CNN-RIME-SVM-based small sample Isa furnace spray gun fault identification method and system

Abstract

The invention discloses a small sample Isa furnace spray gun fault identification method and system based on CNN-RIME-SVM, and the method comprises the steps: collecting related technological parameter data affecting the service life of an Isa furnace spray gun in the smelting of an Isa copper molten pool, and constructing an Isa furnace spray gun data set; filling missing values in each sample in the Isa furnace spray gun data set; normalizing the filled Isa furnace spray gun data set, and dividing the normalized Isa furnace spray gun data set into training set data and test set data according to a preset proportion; building a small sample Isa furnace spray gun fault identification model of CNN-RIME-SVM, adjusting hyper-parameters, and training the small sample Isa furnace spray gun fault identification model of CNN-RIME-SVM by using the training set data; and inputting test set data into the trained small sample Isa furnace spray gun fault identification model of the CNN-RIME-SVM, and verifying the effectiveness of the trained small sample Isa furnace spray gun fault identification model of the CNN-RIME-SVM according to analysis and evaluation between the prediction category and the real category of the Isa furnace spray gun. The method can be effectively used for identifying the spray gun fault of the Isa furnace.

Description

Technical Field

[0001] The invention relates to a small sample Isa furnace spray gun fault identification method and system based on CNN-RIME-SVM, belonging to the technical field of Isa furnace spray gun fault detection. Background Art

[0002] As an efficient top-blown submerged molten pool smelting technology, the Isa copper smelting process uses a lance to top-blow oxygen-enriched air to enhance melt mixing, accelerate chemical reactions, and achieve efficient smelting by utilizing the heat released by oxidation of iron and sulfur in copper concentrate. This method has the advantages of low construction investment, high productivity, wide raw material adaptability, and high degree of automation, and is widely used in copper smelting processes.

[0003] During the ISA copper bath smelting process, the ISA furnace spray gun, as one of the core equipment, is not only responsible for delivering oxygen-enriched air, but also for vigorously stirring the slag layer to enhance the efficiency of mass transfer and heat transfer during the smelting process. The failure of the ISA furnace spray gun will seriously affect the operation rate of the ISA furnace. Replacing the spray gun will not only increase the downtime, but also incur higher maintenance costs. Its stable operation is crucial to ensure product quality. Therefore, reducing the occurrence of failures has become one of the core factors in improving the operation rate of copper bath smelting, and accurately identifying the failure of the ISA furnace spray gun has become particularly critical.

[0004] There is currently a lack of better solutions to the above problems. Summary of the invention

[0005] The present invention provides a small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM, so as to at least solve the problem of small sample Isa furnace spray gun fault identification.

[0006] The technical solution of the present invention is:

[0007] According to a first aspect of the present invention, a small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM is provided, comprising: S1, collecting relevant process parameter data that affects the life of the Isa furnace spray gun in Isa copper molten pool smelting, and constructing an Isa furnace spray gun data set; S2, filling the missing values ​​in each sample in the Isa furnace spray gun data set; S3, normalizing the filled Isa furnace spray gun data set, and dividing the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; S4, building a small sample of CNN-RIME-SVM Isa furnace spray gun fault recognition model, adjust the hyperparameters, use the training set data to train the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, and obtain the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model; S5, input the test set data into the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model, and analyze and evaluate the predicted category and the true category of the Isa furnace spray gun to verify the effectiveness of the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model.

[0008] Furthermore, the relevant process parameter data affecting the life of the Isa furnace spray gun in the Isa copper molten pool smelting includes normal category data and fault category data of the Isa furnace spray gun.

[0009] Furthermore, the missing values ​​in each sample in the Isa furnace spray gun data set are filled, specifically: for a missing value, a predicted value is obtained according to a filling formula to fill the missing value; the filling formula is: in, is the predicted value of the missing value of the i-th sample, N i is the set of K neighbors closest to the i-th sample, y j is the true target value of the jth neighbor; |N i | is the set N i The size of , that is, the number of the nearest K neighbors.

[0010] Furthermore, using The maximum and minimum normalization processing is performed on the process parameter data in the filled Isa furnace spray gun data set; where X norm is the normalized value of the maximum and minimum values; X is the actual sequence data of the current process parameters in the Isa furnace spray gun data set, X min , X max are the minimum and maximum values ​​of each process parameter in the Isa furnace spray gun data set, respectively.

[0011] Furthermore, the CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model includes an improved CNN model and a RIME-SVM module, specifically:

[0012] The improved CNN model uses the traditional CNN model as the framework, and uses four convolution modules connected in sequence to adaptively extract the fault features of the Isa furnace spray gun. By replacing the fully connected layer of the traditional CNN model with a global average pooling layer, the end-to-end fault feature extraction of the Isa furnace spray gun is realized as the output of the improved CNN model. The basic parameters are as follows: the four convolution modules all include convolution layers and maximum pooling layers, and the ELU activation function is used to replace the traditional ReLU activation function. The dimensional changes of the four convolution layers are set to 1*64, 64*128, 128*64, and 64*32; the parameter α in the ELU activation function is 1.0; the loss function selects the cross entropy loss function, and the optimizer selects Adam.

[0013] The RIME-SVM module is used to optimize the regularization parameter λ and kernel parameter σ in the support vector machine SVM within a preset range based on the determined objective function using the frost optimization algorithm; the output of the improved CNN model is used as the input of the optimized support vector machine SVM, and the Isa furnace spray gun category is identified through the optimized support vector machine SVM.

[0014] Furthermore, maximizing the F1 score is used as the objective function, and the preset ranges of the regularization parameter λ and the kernel parameter σ are determined to be [0.1, 100].

[0015] Furthermore, accuracy A, precision P, recall R and F1 score are used as evaluation indicators.

[0016] According to a second aspect of the present invention, a small sample Isa furnace spray gun fault identification system based on CNN-RIME-SVM is provided, comprising: an acquisition module, used to collect relevant process parameter data affecting the life of the Isa furnace spray gun in the Isa copper molten pool smelting, and construct an Isa furnace spray gun data set; a filling module, used to fill the missing values ​​in each sample in the Isa furnace spray gun data set; a processing module, used to normalize the filled Isa furnace spray gun data set, and divide the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; a training module, used to build a CNN-RIME-SVM. A small sample Isa furnace spray gun fault recognition model of VM is constructed, and hyperparameters are adjusted. The small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM is trained using the training set data to obtain the trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM; a verification module is used to input the test set data into the trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, and analyze and evaluate the predicted category and the true category of the Isa furnace spray gun to verify the effectiveness of the trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM.

[0017] According to a third aspect of the present invention, there is provided a terminal comprising a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the processor is configured to execute any one of the above-described small sample Isa furnace spray gun fault identification methods based on CNN-RIME-SVM.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium comprising a stored program, characterized in that when the program is executed by a processor, the processor implements any one of the above-mentioned small sample Isa furnace spray gun fault identification methods based on CNN-RIME-SVM.

[0019] The beneficial effects of the present invention are:

[0020] 1. The present invention provides a small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM. By using the K nearest neighbor method to fill missing values ​​in the collected data affecting the life of the Isa furnace spray gun, the rationality of the data and the integrity of the structure are maintained; by using the maximum and minimum normalization processing, the data is converted to between 0 and 1, the influence of the dimension is eliminated, and the computing power is improved.

[0021] 2. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM of the present invention, in the improved CNN fault feature extraction model, adopts ELU activation function instead of ReLU activation function to improve the robustness to Isa furnace spray gun data noise and input changes, and accelerate the convergence of the fault feature extraction model; adopts global average pooling layer instead of fully connected layer can not only enhance the correlation between Isa furnace spray gun process parameters and fault categories, but also reduce model parameters, avoid overfitting, and better realize the fault feature extraction of Isa furnace spray gun.

[0022] 3. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM of the present invention adopts SVM as the final fault identification model, and optimizes the kernel function parameters and penalty factor of SVM through RIME, thereby improving the classification accuracy of the Isa furnace spray gun fault identification model, and providing a new method for the Isa furnace spray gun fault identification model.

[0023] In summary, the present invention can be effectively used for Isa furnace spray gun fault identification. At the same time, the method solves the problem of low accuracy in Isa furnace spray gun fault identification, thereby achieving the effect of improving the accuracy in Isa furnace spray gun fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of the method of the present invention;

[0025] Figure 2 This is a structural diagram of the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM of the present invention;

[0026] Figure 3 The flowchart of optimizing the kernel function parameters and penalty factors of SVM by RIME of the present invention;

[0027] Figure 4 Partial data of relevant parameters affecting the life of the ISA furnace spray gun according to an embodiment of the present invention;

[0028] Figure 5 A comparison diagram of predicted categories and actual categories of fault identification of an Isa furnace spray gun according to an embodiment of the present invention;

[0029] Figure 6 This is a comparison chart of predicted categories and actual categories in a comparative experiment of an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other arbitrarily without conflict.

[0031] Example 1: Figure 1-6 As shown in the figure, a small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM includes: S1, collecting relevant process parameter data that affect the life of the Isa furnace spray gun in Isa copper molten pool smelting, and constructing an Isa furnace spray gun data set; S2, filling the missing values ​​in each sample in the Isa furnace spray gun data set; S3, normalizing the filled Isa furnace spray gun data set, and dividing the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; S4, building a small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM Fault recognition model, adjust hyperparameters, use training set data to train CNN-RIME-SVM's small sample Isa furnace spray gun fault recognition model, and obtain the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model; S5, input the test set data into the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model, and analyze and evaluate the predicted category and the real category of the Isa furnace spray gun to verify the effectiveness of the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model. Furthermore, the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model can be used to perform fault recognition on the Isa furnace spray gun data to be identified.

[0032] Furthermore, the relevant process parameter data affecting the life of the Isa furnace spray gun in the Isa copper molten pool smelting includes normal category data and fault category data of the Isa furnace spray gun.

[0033] Furthermore, the missing values ​​in each sample in the Isa furnace spray gun data set are filled, specifically: for a missing value, a predicted value is obtained according to a filling formula to fill the missing value; the filling formula is: in, is the predicted value of the missing value of the i-th sample, N i is the set of K neighbors closest to the i-th sample, y j is the true target value of the jth neighbor; |N i | is the set N i The size of , that is, the number of the nearest K neighbors.

[0034] Furthermore, using The maximum and minimum normalization processing is performed on the process parameter data in the filled Isa furnace spray gun data set; where X norm is the normalized value of the maximum and minimum values; X is the actual sequence data of the current process parameters in the Isa furnace spray gun data set, X min , X max are the minimum and maximum values ​​of each process parameter in the Isa furnace spray gun data set, respectively.

[0035] Furthermore, the CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model includes an improved CNN model and a RIME-SVM module, specifically:

[0036] The improved CNN model uses the traditional CNN model as the framework, and uses four convolution modules connected in sequence to adaptively extract the fault features of the Isa furnace spray gun. By replacing the fully connected layer of the traditional CNN model with a global average pooling layer, the fault feature extraction model parameters can be reduced, overfitting can be avoided, the correlation between the process parameters of the Isa furnace spray gun and the fault category can be enhanced, and the end-to-end fault feature extraction of the Isa furnace spray gun can be realized as the output of the improved CNN model. The basic parameters are as follows: the four convolution modules all include convolution layers and maximum pooling layers, and the ELU activation function is used to replace the traditional ReLU activation function to improve the adaptability to the data noise and input changes of the Isa furnace spray gun, while accelerating the convergence of the fault feature extraction model. The dimensional changes of the four convolution layers are set to 1*64, 64*128, 128*64, and 64*32; the parameter α in the ELU activation function is 1.0; the batch-size is set to 32, the learning rate is set to 0.0001, the number of iterations is set to 100 times, the loss function selects the cross entropy loss function, and the optimizer selects Adam;

[0037] The RIME-SVM module specifically comprises: according to the determined objective function, within a preset range, using the frost optimization algorithm to optimize the regularization parameter λ and the kernel parameter σ in the support vector machine SVM to improve the classification accuracy; in the frost optimization algorithm, the number of frost groups is set to 20, and the maximum number of iterations is set to 100; the output of the improved CNN model is used as the input of the optimized support vector machine SVM, and the Isa furnace spray gun category is identified through the optimized support vector machine SVM.

[0038] Furthermore, maximizing the F1 score is used as the objective function, and the preset ranges of the regularization parameter λ and the kernel parameter σ are determined to be [0.1, 100].

[0039] In the above, the frost optimization algorithm is introduced as follows:

[0040] The Rime Ice Optimization Algorithm (RIME) is a meta-heuristic optimization algorithm based on frost growth behavior proposed by SuHang et al. in 2023. The algorithm has the advantages of fast convergence and high precision. By dynamically adjusting the search position, it improves the global search capability and avoids falling into the local optimum. It mainly includes three stages: soft frost search strategy, hard frost puncture mechanism, and greedy selection mechanism.

[0041] (1) Soft frost search strategy: By simulating the movement of soft frost particles, random exploration is achieved in a large range, which helps to quickly find potential optimal solution areas while reducing the risk of falling into local optimality. The position of frost particles can be calculated using formula (1).

[0042]

[0043]

[0044] In the formula, is the new position of the updated frost particle, i and j represent the jth particle in the i-th frost crystal, R best,j is the jth particle of the best frost crystal in the frost population R, and the parameter r1 is a random number in the range of (-1,1); cosθ varies with the number of iterations, as shown in formula (2), where t is the current number of iterations and T2 is the maximum number of iterations of the algorithm; β is the environmental factor, as shown in formula (3), where the value of w defaults to 5, and [] indicates rounding operation; h represents the adhesion, which is a random number in the range of (0,1); Ub ij and Lb ij are the upper and lower bounds of the space respectively; E is the adhesion coefficient, which increases with the number of iterations, as shown in formula (4). r2 is a random number in the range of (0,1), which together with E controls whether the particle position is updated.

[0045] (2) Hard frost penetration mechanism: By simulating the cross-behavior between hard frost particles, information exchange between dimensions is realized, further improving the accuracy of the solution. The replacement formula between frost particles is shown in Equation (5).

[0046]

[0047] Where r3 is a random number in the range of (-1,1); S i F stands for frost crystals. normr (S i ) represents the normalized value of the current frost crystal fitness value.

[0048] (3) Greedy selection mechanism: Compare the fitness values ​​of the frost crystal before and after the update. If the updated value is better, replace the value before the update.

[0049] Furthermore, accuracy A, precision P, recall R and F1 score are used as evaluation indicators.

[0050] The present embodiment is described below by means of specific examples.

[0051] The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM of the present invention comprises the following steps:

[0052] Step (1): Through the data interface established with the DCS system, the relevant parameter data affecting the life of the Isa furnace spray gun were successfully collected in the Isa copper molten pool smelting production workshop, including the normal and fault data of the Isa furnace spray gun. These data mainly come from various sensors, including PT1211 (spray gun oxygen supply pipe pressure), FT_1213 (spray gun oxygen supply pipe oxygen flow), TE1224 (spray gun air supply pipe air temperature), TE1214 (spray gun oxygen supply pipe oxygen temperature), FT_1223 (spray gun air supply pipe air flow), PT1222M (spray gun air supply pipe pressure), FIC1223.MV (flow regulation of spray gun air supply pipe), PIC1222.MV (pressure regulation of air supply pipe evacuation), FT_1712 (flow of fan outlet main pipe), PT1239 (purge pressure of spray gun oil pipe), HZ9 (spray gun load), PT1229 (spray gun back pressure), PT1721 (fuel system outlet pressure), PIC1721.MV (pressure regulation of return oil main pipe), FIC1233A.SV (spray gun oil supply regulation), PQWZ (position of spray gun from furnace bottom), P_AI_B_1 (spray gun feedback signal), PT1249 (spray gun top pressure) 18 process parameter data. After long-term practice and observation, there are three main types of faults in Isa furnace spray guns: burnout of Isa furnace spray gun end, burnout of Isa furnace spray gun end pipe wall and burnout of cyclone. For fault identification, these three fault categories were selected for further study and were labeled as 1, 2, and 3, and the normal category data was labeled as 0. The normal category data and 200 data of each fault category were extracted, totaling 800 samples. Some of the collected data have been Figure 4 Displayed in.

[0053] Step (2): Fill the missing values ​​in each sample of the collected Isa furnace spray gun data: For the missing value of a feature, the basic formula for filling is: Where: is the predicted value after filling the missing value of the i-th sample, N i is the set of K neighbors closest to the i-th sample, y j is the true target value of the jth neighbor, |N i | is the set N iThe size of is the number of the nearest K neighbors. In the embodiment of the present invention, the number of the nearest K neighbors is set to 4.

[0054] Step (3): Use the formula The maximum and minimum normalization processing is performed on the process parameter data in the filled Isa furnace spray gun data set; where X norm is the normalized value of the maximum and minimum values; X is the actual sequence data of the current process parameters in the Isa furnace spray gun data set, X min , X max They are the minimum and maximum values ​​of each process parameter in the Isa furnace spray gun data set. Further, 70% of the normalized data is used to train the model, and the remaining 30% is used as a test set.

[0055] Step (4): Build a small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, adjust the hyperparameters, use the training set data to train the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, and obtain the trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM; the dimensional changes of the four convolutional layers in the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM are set to 1*64, 64*128, 128*64, 64*32; the parameter α in the ELU activation function is 1.0; batch-size is set to 32 , the learning rate is set to 0.0001, the number of iterations is set to 100 times, the cross entropy loss function is selected as the loss function, and the optimizer is selected as Adam; according to the determined objective function, within the preset range, the frost optimization algorithm RIME is used to optimize the regularization parameter λ and the kernel parameter σ in the support vector machine SVM to improve the classification accuracy; in the frost optimization algorithm, the number of frost groups is set to 20, and the maximum number of iterations is set to 100; RIME is used to optimize the parameters of SVM, and the best regularization parameter is 7.4510 and the kernel parameter is 0.2653; finally, the optimized SVM model is used to output the prediction category of the Isa furnace spray gun.

[0056] Furthermore, the specific steps of RIME to optimize SVM parameters are as follows:

[0057] Step 1: Set the initial range of SVM parameter combinations and randomly generate the initial frost population.

[0058] Step 2: Calculate the fitness of each frost ice crystal and select the frost ice crystal with the best fitness.

[0059] Step 3: Determine the adhesion coefficient E, a key parameter that controls the search process in the RIME algorithm. When r2 is lower than E, update the parameter combination through the soft frost search strategy; when r2 reaches or exceeds E, perform cross-update through the hard frost puncture mechanism.

[0060] Step 4: Use the positive greedy selection mechanism to iteratively search for the best solution, replace the suboptimal solution, and select the optimal solution to improve the quality of the frost group until the termination condition is met.

[0061] Step 5: Output the parameter combination corresponding to the frost ice crystal with the best fitness (i.e., the optimal parameters λ and σ) for building the SVM fault recognition model.

[0062] Step (5): Input the test set data into the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model for recognition, verify the effectiveness of this Isa furnace spray gun fault recognition model, and design a comparative test to verify the effectiveness of this fault recognition model; by using accuracy rate, precision rate, recall rate and F1 score as evaluation indicators, the four parameters are calculated as follows:

[0063]

[0064] Where: TP represents the number of samples correctly identified as a certain category; TN represents the number of samples correctly identified as other categories; FP represents the number of samples incorrectly identified as other categories; FN represents the number of samples incorrectly identified as a certain category. The closer the values ​​of accuracy rate, precision rate, recall rate and F1 score are to 100%, the better the fault identification effect of the CNN-RIME-SVM model.

[0065] In this embodiment, the accuracy rate, precision rate, recall rate and F1 score obtained by the CNN-RIME-SVM fault recognition model. The results are shown in Table 1. As can be seen from the table, the accuracy rate is 97.08%, the precision rate is 97.08%, the recall rate is 97.10%, and the F1-score is 97.07%, and the values ​​of the four evaluation indicators are all close to 1. The results show that the small sample Isa furnace spray gun fault recognition model based on CNN-RIME-SVM of the present invention has a very good recognition effect on the Isa furnace spray gun fault category. Further, CNN, RIME-SVM, CNN-SVM and the present invention are selected for comparative experiments to verify the recognition effect of the fault recognition model (CNN represents the traditional CNN model and uses softmax as the classification function, RIME-SVM represents the use of the RIME-SVM module of the present invention without considering CNN; CNN-SVM represents the use of SVM for category recognition under the traditional CNN), and the comparison results are shown in Table 2.

[0066] Table 1

[0067] Model Accuracy Accuracy Recall <![CDATA[F1 score]]> CNN-RIME-SVM 97.08％ 97.08％ 97.10％ 97.07％

[0068] Table 2

[0069] Model Accuracy Accuracy Recall <![CDATA[F1 score]]> CNN 84.58％ 90.46％ 84.58％ 82.96％ RIME-SVM 86.25％ 86.25％ 91.29％ 85.13％ CNN-SVM 87.08％ 87.08％ 91.48％ 86.16％ CNN-RIME-SVM 97.08％ 97.08％ 97.10％ 97.07％

[0070] The predicted categories of the Isa furnace spray guns obtained by the small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM of the present invention are compared and analyzed with the actual categories. The results are as follows: Figure 5 As shown in the figure, the comparison between the predicted category and the real category of the comparative experiment is as follows Figure 6 As shown. Figure 5 and Figure 6 It can be further proved that the fault identification method of the Isa furnace spray gun of the present invention has a high accuracy rate.

[0071] Example 2: A small sample Isa furnace spray gun fault identification system based on CNN-RIME-SVM, comprising: an acquisition module, used to collect relevant process parameter data affecting the life of the Isa furnace spray gun in the Isa copper molten pool smelting, and construct an Isa furnace spray gun data set; a filling module, used to fill the missing values ​​in each sample in the Isa furnace spray gun data set; a processing module, used to normalize the filled Isa furnace spray gun data set, and divide the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; a training module, used to build a small sample of CNN-RIME-SVM The Isa furnace spray gun fault recognition model adjusts the hyperparameters, uses the training set data to train the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, and obtains the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model; the verification module is used to input the test set data into the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model, and analyze and evaluate the predicted category and the real category of the Isa furnace spray gun to verify the effectiveness of the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model. The system is used to implement the above-mentioned embodiments and preferred implementations, and those that have been explained will not be repeated. The term "module" used above can implement a combination of software and / or hardware for a predetermined function. Although the system described in the above embodiment is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived. For the parts that are not described in detail in each module, please refer to the relevant description of other embodiments.

[0072] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0073] Embodiment 3: A terminal comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to execute the small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM as described in any one of Embodiments 1.

[0074] In an exemplary embodiment, the terminal may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0075] Embodiment 4: A computer-readable storage medium, comprising a stored program, which, when executed by a processor, enables the processor to implement the small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM as described in any one of Embodiment 1.

[0076] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0077] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, 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 the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0078] The specific implementation modes of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above implementation modes, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM, characterized in that: include: S1. Collect relevant process parameter data that affect the life of the Isa furnace spray gun in Isa copper molten pool smelting, and construct an Isa furnace spray gun data set; S2, fill in the missing values ​​in each sample in the Isa furnace spray gun dataset; S3, normalizing the padded Isa furnace spray gun data set, and dividing the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; S4. Build a small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, adjust hyperparameters, and use the training set data to train the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM to obtain a trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM; S5. Input the test set data into the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model, and analyze and evaluate the predicted category and the true category of the Isa furnace spray gun to verify the effectiveness of the trained CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model.

2. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 1 is characterized in that: The relevant process parameter data affecting the life of the Isa furnace spray gun in the Isa copper molten pool smelting includes normal category data and fault category data of the Isa furnace spray gun.

3. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 1 is characterized in that: The missing values ​​in each sample in the Isa furnace spray gun data set are filled, specifically: for a missing value, a predicted value is obtained according to a filling formula to fill the missing value; the filling formula is: in, is the predicted value of the missing value of the i-th sample, N i is the set of K neighbors closest to the i-th sample, y j is the true target value of the jth neighbor; |N i | is the set N i The size of , that is, the number of the nearest K neighbors.

4. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 1 is characterized in that: Utilization The maximum and minimum normalization processing is performed on the process parameter data in the filled Isa furnace spray gun data set; where X norm is the normalized value of the maximum and minimum values; X is the actual sequence data of the current process parameters in the Isa furnace spray gun data set, X min , X max are the minimum and maximum values ​​of each process parameter in the Isa furnace spray gun data set, respectively.

5. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 1 is characterized in that: The CNN-RIME-SVM small sample Isa furnace spray gun fault recognition model includes an improved CNN model and a RIME-SVM module, specifically: The improved CNN model uses the traditional CNN model as the framework, and uses four convolution modules connected in sequence to adaptively extract the fault features of the Isa furnace spray gun. By replacing the fully connected layer of the traditional CNN model with a global average pooling layer, the end-to-end fault feature extraction of the Isa furnace spray gun is realized as the output of the improved CNN model. The basic parameters are as follows: the four convolution modules all include convolution layers and maximum pooling layers, and the ELU activation function is used to replace the traditional ReLU activation function. The dimensional changes of the four convolution layers are set to 1*64, 64*128, 128*64, and 64*32; the parameter α in the ELU activation function is 1.0; the loss function selects the cross entropy loss function, and the optimizer selects Adam. The RIME-SVM module is used to optimize the regularization parameter λ and kernel parameter σ in the support vector machine SVM within a preset range based on the determined objective function using the frost optimization algorithm; the output of the improved CNN model is used as the input of the optimized support vector machine SVM, and the Isa furnace spray gun category is identified through the optimized support vector machine SVM.

6. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 5 is characterized in that: Maximizing the F1 score is used as the objective function, and the preset ranges of the regularization parameter λ and the kernel parameter σ are determined to be [0.1, 100].

7. The small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM according to claim 1 is characterized in that: The accuracy A, precision P, recall R and F1 score are used as evaluation indicators.

8. A small sample Isa furnace spray gun fault identification system based on CNN-RIME-SVM, characterized in that: include: The acquisition module is used to collect relevant process parameter data that affect the life of the Isa furnace spray gun in the Isa copper molten pool smelting, and to build an Isa furnace spray gun data set; The imputation module is used to fill in the missing values ​​in each sample in the Isa furnace spray gun dataset; A processing module is used to normalize the filled Isa furnace spray gun data set, and divide the normalized Isa furnace spray gun data set into training set data and test set data according to a preset ratio; A training module is used to build a small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM, adjust hyperparameters, and use the training set data to train the small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM to obtain a trained small sample Isa furnace spray gun fault recognition model of CNN-RIME-SVM; The verification module is used to input the test set data into the trained CNN-RIME-SVM small sample Iissa furnace spray gun fault recognition model, and verify the effectiveness of the trained CNN-RIME-SVM small sample Iissa furnace spray gun fault recognition model based on the analysis and evaluation between the predicted category and the true category of the Iissa furnace spray gun.

9. A terminal, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the processor is configured to execute the small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM as described in any one of claims 1 to 7.

10. A computer-readable storage medium, the computer-readable storage medium comprising a stored program, characterized in that: When the program is executed by a processor, the processor implements the small sample Isa furnace spray gun fault identification method based on CNN-RIME-SVM as described in any one of claims 1 to 7.

Images