A fully automatic boat precision control method and system based on artificial intelligence

Abstract

The invention discloses a fully automatic boat precision control method and system based on artificial intelligence, which relates to the field of control systems. The invention adopts different boat samples to hold different reducing materials respectively and puts them into a high-temperature environment for reduction operation; then collects images of the boat samples and extracts features, so as to determine whether there are cracks on the surface of the boat after the reduction reaction; then constructs a decision tree classification model, and improves and tests the decision tree model by setting boat sample training images and test images; then classifies image feature data in a second boat sample image matrix by using a lion group algorithm to determine the physical properties of the boat; finally, retains and removes boat samples according to the classification results of the boat classification; thereby providing better boat samples and data basis for manufacturing the boat for subsequent reduction operation to control the components of the boat.

Description

Technical Field

[0001] The present invention belongs to the field of control systems, and in particular, relates to a fully automatic material boat precision control method and system based on artificial intelligence. Background Art

[0002] A sintering boat is a high-temperature container used for sintering and melting metal powders in high-temperature environments. It is usually made of molybdenum sheets or molybdenum-based alloys and formed into specific shapes and structures by welding, stamping or other high-temperature processing methods. The sintering boat has high melting point, high strength, good electrical and thermal conductivity, and high-temperature and corrosion resistance, making it an ideal choice for high-temperature industrial applications. The sintering boat has a variety of uses in the fields of materials science and engineering, including: as equipment for melting rare earth materials and vessels for sintering tungsten oxide, carrying and moving high-temperature materials in experimental and industrial applications in high-temperature environments, and in the vacuum coating industry, as a carrier of evaporation sources for precise evaporation and deposition of materials. The shape, size and structure of the sintering boat can be customized according to specific application requirements; for example, they may have a specific trough design to accommodate different material handling and evaporation processes. In addition, the sintering boat can also be surface treated (such as alumina coating) to improve its durability in certain corrosive environments.

[0003] Chinese patent CN117631623A discloses an intelligent and precise control method for silicon content in molten iron, which includes the following steps: Step 1: Through multi-sensor fusion technology, multi-source data collection in the molten iron production process, such as temperature, flow, chemical composition, vibration, and sound, is realized to obtain comprehensive and rich production data; Step 2: Using data analysis and machine learning technology, the collected data is processed and analyzed in real time, a multivariate nonlinear correlation model between the silicon content in molten iron and the production process parameters is constructed, and efficient feature extraction and modeling of the data are realized to achieve precise control of the silicon content in molten iron. Through multi-sensor fusion technology, data analysis, and machine learning technology, comprehensive monitoring and real-time analysis of the molten iron production process is realized, resource waste and energy consumption are reduced, and ultra-precise control of the silicon content in molten iron can be achieved through real-time monitoring and analysis, improving product quality stability.

[0004] A material boat is usually used to hold reactants that need to be reduced at high temperature, and then the material boat holding the reactants is placed in a high temperature environment for a reduction reaction. However, under long-term high temperature conditions, the internal crystals of the material that makes up the existing material boat will deform, causing changes in the overall physical properties of the material boat. The material boat will become fragile, and if it continues to be used, it will easily break, resulting in reduction failure and causing losses to the company. Summary of the invention

[0005] In view of the problems in the related art, the present invention proposes a fully automatic boat precision control method and system based on artificial intelligence to overcome the above-mentioned technical problems existing in the existing related art.

[0006] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] The present invention is a fully automatic boat precision control method based on artificial intelligence, comprising the following steps:

[0008] S1, setting a set of boat contents types, a boat component material type matrix, a corresponding boat contents sample matrix, and a corresponding reduction time-temperature correspondence matrix; then setting a boat component ratio data matrix and obtaining a boat sample matrix to be measured according to the boat component ratio data matrix; then performing a high-temperature reduction operation on the placed boat contents samples according to the reduction time-temperature correspondence matrix to obtain a reduced boat contents sample matrix and a reduced boat sample matrix;

[0009] S2, performing image acquisition on each reduced boat sample in the reduced boat sample matrix to obtain a first boat sample image matrix; then performing corrosion operation on each corresponding reduced boat sample in the reduced boat sample matrix and performing image acquisition to obtain a second boat sample image matrix; setting training images and test images of the boat samples after high-temperature reduction and corresponding image labels;

[0010] S3, setting a first acquisition feature category set; using a convolutional neural network to extract features from the boat sample training image and the test image and the first boat sample image matrix; setting an initial decision tree, using the boat sample training image and the test image to construct and test the initial decision tree to obtain a final decision tree;

[0011] S4, setting a second collection feature category set and using a convolutional neural network model to extract features from the second boat sample image matrix to obtain a feature data set of the corroded boat sample image; constructing a lion group population to perform classification operations on the feature data set of the corroded boat sample image to obtain an optimal sample image feature data set classification set;

[0012] S5, using the final decision tree to perform a classification operation on the first material boat sample image feature data set matrix to obtain a first classification result, and eliminating samples in the restored material boat sample matrix according to the first classification result and the optimal sample image feature data set classification set;

[0013] The material boat is preferably a molybdenum material boat; the types of materials contained in the material boat include, for example, MoO 2 and MoO 3Powder, etc.; the material type matrix of the material boat components includes iron-based and nickel-based high-temperature alloys, and the components are carbon, iron, nickel, chromium, cobalt and molybdenum, etc.; since each type of the contents is different, the high temperature required for reduction under high-temperature conditions is different, and the time required for reduction is also different, so these factors are analyzed and processed separately in the scheme of the present invention; when the material boat is filled with the reducing material for high-temperature reduction, if the material used to manufacture the material boat is not suitable, the material boat will become fragile under long-term high-temperature conditions and easily crack; in the scheme, by setting the material boat component ratio data matrix and the corresponding material boat sample matrix to be measured, these material boat samples are used to hold different reducing materials and put them in a high-temperature environment for reduction operation; after the reduction operation is completed, the images of these material boat samples are collected, and then the features of the images are extracted to determine whether there are cracks on the surface of the material boat after the reduction reaction; at the same time, the material boats are coated with a corrosive agent to corrode the surface of the material boat, and after the corrosion is completed, the corroded surface is photographed with a microscope to obtain to the second material boat sample image matrix; then construct a decision tree classification model, and improve and test the decision tree model by setting the material boat sample training image and test image to obtain the final decision tree; so that the final decision tree can determine the category of the material boat according to the characteristics of the material boat surface image; the categories of the material boat include large cracks, medium cracks, small cracks and no cracks, so as to determine the quality of the material boat; then classify the image feature data in the second material boat sample image matrix by using the lion group algorithm, and determine the physical properties of the material boat according to the aspect ratio of the microscopic crystals of the material boat at the corrosion surface and the distribution of the crystals, and the categories are divided into good physical properties, better physical properties, general physical properties, poor physical properties and very poor physical properties; finally, according to these classification results of the material boat, the material boat samples are retained and eliminated, as well as the corresponding components and proportions of the retained material boats; thereby providing a better material boat sample and a data basis for manufacturing the material boat for the subsequent reduction operation to control the components of the material boat.

[0014] Preferably, the S1 comprises the following steps:

[0015] S11, set the type set of materials contained in the material boat a = {a 1 ,a 2 ,...,a i ,...,a a′}, a i represents the type of the i-th material boat, a′ represents the total number of the set material boat types; according to the material boat type set a={a 1 ,a 2 ,...,a i ,...,a a′}Set the material type matrix of the boat components as follows,

[0016]

[0017] in, Indicates the type of material of the jth material boat component set when the i-th material boat is loaded. Indicates the total number of boat component material types set when the i-th boat is loaded with the contents;

[0018] S12, according to the type set of objects contained in the material boat a={a 1 ,a 2 ,...,a i ,...,a a′} Set the corresponding sample matrix c of the material boat as follows:

[0019]

[0020] Among them, c ij represents the j-th sample of the material boat corresponding to the i-th material boat type, and c′ represents the total number of samples collected for each material boat type;

[0021] According to the sample matrix c of the material boat, the corresponding reduction time-temperature correspondence matrix is ​​set as follows,

[0022]

[0023] in, and They respectively represent the minimum temperature and the corresponding consumption time required for reducing the j-th sample of the material boat corresponding to the i-th type of material boat;

[0024] According to the material type matrix of the boat components Set the boat component ratio data matrix b as follows:

[0025]

[0026] Among them, b ij represents the j-th material boat component material ratio data set when the i-th material boat is loaded, and b′ represents the total number of material boat component material ratio data set when each material boat is loaded;

[0027] Then, the sample matrix of the boat to be measured is obtained according to the boat component ratio data matrix b. as follows,

[0028]

[0029] in, According to bij A boat sample obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type;

[0030] S13, using the sample matrix of the boat to be measured Each of the material boat samples in the material boat sample matrix c is placed in the corresponding material boat sample, and then the reduction time and temperature correspondence matrix is ​​used to calculate the reduction time and temperature correspondence matrix. The sample of the material boat is subjected to high temperature reduction operation to obtain the reduced material boat sample matrix They are as follows:

[0031]

[0032] in, After high temperature reduction, according to b ij A boat sample obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type;

[0033] By setting the boat content type set, boat component material type matrix, boat content sample matrix, reduction time-temperature correspondence matrix, boat component ratio data matrix, boat sample matrix to be measured and reduced boat sample matrix, a sample basis is provided for subsequent boat classification operations, thereby completing the retention and elimination of boat samples.

[0034] Preferably, S2 comprises the following steps:

[0035] S21, the reduced boat sample matrix The images of the reduced boat samples in the sample are collected to obtain the first boat sample image matrix d; as follows,

[0036]

[0037] Among them, d ij Express Image acquisition is performed to obtain the image of the reduced boat sample, d ij ={d ij1 ,d ij2 ,...,d ijk ,...,d ijd′}, d ijk Express The image acquisition is performed to obtain the kth image of the reduced boat sample image, and d′ represents the Perform image acquisition to obtain the total number of images in the restored boat sample image;

[0038] S22, according to the material type matrix of the material boat components Set Corrosive Agent Set Represents the material type matrix of the boat components The etchant corresponding to the material type of each boat component in the i-th row in the etchant set is then used. The various corrosive agents in the reduced material boat sample matrix Corrosion operation is performed on the corresponding reduced boat samples to obtain the corroded boat sample matrix as follows,

[0039]

[0040] in, Indicates the use of Corrosive A corroded boat sample obtained after corrosion;

[0041] Then the corrosion boat sample matrix The image of each boat sample in is collected to obtain the second boat sample image matrix e; as follows,

[0042]

[0043] Among them, e ij Express Perform image acquisition to obtain an image of the boat sample after corrosion;

[0044] e ij ={e ij1 ,e ij2 ,...,e ijk ,...,e ije′},e ijk Express The image acquisition is performed to obtain the kth image of the corroded boat sample image, and e′ represents the Perform image acquisition to obtain the total number of images in the image of the corroded boat sample;

[0045] S23, according to the sample matrix c of the material boat and the corresponding relationship matrix of reduction time and temperature Collect the training image and test image of the material boat sample after high temperature reduction to obtain the training image matrix of the material boat sample after reduction And the image matrix of the boat sample test after reduction They are as follows:

[0046]

[0047] in, Indicates that the ijThe training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation, and f represents the total number of training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation of the boat containing the material type of each boat; Indicates that the ij f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation, and f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation for each type of material in the boat;

[0048] S24, setting the restored boat sample training image matrix And the image matrix of the boat sample test after reduction Corresponding image labels, obtaining a first image label data matrix and a second image label data matrix;

[0049] By performing image acquisition and feature extraction on the reduced material boat sample matrix, corroding the material boats in the reduced material boat sample matrix, and then performing image acquisition and feature extraction on the corroded material boat samples, a data basis is provided for the subsequent classification of the material boats in the reduced material boat sample matrix; and by setting the material boat sample training images and test images and performing feature extraction, data support is provided for the subsequent construction and testing of the decision tree.

[0050] Preferably, S3 comprises the following steps:

[0051] S31. Setting the first collection feature category set represents the i-th feature category in the first collection feature category set, represents the total number of feature categories set in the first collection feature category set; according to the first collection feature category set The convolutional neural network model is used to train the image matrix of the reduced boat sample And the test image matrix of the boat sample after reduction Perform feature extraction to obtain the restored boat sample training image feature data set matrix h and the restored boat sample test image feature data set matrix h′; they are as follows:

[0052]

[0053] Among them, h ij express The feature dataset, h ijk Indicates h ij The kth feature data in h i 'j express The feature dataset, h i ' jk Indicates h i ' j The kth feature data in ;

[0054] Then, a convolutional neural network model is used to extract features from the first material boat sample image matrix d to obtain the first material boat sample image feature data set matrix as follows,

[0055]

[0056] in, Indicates d ij An image feature dataset obtained by feature extraction;

[0057] S32: construct an initial decision tree, calculate the information entropy of the image label data in the first image label data matrix, and obtain the first label data information entropy. The calculation formula is as follows,

[0058]

[0059] In the formula, l i represents the i-th category image label data in the first image label data matrix, represents the probability of the i-th type of image label data appearing in the first image label data matrix, and l′ represents the total number of image label data types in the first image label data matrix;

[0060] S33, according to the first collection feature category set And the restored boat sample training image feature data set matrix h calculates the conditional information entropy of the label data under the condition of the image feature existence, and obtains the first conditional information entropy set Represents the conditional information entropy of the i-th feature category on the image label data;

[0061] S34, selecting the first conditional information entropy set Information entropy of the first label data The feature category corresponding to the conditional information entropy corresponding to the maximum value of the absolute value of the difference is taken as the first layer of the initial decision tree, recorded as the first feature category; at the same time, the first acquisition feature category set Lieutenant General The corresponding feature categories are eliminated to obtain the first collection feature category set after elimination m irepresents the i-th feature category in the first collected feature category set after one elimination;

[0062] S35, recalculating the first collection feature category set based on various feature data of the first feature category The conditional information entropy of each feature category in the second conditional information entropy set is obtained. represents the conditional information entropy of the i-th feature category in the first collected feature category set on various feature data of the first feature category after one elimination;

[0063] S36, continuously repeating S33, S34 and S35, when the first collection feature category set When there are no feature categories left, stop repeating the operation and get the final decision tree;

[0064] S37, using the reduced boat sample test image feature data set matrix h′ to test the final decision tree;

[0065] By calculating the conditional information entropy of each material boat sample image feature category on the material boat category, a decision tree is constructed according to the degree of influence from large to small according to the conditional information entropy, and the final decision tree is obtained after the construction is completed; the final decision tree is then tested, and the decision tree is reconstructed or retained according to the test results, which further improves the classification accuracy of the final decision tree.

[0066] Preferably, the S33 comprises the following steps:

[0067] S331: Based on the first collection feature category set The acquisition feature subcategory matrix m′ is obtained as follows:

[0068]

[0069] Among them, m i ' j represents the jth feature subcategory in the i-th feature category; Represents the total number of feature subcategories in the i-th feature category;

[0070] S332, traverse the acquisition feature subcategory matrix m' row by row, and obtain the acquisition feature subcategory probability data matrix according to the restored boat sample training image feature data set matrix h and the acquisition feature subcategory matrix m'. as follows,

[0071]

[0072] in, represents the probability of the jth feature subcategory in the ith feature category appearing in the boat sample training image feature data set matrix h after the reduction;

[0073] At the same time, the probability of occurrence of various image label data in the first image label data is calculated under the condition that the feature subclass in the acquisition feature subclass matrix m′ appears, and the first conditional probability matrix n is obtained; as follows,

[0074]

[0075] Among them, n ij represents the probability of occurrence of various image label data in the first image label data when the jth feature subcategory in the i-th feature category appears;

[0076] S333: Based on the first conditional probability matrix n and the collection feature subcategory probability data matrix Get the first conditional information entropy set The calculation formula is as follows,

[0077]

[0078] Calculating the conditional information entropy based on the probability of occurrence of feature data can more accurately reflect the degree of influence of the feature data on the material boat category.

[0079] Preferably, the S37 comprises the following steps:

[0080] S371, using the final decision tree to classify the reduced boat sample test image feature data set matrix h′ to obtain a test classification result label data matrix

[0081] S372, set the accuracy threshold to o; the test classification result label data matrix Compare with the second image label data matrix and calculate the test classification result label data matrix The data that is the same as the data at the corresponding position in the second image label data matrix accounts for the test classification result label data matrix The ratio of all data in the test is used to obtain the test accuracy;

[0082] S373. When the test accuracy is greater than or equal to 0, the final decision tree meets the requirements;

[0083] Otherwise, repeat S31 to S36, regenerate a new final decision tree, and perform classification operations on the reduced boat sample test image feature data set matrix h′ until the test accuracy is greater than or equal to o, and then stop repeating;

[0084] By testing the final decision tree, the classification accuracy of the final decision tree is further guaranteed.

[0085] Preferably, S4 comprises the following steps:

[0086] S41, setting the second collection feature category set represents the i-th feature category in the second collection feature category set, represents the total number of feature categories set in the second collection feature category set; according to the second collection feature category set The convolutional neural network model is used to extract features from the second boat sample image matrix e, and the feature dataset matrix p of the boat sample image after corrosion is obtained; as follows,

[0087]

[0088] Among them, p ij Indicates e ij The image feature dataset of the corroded boat sample obtained after characterization; ij ={p ij1 ,p ij2 ,...,p ijk ,...,p ijp′}, p ijk Indicates p ij The kth feature data in p′ represents p ij The total number of feature data in;

[0089] S42, constructing a lion group population, setting the size of the lion group population to n′; the lion group population is represented by represents the i-th lion in the lion population; sets the maximum number of iterations to o′;

[0090] S43, randomly divide each of the boat sample image feature data sets in the corroded boat sample image feature data set matrix p into equal parts Class, get the classification set of the initial sample image feature data set represents the i-th sample image feature data set randomly divided into equal parts; according to the classification set of the initial sample image feature data set Get the classification center set of the first initial sample image feature data set The initial sample image feature data set is classified into a central set As a lion group The initial position of the first lion in the group; repeat the above operation for the lion group population The initial positions of the lions are set to obtain the initial position set q′={q1 ′,q′ 2 ,...,q i ′,...,q′ n′}, q i ′ represents the initial position of the i-th lion in the lion population;

[0091] S44, calculate the discreteness of the classification set obtained by each random classification in S43, and obtain the initial discreteness matrix as follows,

[0092]

[0093] in, represents the discreteness of the classification set of the j-th sample image feature data set when the initial position of the i-th lion in the lion population is set;

[0094] According to the initial discreteness matrix Set the fitness function of the lion population, the function formula is as follows:

[0095]

[0096] S45, start the iteration operation, set the current number of iterations to In each iteration, the position of each lion in the lion population is adjusted according to the fitness function of the lion population;

[0097] when When , the iteration operation is stopped and the optimal sample image feature data set classification set is output

[0098] By constructing a lion population, the image feature data of the corroded material boat sample is first randomly classified, and then the classification center is used as the initial position of the lion in the lion population; at the same time, the inverse of the overall discreteness of the entire image feature data of the corroded material boat sample is used as the fitness of each lion in the lion population. In the continuous iterative process, the overall discreteness of the image feature data of the material boat sample continues to decrease, reflecting that the classification effect is getting better and better.

[0099] Preferably, the S5 comprises the following steps:

[0100] S51, using the final decision tree to analyze the first boat sample image feature data set matrix Perform a classification operation to obtain a first classification result;

[0101] S52, judging whether there are cracks on the surface of the boat according to the first classification result; and then classifying the optimal sample image feature data set according to the optimal sample image feature data set. Determine whether the toughness and strength of the boat are reduced;

[0102] S53, retaining the classification sets without cracks and the boats whose toughness and strength are not reduced; otherwise, discarding them;

[0103] The boat samples are retained and eliminated according to the first classification result and the optimal sample image feature data set classification set, so that the retention and elimination results are more accurate.

[0104] An artificial intelligence-based fully automatic material boat precision control system, comprising a material boat sample collection module, a material boat restored image feature collection module, a material boat corroded image feature collection module, a material boat restored image feature classification module, a material boat image feature classification module and a material boat sample rejection module;

[0105] The boat sample collection module is used to set a boat component ratio data matrix and obtain a boat sample matrix to be measured according to the boat component ratio data matrix;

[0106] The boat restored image feature acquisition module is used to extract features from the boat sample training image and test image using a convolutional neural network;

[0107] The boat corrosion image feature acquisition module is used to extract features from the first boat sample image matrix;

[0108] The boat image feature classification module is used to construct a decision tree and a lion group population to perform classification operations on the first boat sample image feature data set matrix and the corroded boat sample image feature data set respectively;

[0109] The boat sample elimination module is used to eliminate samples in the restored boat sample matrix according to the first classification result and the optimal sample image feature data set classification set.

[0110] The present invention has the following beneficial effects:

[0111] 1. In the present invention, a boat sample collection module, a boat reduction image feature collection module, a boat corrosion image feature collection module, a boat reduction image feature classification module, a boat image feature classification module and a boat sample rejection module are set; a boat component ratio data matrix and a corresponding boat sample matrix to be determined are set, and different reducing materials are placed in these boat samples and placed in a high temperature environment for reduction operation; after the reduction operation is completed, images of these boat samples are collected, and then features of the images are extracted to determine whether there are cracks on the surface of the boat after the reduction reaction; at the same time, a corrosive agent is applied to the boat surface to corrode it, and after the corrosion is completed, the corroded surface is photographed with a microscope to obtain a second boat sample image matrix; a decision tree classification model is then constructed, and the decision tree model is improved and tested by setting boat sample training images and test images to obtain the best final decision tree; the final decision tree can determine the category of the material boat according to the characteristics of the material boat surface image; the categories of the material boat include large crack category, medium crack category, small crack category and no crack category, so as to determine the quality of the material boat; then the image feature data in the second material boat sample image matrix is ​​classified by using the lion group algorithm, and the physical properties of the material boat are determined according to the aspect ratio of the microscopic crystals of the material boat at the corrosion surface and the distribution of the crystals, and the categories are divided into good physical properties, better physical properties, general physical properties, poor physical properties and very poor physical properties; finally, according to the classification results of the material boat, the material boat samples are retained and eliminated, as well as the corresponding components and proportions of the retained material boats; thereby providing better material boat samples and data basis for manufacturing the material boat for controlling the components of the material boat in the subsequent reduction operation, ensuring the success rate of the subsequent reduction reaction and the stability of the material boat.

[0112] 2. In the present invention, by setting a set of types of boat contents, a matrix of types of boat component materials, a matrix of samples of boat contents, a matrix of reduction time-temperature correspondence, a data matrix of boat component ratios, a matrix of boat samples to be measured, and a matrix of reduced boat samples, a sample basis is provided for subsequent boat classification operations, thereby completing the retention and elimination of boat samples, and then performing image acquisition and feature extraction on the reduced boat sample matrix, and then performing corrosion operations on the boats in the reduced boat sample matrix, and then performing image acquisition and feature extraction on the corroded boat samples; providing a data basis for subsequent classification of the boats in the reduced boat sample matrix; and providing data support for subsequent construction and testing of a decision tree by setting boat sample training images and test images and performing feature extraction.

[0113] 3. In the present invention, the conditional information entropy of each material boat sample image feature category on the material boat category is calculated, and a decision tree is constructed according to the degree of influence from large to small according to the conditional information entropy, and the final decision tree is obtained after the construction is completed; then the final decision tree is tested, and the decision tree is reconstructed or retained according to the test results, so as to further improve the classification accuracy of the final decision tree; then, by constructing a lion population, the corroded material boat sample image feature data is first randomly classified, and then the classification center is used as the initial position of the lion in the lion population; at the same time, the inverse of the overall discreteness of the entire corroded material boat sample image feature data is used as the fitness of each lion in the lion population. In the continuous iterative process, the overall discreteness of the material boat sample image feature data is continuously reduced, reflecting that the classification effect is getting better and better.

[0114] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0116] Figure 1 The present invention is a schematic diagram of a process of controlling a material boat by a fully automatic material boat precision control system based on artificial intelligence. DETAILED DESCRIPTION

[0117] The following will be combined with the drawings in the embodiments of the invention to clearly and completely describe the technical solutions in the embodiments of the invention. Obviously, the described embodiments are only part of the embodiments of the invention, not all of the embodiments. Based on the embodiments in the invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the invention.

[0118] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0119] The present invention is a fully automatic boat precision control method based on artificial intelligence, comprising the following steps:

[0120] S1, setting a set of boat contents types, a boat component material type matrix, a corresponding boat contents sample matrix, and a corresponding reduction time-temperature correspondence matrix; then setting a boat component ratio data matrix and obtaining a boat sample matrix to be measured according to the boat component ratio data matrix; then performing a high-temperature reduction operation on the placed boat contents samples according to the reduction time-temperature correspondence matrix to obtain a reduced boat contents sample matrix and a reduced boat sample matrix;

[0121] The S1 comprises the following steps:

[0122] S11, set the type set of materials contained in the material boat a = {a 1 ,a 2 ,...,a i ,...,a a′}, a i represents the type of the i-th material boat, a′ represents the total number of the set material boat types; according to the material boat type set a={a 1 ,a 2 ,...,a i ,...,a a′}Set the material type matrix of the boat components as follows,

[0123]

[0124] in, Indicates the type of material of the jth material boat component set when the i-th material boat is loaded. Indicates the total number of boat component material types set when the i-th boat is loaded with the contents;

[0125] S12, according to the type set of objects contained in the material boat a={a 1 ,a 2 ,...,a i ,...,a a′} Set the corresponding sample matrix c of the material boat as follows:

[0126]

[0127] Among them, c ij represents the j-th sample of the material boat corresponding to the i-th material boat type, and c′ represents the total number of samples collected for each material boat type;

[0128] According to the sample matrix c of the material boat, the corresponding reduction time-temperature correspondence matrix is ​​set as follows,

[0129]

[0130] in, and They respectively represent the minimum temperature and the corresponding consumption time required for reducing the j-th sample of the material boat corresponding to the i-th type of material boat;

[0131] According to the material type matrix of the boat components Set the boat component ratio data matrix b as follows:

[0132]

[0133] Among them, b ij represents the j-th material boat component material ratio data set when the i-th material boat is loaded, and b′ represents the total number of material boat component material ratio data set when each material boat is loaded;

[0134] Then, the sample matrix of the boat to be measured is obtained according to the boat component ratio data matrix b. as follows,

[0135]

[0136] in, According to b ij A boat sample obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type;

[0137] S13, using the sample matrix of the boat to be measured Each of the material boat samples in the material boat sample matrix c is placed in the corresponding material boat sample, and then the reduction time and temperature correspondence matrix is ​​used to calculate the reduction time and temperature correspondence matrix. The sample of the material boat is subjected to high temperature reduction operation to obtain the reduced material boat sample matrix They are as follows:

[0138]

[0139] in, After high temperature reduction, according to b ij A boat sample obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type;

[0140] S2, performing image acquisition on each reduced boat sample in the reduced boat sample matrix to obtain a first boat sample image matrix; then performing corrosion operation on each corresponding reduced boat sample in the reduced boat sample matrix and performing image acquisition to obtain a second boat sample image matrix; setting training images and test images of the boat samples after high-temperature reduction and corresponding image labels;

[0141] The S2 comprises the following steps:

[0142] S21, the reduced boat sample matrix The images of the reduced boat samples in the sample are collected to obtain the first boat sample image matrix d; as follows,

[0143]

[0144] Among them, d ij Express Image acquisition is performed to obtain the image of the reduced boat sample, d ij ={d ij1 ,d ij2 ,...,d ijk ,...,d ijd′}, d ijk Express The image acquisition is performed to obtain the kth image of the reduced boat sample image, and d′ represents the Perform image acquisition to obtain the total number of images in the restored boat sample image;

[0145] S22, according to the material type matrix of the material boat components Set Corrosive Agent Set Represents the material type matrix of the boat components The etchant corresponding to the material type of each boat component in the i-th row in the etchant set is then used. The various corrosive agents in the reduced material boat sample matrix Corrosion operation is performed on the corresponding reduced boat samples to obtain the corroded boat sample matrix as follows,

[0146]

[0147] in, Indicates the use of Corrosive A corroded boat sample obtained after corrosion;

[0148] Then the corrosion boat sample matrix The image of each boat sample in is collected to obtain the second boat sample image matrix e; as follows,

[0149]

[0150] Among them, e ij Express Perform image acquisition to obtain an image of the boat sample after corrosion;

[0151] e ij ={e ij1 ,e ij2 ,...,e ijk ,...,e ije′},e ijk Express The image acquisition is performed to obtain the kth image of the corroded boat sample image, and e′ represents the Perform image acquisition to obtain the total number of images in the image of the corroded boat sample;

[0152] S23, according to the sample matrix c of the material boat and the corresponding relationship matrix of reduction time and temperature Collect the training image and test image of the material boat sample after high temperature reduction to obtain the training image matrix of the material boat sample after reduction And the image matrix of the boat sample test after reduction They are as follows:

[0153]

[0154] in, Indicates that the ij The training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation, and f represents the total number of training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation of the boat containing the material type of each boat; Indicates that the ij f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation, and f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation for each type of material in the boat;

[0155] S24, setting the restored boat sample training image matrix And the image matrix of the boat sample test after reduction Corresponding image labels, obtaining a first image label data matrix and a second image label data matrix;

[0156] S3, setting a first acquisition feature category set; using a convolutional neural network to extract features from the boat sample training image and the test image and the first boat sample image matrix; setting an initial decision tree, using the boat sample training image and the test image to construct and test the initial decision tree to obtain a final decision tree;

[0157] The S3 comprises the following steps:

[0158] S31. Setting the first collection feature category set represents the i-th feature category in the first collection feature category set, represents the total number of feature categories set in the first collection feature category set; according to the first collection feature category set The convolutional neural network model is used to train the image matrix of the reduced boat sample And the test image matrix of the boat sample after reduction Perform feature extraction to obtain the restored boat sample training image feature data set matrix h and the restored boat sample test image feature data set matrix h′; they are as follows:

[0159]

[0160] Among them, h ij express The feature dataset, h ijk Indicates h ij The kth feature data in h i ' j express The feature dataset, h i ' jk Indicates h i ' j The kth feature data in ;

[0161] Then, a convolutional neural network model is used to extract features from the first material boat sample image matrix d to obtain the first material boat sample image feature data set matrix as follows,

[0162]

[0163] in, Indicates d ij An image feature dataset obtained by feature extraction;

[0164] S32: construct an initial decision tree, calculate the information entropy of the image label data in the first image label data matrix, and obtain the first label data information entropy. The calculation formula is as follows,

[0165]

[0166] In the formula, l i represents the i-th category image label data in the first image label data matrix, represents the probability of the i-th type of image label data appearing in the first image label data matrix, and l′ represents the total number of image label data types in the first image label data matrix;

[0167] S33, according to the first collection feature category set And the restored boat sample training image feature data set matrix h calculates the conditional information entropy of the label data under the condition of the image feature existence, and obtains the first conditional information entropy set Represents the conditional information entropy of the i-th feature category on the image label data;

[0168] The S33 comprises the following steps:

[0169] S331: Based on the first collection feature category set The acquisition feature subcategory matrix m′ is obtained as follows:

[0170]

[0171] Among them, m i ' j represents the jth feature subcategory in the i-th feature category; Represents the total number of feature subcategories in the i-th feature category;

[0172] S332, traverse the acquisition feature subcategory matrix m' row by row, and obtain the acquisition feature subcategory probability data matrix according to the restored boat sample training image feature data set matrix h and the acquisition feature subcategory matrix m'. as follows,

[0173]

[0174] in, represents the probability of the jth feature subcategory in the ith feature category appearing in the boat sample training image feature data set matrix h after the reduction;

[0175] At the same time, the probability of occurrence of various image label data in the first image label data is calculated under the condition that the feature subclass in the acquisition feature subclass matrix m′ appears, and the first conditional probability matrix n is obtained; as follows,

[0176]

[0177] Among them, n ij represents the probability of occurrence of various image label data in the first image label data when the jth feature subcategory in the i-th feature category appears;

[0178] S333: Based on the first conditional probability matrix n and the collection feature subcategory probability data matrix Get the first conditional information entropy set The calculation formula is as follows,

[0179]

[0180] S34, selecting the first conditional information entropy set Information entropy of the first label data The feature category corresponding to the conditional information entropy corresponding to the maximum value of the absolute value of the difference is taken as the first layer of the initial decision tree, recorded as the first feature category; at the same time, the first acquisition feature category set Lieutenant General The corresponding feature categories are eliminated to obtain the first collection feature category set after elimination m i represents the i-th feature category in the first collected feature category set after one elimination;

[0181] S35, recalculating the first collection feature category set based on various feature data of the first feature category The conditional information entropy of each feature category in the second conditional information entropy set is obtained. represents the conditional information entropy of the i-th feature category in the first collected feature category set on various feature data of the first feature category after one elimination;

[0182] S36, continuously repeating S33, S34 and S35, when the first collection feature category set When there are no feature categories left, stop repeating the operation and get the final decision tree;

[0183] S37, using the reduced boat sample test image feature data set matrix h′ to test the final decision tree;

[0184] The S37 comprises the following steps:

[0185] S371, using the final decision tree to classify the reduced boat sample test image feature data set matrix h′ to obtain a test classification result label data matrix

[0186] S372, set the accuracy threshold to o; the test classification result label data matrix Compare with the second image label data matrix and calculate the test classification result label data matrix The data that is the same as the data at the corresponding position in the second image label data matrix accounts for the test classification result label data matrix The ratio of all data in the test is used to obtain the test accuracy;

[0187] S373. When the test accuracy is greater than or equal to 0, the final decision tree meets the requirements;

[0188] Otherwise, repeat S31 to S36, regenerate a new final decision tree, and perform classification operations on the reduced boat sample test image feature data set matrix h′ until the test accuracy is greater than or equal to o, and then stop repeating;

[0189] S4, setting a second collection feature category set and using a convolutional neural network model to extract features from the second boat sample image matrix to obtain a feature data set of the corroded boat sample image; constructing a lion group population to perform classification operations on the feature data set of the corroded boat sample image to obtain an optimal sample image feature data set classification set;

[0190] The S4 comprises the following steps:

[0191] S41, setting the second collection feature category set represents the i-th feature category in the second collection feature category set, represents the total number of feature categories set in the second collection feature category set; according to the second collection feature category set The convolutional neural network model is used to extract features from the second boat sample image matrix e, and the feature dataset matrix p of the boat sample image after corrosion is obtained; as follows,

[0192]

[0193] Among them, p ij Indicates e ij The image feature dataset of the corroded boat sample obtained after characterization; ij ={p ij1 ,p ij2 ,...,p ijk ,...,p ijp′}, p ijk Indicates p ij The kth feature data in p′ represents p ij The total number of feature data in;

[0194] S42, constructing a lion group population, setting the size of the lion group population to n′; the lion group population is represented by represents the i-th lion in the lion population; sets the maximum number of iterations to o′;

[0195] S43, randomly divide each of the boat sample image feature data sets in the corroded boat sample image feature data set matrix p into equal parts Class, get the classification set of the initial sample image feature data set represents the i-th sample image feature data set randomly divided into equal parts; according to the classification set of the initial sample image feature data set Get the classification center set of the first initial sample image feature data set The initial sample image feature data set is classified into a central set As a lion group The initial position of the first lion in the group; repeat the above operation for the lion group population The initial positions of the lions are set to obtain the initial position set q′={q 1 ′,q′ 2 ,...,q i ′,...,q′ n′}, q i ′ represents the initial position of the i-th lion in the lion population;

[0196] S44, calculate the discreteness of the classification set obtained by each random classification in S43, and obtain the initial discreteness matrix as follows,

[0197]

[0198] in, represents the discreteness of the classification set of the j-th sample image feature data set when the initial position of the i-th lion in the lion population is set;

[0199] According to the initial discreteness matrix Set the fitness function of the lion population, the function formula is as follows:

[0200]

[0201] S45, start the iteration operation, set the current number of iterations to In each iteration, the position of each lion in the lion population is adjusted according to the fitness function of the lion population;

[0202] when When , the iteration operation is stopped and the optimal sample image feature data set classification set is output

[0203]

[0204] S5, using the final decision tree to perform a classification operation on the first material boat sample image feature data set matrix to obtain a first classification result, and eliminating samples in the restored material boat sample matrix according to the first classification result and the optimal sample image feature data set classification set;

[0205] The S5 comprises the following steps:

[0206] S51, using the final decision tree to analyze the first boat sample image feature data set matrix Perform a classification operation to obtain a first classification result;

[0207] S52, judging whether there are cracks on the surface of the boat according to the first classification result; and then classifying the optimal sample image feature data set according to the optimal sample image feature data set. Determine whether the toughness and strength of the boat are reduced;

[0208] S53. The classification sets without cracks and the ones whose toughness and strength of the boat are not reduced are retained; otherwise, they are discarded.

[0209] An artificial intelligence-based fully automatic material boat precision control system, comprising a material boat sample collection module, a material boat restored image feature collection module, a material boat corroded image feature collection module, a material boat restored image feature classification module, a material boat image feature classification module and a material boat sample rejection module;

[0210] The boat sample collection module is used to set a boat component ratio data matrix and obtain a boat sample matrix to be measured according to the boat component ratio data matrix;

[0211] The boat restored image feature acquisition module is used to extract features from the boat sample training image and test image using a convolutional neural network;

[0212] The boat corrosion image feature acquisition module is used to extract features from the first boat sample image matrix;

[0213] The boat image feature classification module is used to construct a decision tree and a lion group population to perform classification operations on the first boat sample image feature data set matrix and the corroded boat sample image feature data set respectively;

[0214] The boat sample elimination module is used to eliminate samples in the restored boat sample matrix according to the first classification result and the optimal sample image feature data set classification set.

[0215] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0216] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. A fully automatic boat precision control method based on artificial intelligence, characterized in that: The following steps are involved: S1, setting a set of boat contents types, a boat component material type matrix, a corresponding boat contents sample matrix, and a corresponding reduction time-temperature correspondence matrix; then setting a boat component ratio data matrix and obtaining a boat sample matrix to be measured according to the boat component ratio data matrix; then performing a high-temperature reduction operation on the placed boat contents samples according to the reduction time-temperature correspondence matrix to obtain a reduced boat contents sample matrix and a reduced boat sample matrix; S2, performing image acquisition on each reduced boat sample in the reduced boat sample matrix to obtain a first boat sample image matrix; then performing corrosion operation on each corresponding reduced boat sample in the reduced boat sample matrix and performing image acquisition to obtain a second boat sample image matrix; setting training images and test images of the boat samples after high-temperature reduction and corresponding image labels; S3, setting a first acquisition feature category set; performing feature extraction on the boat sample training image and test image and the first boat sample image matrix; An initial decision tree is set, and the initial decision tree is constructed and tested using the boat sample training image and test image to obtain a final decision tree; S4, setting a second acquisition feature category set and performing feature extraction on the second boat sample image matrix to obtain a feature data set of the corroded boat sample image; constructing a lion group population to perform classification operations on the feature data set of the corroded boat sample image to obtain an optimal sample image feature data set classification set; S5, using the final decision tree to perform a classification operation on the first material boat sample image feature data set matrix to obtain a first classification result, and eliminating samples in the restored material boat sample matrix according to the first classification result and the optimal sample image feature data set classification set; The S4 comprises the following steps: S41, setting the second collection feature category set represents the i-th feature category in the second collection feature category set, represents the total number of feature categories set in the second collection feature category set; according to the second collection feature category set The convolutional neural network model is used to extract features from the second boat sample image matrix e, and the feature dataset matrix p of the boat sample image after corrosion is obtained; as follows, Among them, p ij Indicates e ij The image feature dataset of the corroded boat sample obtained after characterization; ij ={p ij1 ,p ij2 ,...,p ijk ,...,p ijp′ }, p ijk Indicates p ij The kth feature data in p′ represents p ij The total number of feature data in; S42, constructing a lion group population, setting the size of the lion group population to n′; the lion group population is represented by represents the i-th lion in the lion population; sets the maximum number of iterations to o′; S43, randomly divide each boat sample image feature data set in the corroded boat sample image feature data set matrix p into equal parts Class, get the classification set of the initial sample image feature data set represents the i-th sample image feature data set randomly divided into equal parts; according to the classification set of the initial sample image feature data set Get the classification center set of the first initial sample image feature data set The first initial sample image feature data set is classified into a central set As a lion group The initial position of the first lion in the group; repeat the above operation for the lion group population The initial positions of the lions are set to obtain the initial position set q′={q1′,q′2,...,q i ′,...,q′ n′ }, q i ′ represents the initial position of the i-th lion in the lion population; S44, calculate the discreteness of the classification set obtained by each random classification in S43, and obtain the initial discreteness matrix as follows, in, represents the discreteness of the classification set of the j-th sample image feature data set when the initial position of the i-th lion in the lion population is set; According to the initial discreteness matrix Set the fitness function of the lion population, the function formula is as follows: S45, start the iteration operation, set the current number of iterations to In each iteration, the position of each lion in the lion population is adjusted according to the fitness function of the lion population; when When , the iteration operation is stopped and the optimal sample image feature data set classification set is output 2. The method for fully automatic boat precision control based on artificial intelligence according to claim 1, characterized in that: The S1 comprises the following steps: S11, set the type set of the material boat to be placed a={a1, a2, ..., a i ,...,a a′ }, a i represents the type of the i-th material boat, a′ represents the total number of the set material boat types; according to the material boat type set a={a1, a2, ..., a i ,...,a a′ }Set the material type matrix of the boat components as follows, in, Indicates the type of material of the jth material boat component set when the i-th material boat is loaded. Indicates the total number of boat component material types set when the i-th boat is loaded with the contents; S12, according to the material boat containing object type set a={a1, a2, ..., a i ,...,a a′ } Set the corresponding sample matrix c of the material boat as follows: Among them, c ij represents the j-th sample of the material boat corresponding to the i-th material boat type, and c′ represents the total number of samples collected for each material boat type; According to the sample matrix c of the material boat, the corresponding reduction time-temperature correspondence matrix is ​​set as follows, in, and They respectively represent the minimum temperature and the corresponding consumption time required for reducing the j-th sample of the material boat corresponding to the i-th type of material boat; According to the material type matrix of the boat components Set the boat component ratio data matrix b as follows: Among them, b ij represents the j-th material boat component material ratio data set when the i-th material boat is loaded, and b′ represents the total number of material boat component material ratio data set when each material boat is loaded; Then, the sample matrix of the boat to be measured is obtained according to the boat component ratio data matrix b. as follows, in, According to b ij A boat sample obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type; S13, using the sample matrix of the boat to be measured Each of the material boat samples in the material boat sample matrix c is placed in the corresponding material boat sample, and then the reduction time and temperature correspondence matrix is ​​used to calculate the reduction time and temperature correspondence matrix. The sample of the material boat is subjected to high temperature reduction operation to obtain the reduced material boat sample matrix They are as follows: in, After high temperature reduction, according to b ij The boat sample is obtained by synthesizing the various boat component materials set when the i-th boat is filled with the content type.

3. The method for fully automatic boat precision control based on artificial intelligence according to claim 2, characterized in that: The S2 comprises the following steps: S21, the reduced boat sample matrix The images of the reduced boat samples in are collected to obtain the first boat sample image matrix d; as follows, Among them, d ij Express Image acquisition is performed to obtain the image of the reduced boat sample, d ij ={d ij1 ,d ij2 ,...,d ijk ,...,d ijd′ }, d ijk Express The image acquisition is performed to obtain the kth image of the reduced boat sample image, and d′ represents the Perform image acquisition to obtain the total number of images in the restored boat sample image; S22, according to the material type matrix of the boat components Set Corrosive Agent Set Represents the material type matrix of the boat components The etchant corresponding to the material type of each boat component in the i-th row in the etchant set is then used. The various corrosive agents in the reduced material boat sample matrix Corrosion operation is performed on the corresponding reduced boat samples to obtain the corroded boat sample matrix as follows, in, Indicates the use of Corrosive A corroded boat sample obtained after corrosion; Then the corrosion boat sample matrix The image of each boat sample in is collected to obtain the second boat sample image matrix e; as follows, Among them, e ij Express Perform image acquisition to obtain an image of the boat sample after corrosion; e ij ={e ij1 ,e ij2 ,...,e ijk ,...,e ije′ },e ijk Express The image acquisition is performed to obtain the kth image of the corroded boat sample image, and e′ represents the Perform image acquisition to obtain the total number of images in the image of the corroded boat sample; S23, according to the sample matrix c of the material boat and the corresponding relationship matrix of reduction time and temperature Collect the training image and test image of the material boat sample after high temperature reduction to obtain the training image matrix of the material boat sample after reduction And the image matrix of the boat sample test after reduction They are as follows: in, Indicates that the c ij The training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation, and f represents the total number of training images of the reduced boat samples are obtained by collecting training images of the boat after the high-temperature reduction operation of the boat containing the material type of each boat; Indicates that the c ij f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation, and f′ represents the total number of reduced boat sample test images obtained by collecting test images of the boat after the high-temperature reduction operation for each type of material in the boat; S24, setting the restored boat sample training image matrix And the image matrix of the boat sample test after reduction Corresponding image labels, a first image label data matrix and a second image label data matrix are obtained.

4. The method for fully automatic boat precision control based on artificial intelligence according to claim 3, characterized in that: The S3 comprises the following steps: S31. Setting the first collection feature category set represents the i-th feature category in the first collection feature category set, represents the total number of feature categories set in the first collection feature category set; according to the first collection feature category set The convolutional neural network model is used to train the image matrix of the reduced boat sample And the test image matrix of the boat sample after reduction Perform feature extraction to obtain the restored boat sample training image feature data set matrix h and the restored boat sample test image feature data set matrix h′; they are as follows: Among them, h ij express The feature dataset, h ijk Indicates h ij The kth feature data in h i ' j express The feature dataset, h i ' jk Indicates h i ' j The kth feature data in ; Then, a convolutional neural network model is used to extract features from the first material boat sample image matrix d to obtain the first material boat sample image feature data set matrix as follows, in, Indicates d ij An image feature data set obtained by feature extraction; S32: construct an initial decision tree, calculate the information entropy of the image label data in the first image label data matrix, and obtain the first label data information entropy. The calculation formula is as follows, In the formula, l i represents the i-th category image label data in the first image label data matrix, represents the probability of the i-th type of image label data appearing in the first image label data matrix, and l′ represents the total number of image label data types in the first image label data matrix; S33, according to the first collection feature category set And the restored boat sample training image feature data set matrix h calculates the conditional information entropy of the label data under the condition of the image feature existence, and obtains the first conditional information entropy set Represents the conditional information entropy of the i-th feature category on the image label data; S34, selecting the first conditional information entropy set Information entropy of the first label data The feature category corresponding to the conditional information entropy corresponding to the maximum value of the absolute value of the difference is taken as the first layer of the initial decision tree, recorded as the first feature category; at the same time, the first acquisition feature category set Lieutenant General The corresponding feature categories are eliminated to obtain the first collection feature category set after elimination m i represents the i-th feature category in the first collected feature category set after one elimination; S35, recalculating the first collection feature category set based on various feature data of the first feature category The conditional information entropy of each feature category in the second conditional information entropy set is obtained. represents the conditional information entropy of the i-th feature category in the first collected feature category set on various feature data of the first feature category after one elimination; S36, continuously repeating S33, S34 and S35, when the first collection feature category set When there are no feature categories left, stop repeating the operation and get the final decision tree; S37, using the reduced boat sample test image feature data set matrix h′ to test the final decision tree.

5. The method for fully automatic boat precision control based on artificial intelligence according to claim 4, characterized in that: The S33 comprises the following steps: S331: Based on the first collection feature category set The acquisition feature subcategory matrix m′ is obtained as follows: Among them, m i ' j represents the jth feature subcategory in the i-th feature category; Represents the total number of feature subcategories in the i-th feature category; S332, traverse the acquisition feature subcategory matrix m' row by row, and obtain the acquisition feature subcategory probability data matrix according to the restored boat sample training image feature data set matrix h and the acquisition feature subcategory matrix m'. as follows, in, represents the probability of the jth feature subcategory in the ith feature category appearing in the boat sample training image feature data set matrix h after the reduction; At the same time, the probability of occurrence of various image label data in the first image label data is calculated under the condition that the feature subclass in the acquisition feature subclass matrix m′ appears, and the first conditional probability matrix n is obtained; as follows, Among them, n ij represents the probability of occurrence of various image label data in the first image label data when the jth feature subcategory in the i-th feature category appears; S333: Obtain the first conditional information entropy set according to the first conditional probability matrix n and the collected feature subcategory probability data matrix m~ The calculation formula is as follows, 6. The method for fully automatic boat precision control based on artificial intelligence according to claim 4, characterized in that: The S37 comprises the following steps: S371, using the final decision tree to classify the reduced boat sample test image feature data set matrix h′ to obtain a test classification result label data matrix S372, set the accuracy threshold to o; the test classification result label data matrix Compare with the second image label data matrix and calculate the test classification result label data matrix The data that is the same as the data at the corresponding position in the second image label data matrix accounts for the test classification result label data matrix The ratio of all data in the test is used to obtain the test accuracy; S373. When the test accuracy is greater than or equal to 0, the final decision tree meets the requirements; Otherwise, repeat S31 to S36, regenerate a new final decision tree, and perform classification operations on the restored boat sample test image feature data set matrix h′ until the test accuracy is greater than or equal to o, then stop repeating.

7. The method for fully automatic boat precision control based on artificial intelligence according to claim 1, characterized in that: The S5 comprises the following steps: S51, using the final decision tree to analyze the first boat sample image feature data set matrix Perform a classification operation to obtain a first classification result; S52, judging whether there are cracks on the surface of the boat according to the first classification result; and then classifying the optimal sample image feature data set according to the optimal sample image feature data set. Determine whether the toughness and strength of the boat are reduced; S53. The classification sets without cracks and the ones whose toughness and strength of the boat are not reduced are retained; otherwise, they are discarded.

8. A system for implementing the fully automatic boat precision control method based on artificial intelligence as claimed in any one of claims 1 to 7, characterized in that: It includes a material boat sample collection module, a material boat restored image feature collection module, a material boat corroded image feature collection module, a material boat restored image feature classification module, a material boat image feature classification module and a material boat sample rejection module; The boat sample collection module is used to set a boat component ratio data matrix and obtain a boat sample matrix to be measured according to the boat component ratio data matrix; The boat restored image feature acquisition module is used to extract features from the boat sample training image and test image using a convolutional neural network; The boat corrosion image feature acquisition module is used to extract features from the first boat sample image matrix; The boat image feature classification module is used to construct a decision tree and a lion group population to perform classification operations on the first boat sample image feature data set matrix and the corroded boat sample image feature data set respectively; The boat sample elimination module is used to eliminate samples in the restored boat sample matrix according to the first classification result and the optimal sample image feature data set classification set.

Images