unchanged;

[0048] The database updating module is used to add the target rare earth solution image color characteristics and the target rare earth element component content to the historical database.

[0049] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0050] The present invention provides a method and system for predicting the content of rare earth element components, comprising: constructing a historical database; the historical database includes a plurality of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents; obtaining a target rare earth solution image color feature; selecting from the historical database a plurality of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color feature meets a preset rule as a learning set; using a least squares support vector machine to establish an element component content prediction model, using the historical rare earth solution image color features in the learning set as the input of the model, and using the corresponding historical rare earth element component content as the target output of the model; inputting the target rare earth solution image color features into the element component content prediction model to obtain the target rare earth element component content. The present invention combines a grey correlation analysis method with an instant learning algorithm, selects a number of historical rare earth solution image color features with strong grey correlation with target rare earth solution image color features from a historical database as a learning set, and establishes an element component content prediction model based on these rare earth solution image color features in the learning set, thereby realizing dynamic establishment of a prediction model, and selecting some rare earth solution image color features with strong similarity in the historical database to train the model. Compared with the prior art, which uses rare earth solution image color features of all working conditions to establish a model and predict the element component content, the accuracy of rare earth element component content prediction under real-time working conditions can be improved, and the problem of decreased offline model prediction accuracy caused by variable on-site working conditions is solved. In addition, the grey correlation analysis method is superior to the traditional k-nearest neighbor method, and the model prediction accuracy is also improved to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 A flowchart of a method for predicting the content of rare earth elements provided in Example 1 of the present invention;

[0053] Figure 2 This is a flow chart of steps S6 to S8 in the method provided in Example 1 of the present invention;

[0054] Figure 3 A specific flow chart of step S3 in the method provided in Example 1 of the present invention;

[0055] Figure 4This is a flow chart of steps A1 to A6 in the method provided in Example 1 of the present invention;

[0056] Figure 5 A specific flow chart of step A6 in the method provided in Example 1 of the present invention;

[0057] Figure 6 This is a flow chart of steps A7 to A9 in the method provided in Example 1 of the present invention;

[0058] Figure 7 A diagram showing the prediction results after optimizing the parameters of the instant learning algorithm in the method provided in Example 1 of the present invention;

[0059] Figure 8 A schematic diagram of the structure of a rare earth element component content prediction system provided in Example 2 of the present invention.

[0060] Symbol explanation: 1: Historical database construction module; 2: Grey correlation calculation module; 3: Target input acquisition module; 4: Learning set selection module; 5: Prediction model establishment module; 6: Target output acquisition module; 7: Update criterion determination module; 8: Comparison module; 9: Database update module. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] The purpose of the present invention is to provide a method and system for predicting the content of rare earth element components, thereby improving the accuracy of predicting the content of rare earth element components.

[0063] In the prior art, due to the unstable working conditions of the extraction site and many variable factors, the prediction accuracy of the offline detection model will decrease. In order to overcome the shortcomings of the current problem, the present invention proposes a method for real-time prediction of the content of rare earth element components using an improved grey correlation analysis method-instant learning algorithm model, which improves the prediction accuracy of the content of rare earth element components.

[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] Embodiment 1:

[0066] This embodiment provides a method for predicting the content of rare earth elements. Figure 1 As shown, the method comprises the following steps:

[0067] S1. Constructing a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and the corresponding historical rare earth element component contents;

[0068] S2, obtaining the color characteristics of the target rare earth solution image;

[0069] S3, selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color features meets a preset rule as a learning set;

[0070] S4, using a least squares support vector machine to establish an element component content prediction model, using the color features of the historical rare earth solution images in the learning set as the input of the model, and using the corresponding historical rare earth element component content as the target output of the model;

[0071] S5. Inputting the color features of the target rare earth solution image into the element component content prediction model to obtain the target rare earth element component content.

[0072] In order to maintain the diversity of the database, such as Figure 2 As shown, the method also includes:

[0073] S6, calculating the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and selecting the maximum value among the root mean squares as the update criterion;

[0074] S7, judging the size of the update criterion and the update threshold, if the update criterion is greater than the update threshold, jumping to step S8, otherwise, keeping the historical database unchanged;

[0075] S8. Add the target rare earth solution image color features and the target rare earth element component content to the historical database.

[0076] In this embodiment, if Figure 3 As shown, the step S3 specifically includes:

[0077] S31, calculating the grey correlation between the target rare earth solution image color feature and each historical rare earth solution image color feature, and determining the similar rare earth solution image color feature with the strongest correlation;

[0078] S32, selecting from the historical database several historical rare earth solution image color features whose gray correlation degree with the similar rare earth solution image color features is greater than a preset comparison threshold as a learning set.

[0079] In this embodiment, the gray correlation degree between the color features of two rare earth solution images is calculated according to the following formula:

[0080]

[0081] Where n represents the dimension of the color feature of the rare earth solution image, r ij,m Represents the color feature x of the rare earth solution image i and rare earth solution image color feature x j Grey relational coefficient of mth dimension, grey relational coefficient r ij,m Calculated according to the following formula:

[0082]

[0083] Among them, x i,m Represents the color feature x of the rare earth solution image i The value of the mth dimension, x j,m Represents the color feature x of the rare earth solution image j The value of the mth dimension, ρ is the resolution coefficient.

[0084] In order to make the instant learning algorithm more accurate, the prediction error between the prediction result and the target output is obtained, such as Figure 4 As shown, before step S3, the method further includes:

[0085] A1, sequentially selecting the color features of the historical rare earth solution images in the historical database as the iterative rare earth solution image color features, and the corresponding historical rare earth element component contents as the iterative rare earth element component contents;

[0086] A2, selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the iterative rare earth solution image color features meets a preset rule as a learning set;

[0087] A3. Using a least squares support vector machine to establish an element component content prediction model, using the color features of the historical rare earth solution images in the learning set as the input of the model, and using the corresponding historical rare earth element component content as the target output of the model;

[0088] A4, inputting the iterative rare earth solution image color feature into the element component content prediction model to obtain an iterative rare earth element component content prediction value;

[0089] A5. Compare the iterative rare earth element component content prediction value with the iterative rare earth element component content to obtain a prediction error.

[0090] In order to further improve the prediction accuracy of the model, the present invention uses a genetic algorithm with a stagnation backtracking strategy to optimize the parameters of the instant learning algorithm. The method also includes: A6, using a genetic algorithm to iteratively optimize the resolution coefficient and the preset comparison threshold to minimize the prediction error; the specific steps are as follows Figure 5 As shown, step A6 specifically includes:

[0091] A61, establish a parent population, bring each individual in the population into steps A1-A5, obtain multiple prediction errors, and select the reciprocal of the best prediction error as the fitness value of the corresponding individual;

[0092] A62, selecting a number of individuals whose fitness values ​​are greater than a set threshold from the parent population by using a roulette method to obtain a screening population;

[0093] A63. In the screening population, genetic operations are performed on the population according to the initial crossover probability and mutation probability to obtain an offspring population;

[0094] A64. Use the offspring population as a new parent population and repeat steps A61 to A63 until a preset number of inheritance times is reached.

[0095] To prevent the genetic algorithm from getting stagnant, Figure 6 As shown, the method of the present invention also includes:

[0096] A7, judging the change of the single optimal fitness value, if the number of times the single optimal fitness value remains unchanged in several consecutive genetic operations is greater than the stagnation threshold, calling step A8, otherwise jumping to step A9; the single optimal fitness value is the maximum value of each fitness value in the single genetic operation;

[0097] A8, modifying the crossover probability and mutation probability of the genetic algorithm; until the single optimal fitness value changes, restoring the unchanged number of times, the crossover probability and the mutation probability to the initial values;

[0098] A9. Reset the unchanged number of times and keep the crossover probability and the mutation probability unchanged.

[0099] The following is a specific example to illustrate the method for predicting the content of rare earth elements provided by the present invention:

[0100] (1) Learning set selection for just-in-time learning algorithm

[0101] The real image data of Pr / Nd rare earth solution of a rare earth company was selected for simulation experiment. 85 sample solutions with different component contents were collected, and the Nd element component content was distributed in the range of 18%-99%. The solution images at the time of collection were recorded, and the collected solution samples were analyzed in color space using machine vision technology to obtain the color features of R, G, B, H, S, I, RR, RG, RB and CVA historical rare earth solution images; the standard component content value corresponding to the solution was obtained using an offline assay method as the historical rare earth element component content, and the above historical rare earth solution image color features and the corresponding historical rare earth element component content were used as the real-time learning algorithm modeling data. Among them, 70 groups of data were used as historical databases and 15 groups of data were used as test databases.

[0102] First, the grey correlation analysis is used to determine the grey correlation degree between the variables in the historical database. The grey correlation analysis formula is as follows:

[0103]

[0104] in

[0105]

[0106]

[0107] n is the total number of input variables and output variables in a single data, x i,m , x j,m is the value of the mth dimension of the i-th and j-th data in the database, ρ is the resolution coefficient, taking values ​​of 0,1), here ρ=0.5, r ij,m For data x i With x j The grey correlation coefficient of the mth dimension, r ij For data x i With x j The grey relational value between ij The larger it is, the stronger the correlation between the data.

[0108] According to the grey correlation values ​​between each data in the historical database, a grey correlation matrix is ​​established to form a hash table as follows:

[0109]

[0110] Where N represents the data size of the historical database, r ab It represents the grey correlation value between the a-th data and the b-th data in the historical database.

[0111] Determine the new input condition data x0 and the database data {x1, x2, …, x N The maximum grey relational value r in0t The corresponding data x t , according to experience, the hash table threshold k is set manually δ = 0.7, the tth row in the hash table is greater than k δ The data corresponding to the grey relational value is taken as the learning set data.

[0112] (2) Model building of the just-in-time learning algorithm

[0113] After the learning set required for model building has been determined, the least squares support vector machine LSSVM is used to build the prediction model. The advantages of LSSVM are fast learning speed and relatively simple calculation. It is very suitable for modeling small samples. The objective function of LSSVM is shown as follows:

[0114]

[0115] where y i represents the constraints of the objective function, e i is the error, γ is the error penalty coefficient, ω represents the weight vector, b is the bias, It can map data from low-dimensional space to high-dimensional space.

[0116] Since the objective function is a constrained problem, the Lagrangian function is established to transform the constrained problem into an unconstrained problem, so as to make it easier to solve the problem, namely:

[0117]

[0118] Finally, the output function of the prediction result of the LSSVM model can be obtained as:

[0119]

[0120] Where x0 is the input of the prediction data, i.e. the input of the new working condition data, and x i is the input of the training sample, i.e., the selected learning set input, b is the deviation value, α i is the Lagrangian operator, K(·) is the kernel function, and the Gaussian kernel function is selected as the kernel function of the model, as shown in the following formula:

[0121]

[0122] In the formula, σ is the width of the Gaussian kernel function. The leave-one-out method is used to optimize the parameters (γ, σ) in the LSSVM model.

[0123] (3) Database update criteria

[0124] Due to the change of on-site working conditions, the historical database cannot completely contain the data of the current working conditions. Therefore, it is necessary to adopt the database update criteria to determine whether to put the current new input data into the database after predicting the component content of the new data. i The corresponding maximum RMS value of the Euclidean distance is used as the update criterion x error .

[0125]

[0126] Where k is the number of data samples, n is the number of dimensions of the data, x0 is the input data, x i is the database data. The database update threshold x is determined by experience δ , x δ The value range is (0,1), when 1≥x error >x δ When x δ >x error When ≥0, it means that the input operating condition data is generally similar to the database data, so the database remains unchanged.

[0127] (4) Genetic Algorithm Optimization of Stagnation Backtracking Strategy

[0128] Genetic algorithm is inspired by genetic evolution theory and genetic evolution mechanism. It is a genetic method for optimizing data search. First, the data is binary encoded as chromosomes, and then selected and evolved according to the fitness value. The chromosomes are operated with the help of gene selection, crossover and gene mutation in genetics, and some chromosomes with low fitness values ​​are eliminated. Individuals with high fitness values ​​are selected to be inherited to the next generation to form new chromosomes, and iterate again and again until a satisfactory result is finally obtained. The stagnation backtracking strategy is to change the crossover and mutation probability in the genetic algorithm when the number of iterations of the fitness value in the genetic algorithm reaches a certain number, so as to achieve the purpose of jumping out of the current solution and avoid falling into the local optimum.

[0129] In this example, the initial population is first established. Each individual randomly generates 3 chromosomes, corresponding to the 3 parameters that need to be optimized. Then, the parameters of each individual are substituted into the instant learning algorithm to obtain the prediction error solution corresponding to each individual. The value of the solution is used as the fitness value f of the population. i ;

[0130] The roulette method is then used to select the population, and the selection probability of each individual is The purpose of the roulette method is to eliminate individuals with poor fitness values, thereby achieving the purpose of optimizing the parameters in the population;

[0131] Then, a chromosome crossover with a probability of 70% and a chromosome mutation with a probability of 1% are performed in the population, and then the population is modeled and predicted using the real-time learning algorithm.

[0132] When the optimal solution in the population remains unchanged for 4 consecutive times, the corresponding population is saved; when the optimal solution remains unchanged for 12 consecutive times, it means that the population evolution has stagnated, so the chromosome crossover probability is increased to 90%, and the chromosome mutation probability is increased to 10%, increasing the diversity of the population and the range of solutions; when the optimal solution of the population changes, the chromosome crossover probability and mutation probability return to the initial value. The optimal parameters obtained after optimization are substituted into the improved GRA-instant learning algorithm for experimental verification. The experimental prediction results are as follows Figure 7 shown.

[0133] Embodiment 2:

[0134] like Figure 8 The structural schematic diagram shown corresponds to a rare earth element component content prediction method provided in Example 1. This embodiment provides a rare earth element component content prediction system, including:

[0135] A historical database construction module 1 is used to construct a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and the corresponding historical rare earth element component contents;

[0136] A grey correlation calculation module 2 is used to calculate the grey correlation between any two rare earth solution image color features;

[0137] The target input acquisition module 3 is used to obtain the color characteristics of the target rare earth solution image;

[0138] A learning set selection module 4 is used to select from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color features meets a preset rule as a learning set;

[0139] A prediction model building module 5 is used to establish an element component content prediction model using a least squares support vector machine, using the color features of the historical rare earth solution images in the learning set as the input of the model and using the corresponding historical rare earth element component content as the target output of the model;

[0140] The target output acquisition module 6 is used to input the color characteristics of the target rare earth solution image into the element component content prediction model to obtain the target rare earth element component content.

[0141] In order to maintain the diversity of the database, the system further includes:

[0142] An update criterion determination module 7 is used to calculate the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and select the maximum value of each root mean square as the update criterion;

[0143] A comparison module 8 is used to compare the update criterion and the update threshold; if the update criterion is greater than the update threshold, the database update module is called to update the historical database; otherwise, the historical database is kept unchanged;

[0144] The database updating module 9 is used to add the target rare earth solution image color characteristics and the target rare earth element component content to the historical database.

[0145] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0146] All or part of the software may sometimes communicate over a network, such as the Internet or other communications network. Such communications can load software from one computer device or processor to another. For example: from a server or host computer of a video target detection device to a hardware platform of a computer environment, or other computer environment that implements the system, or a system with similar functions related to providing information required for target detection. Therefore, another medium capable of transmitting software elements can also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., transmitted through cables, optical cables or air. Physical media used to carry carriers, such as cables, wireless connections or optical cables and the like, can also be considered as media that carry software. As used herein, unless limited to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media involved in the process of the processor executing any instructions.

[0147] Specific examples are used in this article, but the above description is only to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas; technicians in this field should understand that the modules or steps of the present invention can be implemented by a general computer device, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by a computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps in them can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0148] Meanwhile, for those skilled in the art, according to the concept of the present invention, there will be changes in the specific implementation and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting the content of rare earth elements, characterized in that: The method comprises the following steps: Constructing a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents; Acquire the color characteristics of the target rare earth solution image; Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color feature meets a preset rule as a learning set; A least squares support vector machine is used to establish an element component content prediction model, the color features of the historical rare earth solution images in the learning set are used as the input of the model, and the corresponding historical rare earth element component content is used as the target output of the model; Inputting the target rare earth solution image color feature into the element component content prediction model to obtain the target rare earth element component content; Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color feature meets a preset rule as a learning set, specifically including: The gray correlation between the target rare earth solution image color feature and each historical rare earth solution image color feature is calculated to determine the similar rare earth solution image color feature with the strongest correlation; the gray correlation between the two rare earth solution image color features is calculated according to the following formula: Where n represents the dimension of the color feature of the rare earth solution image, r ij,m Represents the color feature x of the rare earth solution image i and rare earth solution image color feature x j The grey relational coefficient of the mth dimension is calculated according to the following formula: Among them, x i,m Represents the color feature x of the rare earth solution image i The value of the mth dimension, x j,m Represents the color feature x of the rare earth solution image j The value of the mth dimension, ρ is the resolution coefficient; Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation degree with the similar rare earth solution image color feature is greater than a preset comparison threshold as a learning set; Before selecting a plurality of historical rare earth solution image color features having a gray correlation degree close to the target rare earth solution image color feature from the historical database as a learning set, the method further includes: Selecting the historical rare earth solution image color features in the historical database as iterative rare earth solution image color features, and the corresponding historical rare earth element component content as iterative rare earth element component content; Selecting from the historical database a number of historical rare earth solution image color features whose gray correlation with the iterative rare earth solution image color feature meets a preset rule as a learning set; A least squares support vector machine is used to establish an element component content prediction model, the color features of the historical rare earth solution images in the learning set are used as the input of the model, and the corresponding historical rare earth element component content is used as the target output of the model; Inputting the iterative rare earth solution image color feature into the element component content prediction model to obtain an iterative rare earth element component content prediction value; The prediction error is obtained by comparing the iterative rare earth element component content prediction value with the iterative rare earth element component content.

2. The method according to claim 1, characterized in that The method further comprises: Calculating the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and selecting the maximum value among each root mean square as an update criterion; The size of the update criterion and the update threshold is determined. If the update criterion is greater than the update threshold, the target rare earth solution image color feature and the target rare earth element component content are added to the historical database; otherwise, the historical database is kept unchanged.

3. The method according to claim 1, characterized in that The method further comprises: using a genetic algorithm to iteratively optimize the resolution coefficient and the preset comparison threshold to minimize the prediction error.

4. The method according to claim 3, characterized in that: The iterative optimization of the resolution coefficient and the preset comparison threshold by using a genetic algorithm specifically includes: Establish a parent population, each individual in the population has a corresponding fitness value; A roulette wheel method is used to select a number of individuals whose fitness values ​​are greater than a set threshold from the parent population to obtain a screening population; In the screening population, genetic operations are performed on the population according to the initial crossover probability and mutation probability to obtain an offspring population; The offspring population is used as a new parent population, and the steps of "using the roulette method to select a number of individuals with fitness values ​​greater than a set threshold from the parent population to obtain a screening population; in the screening population, performing genetic operations on the population according to the initial crossover probability and mutation probability to obtain an offspring population" are repeated until the preset number of genetic times is reached.

5. The method according to claim 4, characterized in that The method further comprises: Determine the change of a single optimal fitness value; the single optimal fitness value is the maximum value of each fitness value in a single genetic operation; If the number of times the single optimal fitness value remains unchanged in a number of consecutive genetic operations is greater than a stagnation threshold, modifying the crossover probability and mutation probability of the genetic algorithm; until the single optimal fitness value changes, restoring the number of times, the crossover probability and the mutation probability to initial values; If the single optimal fitness value changes, the constant number is reset, and the crossover probability and the mutation probability are kept unchanged.

6. A rare earth element component content prediction system, characterized in that: The system is used to implement the method for predicting the content of rare earth elements as claimed in claims 1 to 5, and the system comprises: A historical database construction module is used to construct a historical database; the historical database includes a number of historical rare earth solution image color features under different working conditions and corresponding historical rare earth element component contents; A grey correlation calculation module is used to calculate the grey correlation between any two rare earth solution image color features; A target input acquisition module is used to acquire the color characteristics of the target rare earth solution image; A learning set selection module, used to select from the historical database a number of historical rare earth solution image color features whose gray correlation with the target rare earth solution image color features meets a preset rule as a learning set; A prediction model establishment module, used to establish an element component content prediction model using a least squares support vector machine, using the color features of the historical rare earth solution images in the learning set as the input of the model, and using the corresponding historical rare earth element component content as the target output of the model; The target output acquisition module is used to input the color characteristics of the target rare earth solution image into the element component content prediction model to obtain the target rare earth element component content.

7. The system according to claim 6, characterized in that The system further comprises: An update criterion determination module is used to calculate the root mean square of the Euclidean distance between the target rare earth solution image color feature and each historical rare earth solution image color feature in the historical database, and select the maximum value of each root mean square as the update criterion; A comparison module, used for comparing the update criterion and the update threshold; if the update criterion is greater than the update threshold, calling the database update module to update the historical database; otherwise, keeping the historical database unchanged; The database updating module is used to add the target rare earth solution image color characteristics and the target rare earth element component content to the historical database.

Images