Foam flotation working condition identification method based on undersampling and related equipment

Abstract

The invention provides an undersampling-based froth flotation working condition identification method and related equipment. The method comprises the following steps: acquiring historical froth image samples under normal working conditions as majority class samples and historical froth image samples under abnormal working conditions as minority class samples; inputting the majority of samples into the constructed multi-target optimization under-sampling model, and solving the multi-target optimization under-sampling model through a binary multi-target state transition algorithm to obtain an optimal Pareto solution set; for each undersampling result in the optimal Pareto solution set, merging the undersampling results with minority class samples to obtain a training set to train the base classifiers, and obtaining a working condition identification model integrated by a plurality of base classifiers; the foam image data collected in real time in the industrial production process are input into the working condition recognition model for working condition recognition, and a working condition recognition result in the industrial production process is obtained; and the recognition precision and the classification confidence coefficient of minority class abnormal working conditions are remarkably improved while the high overall recognition rate is kept.

Description

Technical Field

[0001] This invention relates to the field of industrial process anomaly identification technology, and in particular to a method and related equipment for identifying foam flotation conditions based on undersampling. Background Technology

[0002] Foam flotation is a crucial step in non-ferrous metal beneficiation, and its abnormal operating conditions directly impact production indicators and energy consumption. Abnormal operating conditions include overflowing of the flotation tank, settling tank malfunction, and slurry churn. Traditional methods for identifying abnormal operating conditions rely on visual inspection by operators, which suffers from strong subjectivity and delayed response.

[0003] In recent years, machine vision-based methods have been introduced to achieve automatic identification of working conditions by analyzing the features of foam images (such as color, size, texture, and speed). Machine vision-based working condition identification can be regarded as a typical image classification problem. By extracting multidimensional features of foam images and using classification models to distinguish different working conditions, intelligent identification of production status can be achieved.

[0004] However, building effective classification models from real-world industrial data faces two core challenges:

[0005] 1. In actual production, the frequency of abnormal working conditions is much lower than that of normal working conditions, resulting in a significant imbalance in the number of collected foam image samples. As a result, the classification model tends to optimize the recognition performance of normal working conditions during training, while ignoring the feature learning of abnormal working conditions. This results in a high overall accuracy of the model, but a low recognition rate of abnormal working conditions, leading to class imbalance.

[0006] 2. In actual production, the appearance of foam is affected by various factors such as the properties of slurry, the amount of reagent added, lighting conditions and shooting angle. Even under different working conditions, their visual characteristics may be similar. Conversely, the image features of the same working condition under different conditions may also have significant differences. This complexity leads to the blurring and intertwining of the distribution boundaries of samples from different working conditions in the feature space, which reduces the discrimination confidence of the classifier in these areas, makes it easy to make misjudgments, and causes class overlap problems.

[0007] Moreover, the two situations mentioned above often coexist and exacerbate each other, further deteriorating the classification performance.

[0008] To address the above problems, existing technical solutions mainly fall into two categories:

[0009] 1. Imbalance can be alleviated by adjusting the sample distribution of the training set, but existing sample collection methods are mostly random strategies, which cannot systematically balance the conflicting goals of "preserving normal working condition information" and "performing abnormal working condition identification", and generally do not consider the classification uncertainty caused by class overlap.

[0010] 2. Modifying the classifier itself to improve attention to abnormal operating conditions can alleviate the performance bias caused by imbalance to some extent, but it generally fails to effectively deal with the decision uncertainty caused by class overlap, and the model structure is complex, the parameters are sensitive, and the generalization ability is limited.

[0011] In summary, the closest existing technology has failed to achieve coordinated processing of class imbalance and class overlap, and lacks the quantification and utilization of uncertainty in the classification decision process, resulting in insufficient accuracy, robustness and reliability in identifying complex industrial data. Summary of the Invention

[0012] This invention provides a method and related equipment for identifying foam flotation conditions based on undersampling. The purpose is to significantly improve the identification accuracy and classification confidence of a few abnormal conditions while maintaining a high overall recognition rate.

[0013] To achieve the above objectives, the present invention provides a method for identifying foam flotation conditions based on undersampling, comprising:

[0014] Step 1: Collect historical foam image samples from the industrial production process, and use historical foam image samples collected under normal operating conditions as the majority class samples and historical foam image samples collected under abnormal operating conditions as the minority class samples.

[0015] Step 2: Input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model through the binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results;

[0016] Step 3: For each undersampling result in the optimal Pareto solution set, merge the undersampling result with the minority class sample to obtain the training set, and use the training set to train the base classifier to obtain the working condition recognition model integrated by multiple base classifiers.

[0017] Step 4: Input the foam image data collected in real time during the industrial production process into the working condition recognition model for working condition recognition, and obtain the working condition recognition results during the industrial production process.

[0018] Furthermore, the objective of optimizing the undersampling model is:

[0019]

[0020] in, Denotes the first objective function. Indicates the number of minority class samples. Indicates the number of samples in the majority class. undersampling scheme The Middle The predicted label vector of each sample. , undersampling scheme The Middle The true label vector of each sample undersampling scheme The Middle The overall quality value of each sample in the dataset. Describes the second objective function. Indicates the first One sample, , .

[0021] Furthermore, by solving the multi-objective optimization undersampling model using a binary multi-objective state transition algorithm, the optimal Pareto solution set is obtained, including:

[0022] Step 21: Randomly generate the initial population;

[0023] Step 22: Apply four state transition operators to each candidate solution in the current population to perform state transition, forming a set of offspring candidate solutions. The set of offspring candidate solutions includes multiple offspring candidate solutions.

[0024] Step 23: Merge the parent population with the offspring candidate solutions to obtain a joint population, which includes multiple individual solutions. Calculate the first and second objective values ​​of the undersampling scheme corresponding to each individual solution in the joint population.

[0025] Step 24: Sort the individual solutions in the joint population using the first objective value and the second objective value to obtain the Pareto solution set;

[0026] Step 25: Repeat steps 22-24 until the maximum number of iterations is reached, and obtain the optimal Pareto solution set.

[0027] Furthermore, the real-time foam image data collected during industrial production is input into the working condition recognition model for working condition identification, resulting in the working condition identification results for the industrial production process, including:

[0028] The foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in multiple evidence bodies;

[0029] By combining all the pieces of evidence using Dempster's combination rule, we obtain the complete piece of evidence.

[0030] The overall evidence body is mapped to a probability distribution using the Pignistic probability transformation, and the probability distribution is processed based on the principle of maximizing the posterior probability to obtain the working condition identification results in the industrial production process.

[0031] Furthermore, by inputting the real-time foam image data collected during industrial production into the working condition recognition model for working condition identification, the expressions for multiple evidence bodies are obtained as follows:

[0032]

[0033] in, Indicates the first The evidence body output by each base classifier Indicates the first Each base classifier evaluates the samples Category Confidence level, , Indicates the number of categories. This represents uncertain quality that is not assigned to any specific category.

[0034] Furthermore, by using Dempster's combination rule to fuse all the evidence, the expression for the overall evidence is:

[0035]

[0036]

[0037] in, This indicates the evidence body after fusion. This indicates the evidence body after fusion, including the sample. Category Confidence level, This indicates the combination rules.

[0038] Furthermore, the expression for mapping the entire body of evidence to a probability distribution using the Pignistic probability transformation is as follows:

[0039]

[0040] in, Indicates sample Category The probability of.

[0041] The present invention also provides a foam flotation condition identification device based on undersampling, comprising:

[0042] The acquisition module is used to collect historical foam image samples during industrial production processes. Historical foam image samples collected under normal operating conditions are used as majority class samples, and historical foam image samples collected under abnormal operating conditions are used as minority class samples.

[0043] The solution module is used to input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model through a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results.

[0044] The training module is used to combine the undersampled results with the minority class samples for each undersampled result in the optimal Pareto solution set to obtain a training set, and to use the training set to train the base classifiers to obtain a working condition recognition model integrated by multiple base classifiers.

[0045] The identification module is used to input real-time foam image data collected during industrial production into the working condition identification model for working condition identification, and obtain the working condition identification results during industrial production.

[0046] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a foam flotation condition identification method based on undersampling.

[0047] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for identifying foam flotation conditions based on undersampling.

[0048] The above-described solution of the present invention has the following beneficial effects:

[0049] This invention collects historical foam image samples from industrial production processes, using historical foam image samples collected under normal operating conditions as majority class samples and historical foam image samples collected under abnormal operating conditions as minority class samples. The majority class samples are input into a constructed multi-objective optimization undersampling model, which is then solved using a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set. For each undersampling result in the optimal Pareto solution set, the undersampling result is merged with the minority class samples to obtain a training set. This training set is then used to train base classifiers, resulting in a working condition recognition model integrated from multiple base classifiers. Real-time foam image data collected during industrial production is input into the working condition recognition model for working condition identification, yielding the working condition identification results for the industrial production process. Compared with existing technologies, this invention optimizes both the conflicting objectives of "improving minority class recognition performance" and "preserving representative information of the majority class" through a multi-objective optimization undersampling model, fundamentally addressing class imbalance and class overlap issues. While maintaining a high overall recognition rate, it significantly improves the recognition accuracy and classification confidence for minority class abnormal working conditions.

[0050] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the structure of the foam flotation condition identification device in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of the structure of the terminal device in an embodiment of the present invention. Detailed Implementation

[0054] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] In the description of this invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0056] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0057] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0058] This invention addresses existing problems by providing a method and related equipment for identifying foam flotation conditions based on undersampling.

[0059] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying foam flotation conditions based on undersampling, including:

[0060] Step 1: Collect historical foam image samples from the industrial production process, and use historical foam image samples collected under normal operating conditions as the majority class samples and historical foam image samples collected under abnormal operating conditions as the minority class samples.

[0061] Step 2: Input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model through the binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results;

[0062] Step 3: For each undersampling result in the optimal Pareto solution set, merge the undersampling result with the minority class sample to obtain the training set, and use the training set to train the base classifier to obtain the working condition recognition model integrated by multiple base classifiers.

[0063] Step 4: Input the foam image data collected in real time during the industrial production process into the working condition recognition model for working condition recognition, and obtain the working condition recognition results during the industrial production process.

[0064] In this embodiment of the invention, 1007 historical foam image samples were acquired using a visual acquisition device located above the metal foam flotation unit. Historical bubble image samples The majority class samples in Minority class samples are .

[0065] This invention extracts 37-dimensional features from each historical bubble image to comprehensively characterize the apparent state of the bubble. These features mainly include:

[0066] Texture features: Statistical analysis of the co-occurrence frequency of pixel values ​​in local neighborhoods (simplified version of gray-level co-occurrence matrix), calculation of foam coarseness (based on statistical features of gray-level co-occurrence matrix), calculation of numerical non-uniformity, calculation of second moment, and calculation of high-frequency energy;

[0067] Color features: Calculate the mean saturation of the image;

[0068] Morphological characteristics: Calculate the average aspect ratio of the foam region, calculate the size distribution and average radius of the foam, and calculate the foam load-bearing capacity;

[0069] Frequency domain features: Wavelet decomposition of grayscale images, calculation of low-frequency coefficients and high-frequency coefficients on the third diagonal.

[0070] In order to select the most representative subset from the majority class samples without degrading classification performance, this invention formalizes the undersampling process of the majority class samples as a binary selection problem. Specifically, each majority class sample in the historical bubble image samples is assigned a binary decision variable to indicate whether the sample should be retained in the current undersampling scheme. When the decision variable is 1, it means that the sample is selected into the undersampled training set; when the decision variable is 0, it means that the sample is discarded. Since the number of majority class samples is usually large, the selection variables of all samples are arranged in sample index order, which forms a binary vector with a length equal to the number of majority class samples in the historical bubble image samples. This vector is then used to describe a specific undersampling scheme.

[0071] In this embodiment of the invention, the purpose of the multi-objective optimization undersampling model is to find an optimal set of vectors among all possible binary vectors, thereby selecting a majority class sample set that can balance classification performance and sample representativeness as the undersampling scheme. .

[0072] Specifically, the multi-objective optimization undersampling model is as follows:

[0073]

[0074] in, Denotes the first objective function. undersampling scheme of The norm represents the number of non-zero elements in the majority class samples. Indicates the number of minority class samples. Indicates the number of samples in the majority class. undersampling scheme The Middle The predicted label vector of each sample. , undersampling scheme The Middle The true label vector of each sample undersampling scheme The Middle The overall quality value of each sample in the dataset. Describes the second objective function. Indicates the first One sample, , .

[0075] In the multi-objective optimization undersampling model of this invention, the first objective function is used to measure the performance of the classifier under the current undersampling scheme, including classification error and uncertainty derived from evidence theory. The core idea is that if some majority class samples are located in class overlap regions or interfere with classification boundaries, they often exhibit higher error rates and uncertainty quality. Therefore, the larger the value of the first objective function, the worse the classification effect of the undersampling scheme; conversely, the smaller the value of the first objective function, the better the discriminative ability and the lower the uncertainty of the model under the undersampling scheme. By minimizing the first objective function, this invention can preferentially select majority class samples that contribute positively to the classifier. The first objective function is used to measure the number of majority class samples discarded in the undersampling scheme, aiming to maintain the representativeness of the majority class data as much as possible. The majority class samples are usually widely distributed in historical bubble image samples. If the undersampling process deletes too many samples, it may destroy the original structure of the majority class samples, thereby reducing the model's ability to characterize normal working conditions. Therefore, the value of the second objective function needs to be one of the optimization objectives, so that the algorithm can improve the performance of minority class samples while retaining the majority class information as much as possible. By minimizing the second objective function, the model can avoid losing the ability to learn the majority class sample patterns due to excessive deletion.

[0076] Specifically, the optimal Pareto solution set is obtained by solving the multi-objective optimization undersampling model using a binary multi-objective state transition algorithm, including:

[0077] Step 21, randomly generate the initial population , of which each For length is binary vector;

[0078] Step 22: Apply the four state transition operators to each candidate solution in the current population to perform state transitions, forming a set of offspring candidate solutions. The offspring candidate set includes multiple offspring candidate solutions. ;

[0079] Step 23, transfer the parent population With offspring candidate solutions Merging, resulting in a joint population , The joint population comprises multiple individual solutions, and the first objective value of the undersampling scheme corresponding to each individual solution in the joint population is calculated. Second target value ;

[0080] Step 24: Sort the individual solutions in the joint population using the first objective value and the second objective value to obtain the Pareto solution set in order to maintain the diversity of the Pareto front.

[0081] Step 25: Repeat steps 22-24 until the maximum number of iterations is reached to obtain the optimal Pareto solution set. The optimal Pareto solution set contains several non-dominated undersampling schemes, each with different trade-offs between classification performance and information preservation.

[0082] In this embodiment of the invention, each undersampling scheme in the optimal Pareto solution set is combined with all minority class samples to train multiple base classifiers, thereby obtaining multiple working condition recognition models.

[0083] In this embodiment of the invention, the entire historical foam image sample is divided into 4 categories: normal working condition, trough overflow, settling trough, and slurry churn, with the sample quantity distribution being 721, 112, 114, and 60. The samples are randomly divided into training set and test set in a ratio of 8:2. The training set is used to train the base classifier, and the test set is used to test the performance of the base classifier.

[0084] Specifically, real-time foam image data collected during industrial production is input into the working condition recognition model for working condition identification, resulting in the working condition identification results for the industrial production process, including:

[0085] Foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in multiple evidence bodies;

[0086] By combining all the pieces of evidence using Dempster's combination rule, we obtain the complete piece of evidence.

[0087] The overall evidence body is mapped to a probability distribution using the Pignistic probability transformation, and the probability distribution is processed based on the principle of maximizing the posterior probability to obtain the working condition identification results in the industrial production process.

[0088] Specifically, the foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in the following expressions for multiple evidence bodies:

[0089]

[0090] in, Indicates the first The evidence body output by each base classifier Indicates the first Each base classifier evaluates the samples Category Confidence level, , Indicates the number of categories. This represents uncertain quality that is not assigned to any specific category.

[0091] Specifically, by using Dempster's combination rule to fuse all pieces of evidence, the expression for the overall piece of evidence is as follows:

[0092]

[0093]

[0094] in, This indicates the evidence body after fusion. This indicates the evidence body after fusion, including the sample. Category Confidence level, This indicates the combination rules.

[0095] Specifically, in order to achieve probabilistic decision-making, this embodiment of the invention uses the Pignistic probability transformation to map the overall evidence body into a probability distribution, as expressed in the following expression:

[0096]

[0097] in, Indicates sample Category The probability of;

[0098] This embodiment of the invention also normalizes the probability distribution to obtain a normalized probability vector, expressed as:

[0099]

[0100] The above equation satisfies ;

[0101] The expression for processing the probability distribution based on the principle of maximizing the posterior probability is:

[0102]

[0103] in, This indicates the working condition identification result, which may be normal working condition, overflowing trough, settling trough, or slurry overflow.

[0104] This invention collects historical foam image samples from industrial production processes, using those collected under normal operating conditions as the majority class samples and those collected under abnormal operating conditions as the minority class samples. The majority class samples are input into a constructed multi-objective optimization undersampling model, which is then solved using a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set. For each undersampling result in the optimal Pareto solution set, the undersampling result is merged with the minority class samples to obtain a training set. This training set is then used to train base classifiers, resulting in a working condition recognition model integrated from multiple base classifiers. Real-time foam image data collected during industrial production is input into the working condition recognition model for working condition identification, yielding the working condition identification results. Compared to existing technologies, this invention optimizes both the conflicting objectives of "improving minority class recognition performance" and "preserving representative information of the majority class" through a multi-objective optimization undersampling model. This addresses class imbalance and class overlap issues at their root, significantly improving the recognition accuracy and classification confidence for minority class abnormal working conditions while maintaining a high overall recognition rate.

[0105] Corresponding to the undersampling-based foam flotation condition identification method described in the above embodiments, such as Figure 2 As shown, this embodiment of the invention also provides a foam flotation condition identification device 100 based on undersampling, the foam flotation condition identification device 100 comprising:

[0106] The acquisition module 101 is used to acquire historical foam image samples during industrial production processes, and to use historical foam image samples acquired under normal operating conditions as majority class samples and historical foam image samples acquired under abnormal operating conditions as minority class samples.

[0107] The solver module 102 is used to input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model through a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results.

[0108] The training module 103 is used to merge the undersampled results with the minority class samples for each undersampled result in the optimal Pareto solution set to obtain a training set, and to use the training set to train the base classifiers to obtain a working condition recognition model integrated by multiple base classifiers.

[0109] The recognition module 104 is used to input the foam image data collected in real time during the industrial production process into the working condition recognition model for working condition recognition, and obtain the working condition recognition result of the industrial production process.

[0110] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0111] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0112] This invention also provides a terminal device, such as... Figure 3 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 3 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, it implements the above-described undersampling-based foam flotation condition identification method.

[0113] The terminal device D10 can be a desktop computer, laptop, handheld computer, server, server cluster, or cloud server, etc. This terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will understand that... Figure 3 This is merely an example of terminal device D10 and does not constitute a limitation on terminal device D10. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0114] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0115] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0116] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0117] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0118] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for identifying foam flotation conditions based on undersampling.

[0119] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0120] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying foam flotation conditions based on undersampling, characterized in that, include: Step 1: Collect historical foam image samples from the industrial production process, and use historical foam image samples collected under normal operating conditions as the majority class samples and historical foam image samples collected under abnormal operating conditions as the minority class samples. Step 2: Input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model using a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results; Step 3: For each undersampling result in the optimal Pareto solution set, merge the undersampling result with the minority class sample to obtain a training set, and use the training set to train the base classifier to obtain a working condition recognition model integrated by multiple base classifiers. Step 4: Input the foam image data collected in real time during the industrial production process into the working condition recognition model for working condition recognition, and obtain the working condition recognition result of the industrial production process.

2. The method for identifying foam flotation conditions based on undersampling according to claim 1, characterized in that, The multi-objective optimization undersampling model is as follows: in, Denotes the first objective function. Indicates the number of minority class samples. Indicates the number of samples in the majority class. undersampling scheme The Middle The predicted label vector of each sample. , undersampling scheme The Middle The true label vector of each sample undersampling scheme The Middle The overall quality value of each sample in the dataset. This represents the second objective function. Indicates the first One sample, , .

3. The method for identifying foam flotation conditions based on undersampling according to claim 1, characterized in that, The multi-objective optimization undersampling model is solved using a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, including: Step 21: Randomly generate the initial population; Step 22: Apply four state transition operators to each group of candidate solutions in the current population to perform state transition, forming a set of offspring candidate solutions, wherein the set of offspring candidate solutions includes multiple offspring candidate solutions; Step 23: Merge the parent population with the offspring candidate solutions to obtain a joint population, which includes multiple individual solutions, and calculate the first target value and the second target value of the undersampling scheme corresponding to each individual solution in the joint population; Step 24: Sort the individual solutions in the joint population using the first target value and the second target value to obtain the Pareto solution set; Step 25: Repeat steps 22-24 until the maximum number of iterations is reached, and obtain the optimal Pareto solution set.

4. The method for identifying foam flotation conditions based on undersampling according to claim 1, characterized in that, The foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in the working condition recognition results for the industrial production process, including: The foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in multiple evidence bodies; By combining all the pieces of evidence using Dempster's combination rule, we obtain the complete piece of evidence. The overall evidence body is mapped to a probability distribution using the Pignistic probability transformation, and the probability distribution is processed based on the principle of maximizing the posterior probability to obtain the working condition identification results in the industrial production process.

5. The method for identifying foam flotation conditions based on undersampling according to claim 4, characterized in that, The foam image data collected in real time during industrial production is input into the working condition recognition model for working condition recognition, resulting in the following expressions for multiple evidence bodies: in, Indicates the first The evidence body output by each base classifier Indicates the first A base classifier evaluates samples Category confidence level , Indicates the number of categories. This represents uncertain quality that is not assigned to any specific category.

6. The method for identifying foam flotation conditions based on undersampling according to claim 5, characterized in that, By combining all the evidence using Dempster's combination rule, the expression for the overall evidence is: in, This indicates the evidence body after fusion. This indicates the evidence body after fusion, including the sample. Category confidence level This indicates the combination rules.

7. The method for identifying foam flotation conditions based on undersampling according to claim 6, characterized in that, The expression for mapping the entire body of evidence to a probability distribution using the Pignistic probability transformation is as follows: in, Indicates sample Category The probability of.

8. A foam flotation condition identification device based on undersampling, characterized in that, include: The acquisition module is used to collect historical foam image samples during industrial production processes. Historical foam image samples collected under normal operating conditions are used as majority class samples, and historical foam image samples collected under abnormal operating conditions are used as minority class samples. The solution module is used to input the majority class samples into the constructed multi-objective optimization undersampling model, and solve the multi-objective optimization undersampling model through a binary multi-objective state transition algorithm to obtain the optimal Pareto solution set, which includes multiple undersampling results; The training module is used to merge the undersampling result with the minority class sample for each undersampling result in the optimal Pareto solution set to obtain a training set, and to train the base classifier using the training set to obtain a working condition recognition model integrated by multiple base classifiers. The identification module is used to input the foam image data collected in real time during the industrial production process into the working condition identification model for working condition identification, and obtain the working condition identification result of the industrial production process.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the undersampling-based foam flotation condition identification method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the undersampling-based foam flotation condition identification method as described in any one of claims 1 to 7.

Images