Prebaked anode production equipment health assessment method and device based on artificial intelligence

Abstract

The invention provides a prebaked anode production equipment health assessment method and device based on artificial intelligence, and relates to the technical field of data processing, and the method comprises the steps: obtaining the monitoring data of prebaked anode production equipment, carrying out the data processing through a neural network, determining a corresponding neural network weight based on the monitoring data, and then extracting a target parameter. The key features can be intelligently identified and highlighted, and meanwhile the importance difference between the features is balanced. And determining an output layer weight corresponding to the target parameter by using an extreme learning machine combined with a single-step learning algorithm so as to determine a prediction result of the target parameter and determine the health state of the prebaked anode production equipment. Through the above improvement, the extreme learning machine can adapt to the dynamic change of data, thereby improving the intelligent level of health prediction of the prebaked anode production equipment, and achieving the more efficient fault prediction and prevention.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a health assessment method and device for prebaked anode production equipment based on artificial intelligence. Background Art

[0002] With the development of industrial automation and information technology, prebaked anode production equipment plays an important role in the aluminum electrolysis industry. The operating status of these equipment directly affects production efficiency and product quality. Traditional equipment maintenance methods rely on regular manual inspections and empirical judgments, facing bottlenecks such as poor timeliness and insufficient monitoring accuracy, and are difficult to meet the needs of modern industrial automation upgrades. In view of this, intelligent equipment health management technology that integrates machine learning and artificial intelligence has emerged, aiming to monitor equipment conditions in real time through continuous data collection and analysis, and proactively predict the occurrence of failures, thereby reducing unplanned downtime, controlling maintenance costs, and promoting a significant improvement in production efficiency.

[0003] However, in the implementation of existing intelligent maintenance technologies, the feature selection process fails to fully consider the relative importance of each feature, which can easily lead to gradient vanishing or gradient explosion during training, threatening the convergence performance and long-term stability of the model. In addition, in the face of a rapidly changing industrial data environment, traditional classification algorithms are slow in adjusting learning rates and weight update strategies, which restricts the accuracy and response speed of classification.

[0004] In response to the above-mentioned technical bottlenecks, exploring innovative data processing strategies, optimizing feature engineering methods, developing efficient feature processing technologies, and designing advanced classification models that can dynamically adapt to data flow have important scientific value and practical significance for promoting a leap in the level of intelligent health management of prebaked anode production equipment in the aluminum electrolysis industry. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a health assessment method and device for prebaked anode production equipment based on artificial intelligence, which can improve the intelligence level of the health management system of prebaked anode production equipment and achieve more efficient fault prediction and prevention.

[0006] In a first aspect, an embodiment of the present invention provides a health assessment method for prebaked anode production equipment based on artificial intelligence, wherein the method comprises: obtaining monitoring data of the prebaked anode production equipment; inputting the monitoring data into a preset neural network, determining the neural network weights corresponding to the monitoring data, and adjusting the neural network weights using fuzzy logic; extracting target parameters of the monitoring data based on the adjusted neural network weights; inputting the target parameters into a pre-trained extreme learning machine, calculating the output layer weights of the extreme learning machine using a single-step learning algorithm, and outputting prediction results corresponding to the target parameters based on the output layer weights; and determining the health status of the prebaked anode production equipment based on the prediction results.

[0007] In combination with the first aspect, an embodiment of the present invention provides a first implementation of the first aspect, wherein the monitoring data is input into a preset neural network, the neural network weights corresponding to the monitoring data are determined, and the steps of adjusting the neural network weights using fuzzy logic include: forward propagating the monitoring data through a preset neural network to obtain a neural network output; wherein the neural network includes initial neural network weights; based on the classification error corresponding to the neural network output and a preset feature importance threshold, using fuzzy logic to determine the weight adjustment factor corresponding to the neural network output; based on the weight adjustment factor, adjusting the initial neural network weights of the neural network to obtain adjusted neural network weights.

[0008] In combination with the first aspect, an embodiment of the present invention provides a second implementation of the first aspect, wherein the prediction result includes one of the following health levels: healthy, sub-healthy, mild abnormality, serious abnormality, and failure; based on the prediction result, the step of determining the health status of the prebaked anode production equipment includes: determining the health level in the prediction result as the health status level of the prebaked anode production equipment.

[0009] In combination with the first aspect, an embodiment of the present invention provides a third implementation of the first aspect, wherein the step of calculating the output layer weight of an extreme learning machine using a single-step learning algorithm includes: obtaining a hidden layer output matrix of the extreme learning machine; performing matrix identity calculations on the hidden layer output matrix to determine the output layer weight of the extreme learning machine.

[0010] In combination with the first aspect, an embodiment of the present invention provides a fourth implementation of the first aspect, wherein the above method also includes: performing dimensionality reduction processing on the target parameter through an autoencoder; wherein the hidden layer of the autoencoder adopts a sparse regularization strategy to activate some neurons of the autoencoder, and the activation state of the neurons is determined based on the dynamic threshold of the target parameter.

[0011] In combination with the first aspect, an embodiment of the present invention provides a fifth implementation of the first aspect, wherein the above method also includes: calculating the parameter gradient of the autoencoder; and, based on the weight vectors of all layers of the autoencoder, calculating the adjustment vector corresponding to the target parameter; based on the gradient descent of the parameter gradient and the adjustment vector, updating the weights and biases of the autoencoder.

[0012] In combination with the first aspect, an embodiment of the present invention provides a sixth implementation of the first aspect, wherein the method also includes: obtaining equipment monitoring samples of prebaked anode production equipment; the equipment monitoring samples include data parameters corresponding to the equipment operation; data labeling of the equipment monitoring samples based on the equipment health status of the prebaked anode production equipment to construct an initial sample set; converting the initial sample set into a quantum state, and based on the quantum state of the initial sample set, data expansion of the initial sample set to construct a training sample set.

[0013] In combination with the first aspect, an embodiment of the present invention provides a seventh implementation of the first aspect, wherein the step of converting the initial sample set into a quantum state includes: defining a quantum state conversion function; operating the initial sample set through a generative adversarial network, and determining the angular parameters of the Bloch sphere of the quantum bit based on the weight matrix and bias term of the generative adversarial network; determining the state of a single quantum bit based on the angular parameters; and mapping the initial sample set to a quantum state through the quantum state conversion function based on the state of the quantum bit.

[0014] In combination with the first aspect, an embodiment of the present invention provides an eighth implementation of the first aspect, wherein the step of performing data expansion on the initial sample set based on the quantum state of the initial sample set includes: calculating the Hamiltonian of the quantum state through Pauli-X operation and Pauli-Z operation of quantum bits, and identity operation; based on the Hamiltonian and quantum gate operation, performing data expansion on the quantum state through a generative adversarial network.

[0015] In a second aspect, an embodiment of the present invention provides a health assessment device for prebaked anode production equipment based on artificial intelligence, wherein the device includes: a data acquisition module, used to acquire monitoring data of prebaked anode production equipment; a data processing module, used to input the monitoring data into a preset neural network, determine the neural network weights corresponding to the monitoring data, and adjust the neural network weights using fuzzy logic; a feature extraction module, used to extract target parameters of the monitoring data based on the adjusted neural network weights; an execution module, used to input the target parameters into a pre-trained extreme learning machine, calculate the output layer weights of the extreme learning machine using a single-step learning algorithm, and output the prediction results corresponding to the target parameters based on the output layer weights; an output module, used to determine the health status of the prebaked anode production equipment based on the prediction results.

[0016] The embodiments of the present invention bring the following beneficial effects: The present invention provides a method and device for health assessment of prebaked anode production equipment based on artificial intelligence, which can intelligently identify and highlight key features while balancing the importance differences between features. It can effectively prevent the problems of gradient vanishing and gradient explosion, enhance the learning efficiency and stability of the model, and enable the model to extract early signs of faults from complex signals more quickly. Moreover, the extreme learning machine can automatically optimize the learning parameters based on the real-time changes of the input data based on the single-step learning algorithm, ensuring the efficient learning and rapid response of the model in a dynamic data environment. It can not only improve the classification accuracy, but also speed up the recognition of new fault types.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flowchart of a method for health assessment of prebaked anode production equipment based on artificial intelligence provided by an embodiment of the present invention; Figure 2 A flowchart of another artificial intelligence-based health assessment method for prebaked anode production equipment provided by an embodiment of the present invention; Figure 3 A flow chart of a method for constructing a training sample set provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a health assessment device for prebaked anode production equipment based on artificial intelligence provided by an embodiment of the present invention; Figure 5 A schematic diagram of the structure of another artificial intelligence-based health assessment device for prebaked anode production equipment provided in an embodiment of the present invention; Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the following embodiments of the present disclosure are described by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0022] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described in the present invention may be embodied in a wide variety of forms, and any specific structure and / or function described in the present invention is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described in the present invention may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described in the present invention may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described in the present invention may be used to implement this device and / or practice this method.

[0023] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. The type, quantity and proportion of each component in actual implementation may be changed at will, and the component layout type may also be more complicated. In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the aspects described can be practiced without these specific details.

[0024] The embodiments of the present invention provide a method and device for health assessment of prebaked anode production equipment based on artificial intelligence, which can improve the intelligence level of the health management system of prebaked anode production equipment and achieve more efficient fault prediction and prevention.

[0025] For ease of understanding, a health assessment method for prebaked anode production equipment based on artificial intelligence provided by an embodiment of the present invention is first described. Figure 1 A flowchart of a method for health assessment of prebaked anode production equipment based on artificial intelligence provided by an embodiment of the present invention is shown. Figure 1 As shown, the method comprises the following steps: Step S102, obtaining monitoring data of prebaked anode production equipment.

[0026] The data collected in the present invention is the monitoring data of the prebaked anode production equipment, including the belt conveyor, vibrating screen and demoulding machine used in the prebaked anode production. Their stable operation is directly related to the production efficiency and product quality. Monitoring the operating status of these equipment is very important for preventing failures and optimizing the process flow. Among them, the collected data is stored in a structured JSON format. In one embodiment, the attributes of the data include: is the device operating speed, is the temperature index, is the vibration frequency, is the noise level, is the current intensity, is the voltage level, For routine maintenance times, is the number of failures, is the material type, is the ambient humidity.

[0027] It should be noted that this embodiment is only used to illustrate a data format and type of the present invention. In actual applications, the attributes of data are usually more than 10 attributes, and the number of attributes of data may reach dozens or even hundreds. In the embodiment of the present invention, the monitoring attributes also involve more dimensions of the equipment, such as the specific frequency spectrum of equipment vibration, energy efficiency, emission quality, etc. The embodiment of the present invention collects the above-mentioned types of data of the equipment to identify equipment anomalies, predict failures, and optimize production processes.

[0028] Step S104, input the monitoring data into a preset neural network, determine the neural network weights corresponding to the monitoring data, and use fuzzy logic to adjust the neural network weights.

[0029] Step S106, extracting target parameters of the monitoring data based on the adjusted neural network weights.

[0030] After acquiring the monitoring data, a pre-trained neural network model is used to extract features. In an embodiment of the present invention, the weights corresponding to the monitoring data are determined by a neural network model based on historical monitoring data to understand the characteristics of the input data. Furthermore, an embodiment of the present invention uses fuzzy logic to introduce rule-based decision-making capabilities for weight adjustment to extract target parameters from monitoring data, which can improve the robustness of the model in the face of uncertain data. Among them, the target parameters are indicators directly related to the health of the equipment, such as potential fault signals, degree of performance degradation, etc.

[0031] Step S108, input the target parameters into a pre-trained extreme learning machine, calculate the output layer weights of the extreme learning machine using a single-step learning algorithm, and output the prediction results corresponding to the target parameters based on the output layer weights.

[0032] Step S110, determining the health status of the prebaked anode production equipment based on the prediction result.

[0033] For the extracted target parameters, the embodiment of the present invention uses an extreme learning machine for classification prediction. The advantage of the extreme learning machine is its fast learning speed, which is suitable for real-time or near real-time prediction. In actual operation, the embodiment of the present invention uses a pre-trained extreme learning machine model that is fine-tuned for the application scenario to process the target parameters output from the neural network.

[0034] Among them, the single-step learning algorithm is a machine learning method that allows the model to be updated immediately after receiving each new sample, and is adapted to scenarios where data streams are constantly changing or memory resources are limited. Based on this, it is possible to adapt to dynamic changes in data and output prediction results that are adapted to current data. Finally, based on the prediction results, the health status of the device is judged by the set thresholds or classification rules. For example, if the prediction results show that certain key indicators deviate from the normal range, it can be determined that the device may have health risks or is about to fail. Specifically, the prediction results of an embodiment of the present invention include various health levels of the device, specifically: healthy, sub-healthy, mildly abnormal, severely abnormal, and faulty, 5 categories.

[0035] Furthermore, based on the above embodiment, the embodiment of the present invention also provides another artificial intelligence-based health assessment method for prebaked anode production equipment. Figure 2 A flowchart of another artificial intelligence-based prebaked anode production equipment health assessment method provided by an embodiment of the present invention is shown, referring to Figure 2 , the method comprises the following steps: Step S202, obtaining monitoring data of prebaked anode production equipment.

[0036] Step S204, forward propagating the monitoring data through a preset neural network to obtain a neural network output.

[0037] The present invention uses a 6-layer fully connected neural network for feature extraction. Among them, by extracting a batch of data from the data, forward propagation is performed through the neural network, and the output of each layer will be used as the input of the next layer until the last layer, and the neural network output is obtained. The process of forward propagation is expressed as:

[0038] In the formula, is the Sigmoid activation function; The neural network is The input data of the layer; The neural network is The output of the layer.

[0039] In the prior art, some solutions use neural networks for feature extraction. In some neural network structures, gradient vanishing, gradient exploding, or falling into local optimal solutions may occur, affecting the stability of training and the performance of the model. The present invention uses a neural network algorithm based on feature weight learning for feature extraction. By adopting a dynamic weight adjustment mechanism, the neural network can automatically adjust its weight according to the contribution of different features, and use fuzzy logic to evaluate the importance of each feature to the final prediction task, thereby improving the accuracy of feature extraction. Specifically, refer to the following steps S206-S210.

[0040] Step S206, based on the classification error corresponding to the neural network output and a preset feature importance threshold, fuzzy logic is used to determine the weight adjustment factor corresponding to the neural network output.

[0041] Step S208, adjusting the initial neural network weights of the neural network based on the weight adjustment factor to obtain adjusted neural network weights.

[0042] In the process of forward propagation, the embodiment of the present invention dynamically adjusts the weights of each layer according to fuzzy logic. Specifically, the weights are adjusted according to the influence of each feature on the output target, so that more influential features obtain higher weights. The adjustment method is expressed as:

[0043]

[0044] In the formula, is to adjust the sensitivity, is the feature importance threshold, The neural network The fuzzy logic weight adjustment factor for the layer. is the first neural network after updating according to fuzzy logic The weight of the layer; According to the neural network The output of the layer is classified using the preset Softmax classification function. Preferably, Set to 2.

[0045] Step S210, extracting target parameters of the monitoring data based on the adjusted neural network weights.

[0046] The neural network of the embodiment of the present invention is obtained by pre-training the training sample set. In the specific implementation, the training sample set is learned by the neural network, and the dynamic adjustment factor is determined based on the performance of the model and the influence of the data, so as to update the parameters of the neural network to improve the robustness of the model when facing uncertain data.

[0047] Specifically, the performance of the current model is measured using the loss function of the neural network. In one embodiment, the loss function is calculated as follows:

[0048] In the formula, is the loss function of the neural network, It is The true labels of samples, It is The predicted output of each sample is calculated by the output of the neural network through the preset Softmax classification function. is the number of samples input to the neural network in the current batch, is the regularization term of the neural network loss function.

[0049] In an embodiment of the present invention, the method of initializing the parameters of the neural network is expressed as:

[0050]

[0051] In the formula, It means the mean is 0 and the variance is Normal distribution of The neural network The weight of the layer; The neural network Preferably, Set to 0.01.

[0052] It can be seen from the above loss function that the loss function of the embodiment of the present invention is combined with a regularization term. In combination with the embodiment of the present invention, the weight of the neural network is adjusted by calculating the weight adjustment factor. When training the neural network, the embodiment of the present invention adopts an adaptive fuzzy dynamic regularization method, and the real-time feedback obtained is dynamically adjusted to define the dynamic regularization term. The calculation method of the regularization term of the neural network loss function is expressed as:

[0053] In the formula, is the weight The square of the L2 norm of represents the L2 norm, It is a dynamic adjustment factor obtained based on fuzzy logic.

[0054] Furthermore, the calculation method of the dynamic adjustment factor obtained according to fuzzy logic is expressed as:

[0055] here, is the sensitivity parameter of fuzzy logic, is the feature importance threshold. Preferably, Set to 0.3, Set to 2.

[0056] Back propagation of errors and parameter updates are performed based on the loss function to ensure that each parameter update is in the direction of reducing the total loss. Specifically, the calculation method for the update amount of the neural network weights and biases is expressed as:

[0057]

[0058] In the formula, is the update amount of the neural network weight parameters, is the update amount of the neural network bias parameter, is the learning rate of the neural network, is the symbol of partial derivative.

[0059] Repeat the above steps until the preset stop iteration condition is met, which means that the model training is completed. In one embodiment, the preset stop iteration condition is reaching a preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0060] In summary, the embodiment of the present invention determines the weight adjustment factor based on the importance threshold of the feature, thereby updating the weight of the neural network used to extract the target parameters in the monitoring data based on the weight. At the same time, the neural network used in the embodiment of the present invention adjusts the neural network parameters based on the above-mentioned weight adjustment factor according to the degree of influence of each feature on the output target during the training process, so that the neural network can adapt to the data during the training stage and improve the robustness of the model when facing uncertain data. In summary, the feature extraction of the embodiment of the present invention adopts a fully connected neural network based on feature weight learning. Through a dynamic weight adjustment mechanism, the weight is automatically adjusted according to the degree of influence of each feature on the output target (that is, the importance of the feature), so that more influential features obtain higher weights, effectively deal with the problem of gradient disappearance or explosion, and enhance the model's learning ability for important features. Feature weight learning enables the neural network model to more accurately identify and strengthen the learning of key features, thereby improving the accuracy of feature extraction and the overall effectiveness of the model.

[0061] Furthermore, in high-dimensional data processing, the current feature dimensionality reduction technology fails to accurately distinguish and retain the most critical features, affecting the efficiency and accuracy of the model in complex data analysis. In this regard, the embodiment of the present invention also performs dimensionality reduction processing on the target parameters through an autoencoder. Among them, the hidden layer of the autoencoder uses a sparse regularization strategy to activate some neurons of the autoencoder, and the activation state of the neurons is determined based on the dynamic threshold of the target parameter.

[0062] In the specific implementation, the present invention adopts an autoencoder neural network algorithm based on transform domain sparse representation as a feature dimensionality reduction model. On the basis of the autoencoder algorithm, the hidden layer of the autoencoder adopts a sparse regularization term constraint to only promote the sparse representation of features in the hidden layer, thereby enhancing the model's ability to capture important features in the input data.

[0063] Specifically, the steps for reducing the dimension of the target parameters are as follows: 1) The input feature-extracted data is processed by the encoder part, where the output of each layer is a function of the output of the previous layer and the parameters of the current layer. Specifically, during the training process, the forward propagation process of the data in the autoencoder is expressed as:

[0064] In the formula, is the ReLU activation function, It is the autoencoder The output of the layer, It is the autoencoder The output of the layer.

[0065] The parameters of the autoencoder include the weight and bias of the autoencoder. In one embodiment, the initialization method of the parameters of the autoencoder is expressed as follows:

[0066]

[0067] In the formula, For the autoencoder The weight of the layer, For the autoencoder Bias of the layer; It means the mean is 0 and the variance is Normal distribution of Indicates that in [- , ] is uniformly distributed within the interval. Preferably, Set to 0.1, Set to 0.1.

[0068] 2) In the hidden layer, a sparse regularization strategy is used to force a small number of neurons to be activated, ensuring that only the most important features are retained. Specifically, the activation state of the hidden layer neurons is controlled by a dynamic threshold that depends on the data, expressed as:

[0069] In the formula, is the sparse regularization function; is the output of the hidden layer of the autoencoder; is the regularization strength, which is used to control the degree of sparsity. In the embodiment of the present invention, the regularization strength is set manually; It is the autoencoder The output of the layer.

[0070] Furthermore, the calculated sparse regularization function value is used as a dynamic threshold to determine the activation state of the hidden layer neurons. For example, if the calculated sparse regularization function value is greater than 0.5, 10 neurons are randomly inactivated; if the calculated sparse regularization function value is less than or equal to 0.5 and greater than 0.3, 20 neurons are randomly inactivated; if the calculated sparse regularization function value is less than or equal to 0.3, 30 neurons are randomly inactivated.

[0071] 3) The encoded features are then sent to the decoder for decoding and reconstruction. The task of the decoder is to reconstruct the original input data as much as possible using the encoded data, expressed as:

[0072] In the formula, is the decoder reconstructed input; is the decoder weight, is the decoder bias.

[0073] Specifically, the autoencoder used in the embodiment of the present invention is pre-trained, wherein the responsiveness of the model to changes in input data is enhanced by adopting dynamic weight adjustment. In the specific implementation, the parameter gradient of the autoencoder is calculated; and, based on the weight vectors of all layers of the autoencoder, the adjustment vector corresponding to the target parameter is calculated; based on the gradient descent of the parameter gradient and the adjustment vector, the weight and bias of the autoencoder are updated.

[0074] In the specific implementation, the above steps are used to perform data encoding and decoding operations through the training sample set to perform model training. Furthermore, the loss function of the autoencoder is calculated based on the reconstruction error and the sparse regularization term to find the best balance between the reconstruction accuracy and the sparsity of feature extraction. The calculation method of the loss function of the autoencoder is expressed as:

[0075] In the formula, is the loss function of the autoencoder, is the regularization term weight of the autoencoder loss function. Preferably, Set to 0.3.

[0076] Furthermore, based on the calculated loss, the back propagation algorithm is used to calculate the gradient of each parameter, and the weights and biases of the autoencoder are updated using gradient descent. The update method is expressed as:

[0077]

[0078] In the formula, is the learning rate of the autoencoder; is the updated autoencoder The weight of the layer; is the updated autoencoder Bias of the layer; represents element-wise multiplication; is the dynamic weight adjustment factor, whose elements are , Indicates that it is used for The adjustment vector of the features.

[0079] In one embodiment, dynamic weight adjustment is used to enhance the responsiveness of the model to changes in input data. Specifically, the dynamic weight adjustment factor is dynamically calculated based on the characteristics of the input data and is expressed as:

[0080] In the formula, is the Sigmoid activation function, Indicates that it is used for The adjustment vector of the feature is obtained by taking the average of the weight vectors of all layers of the autoencoder; is a bias term used to adjust the dynamic adjustment factor.

[0081] Further, the above steps are repeated until the preset stop iteration condition is met, indicating that the model training is completed. In one embodiment, the preset stop iteration condition is reaching a preset maximum number of iterations, preferably, the preset maximum number of iterations is set to 1000 times.

[0082] The feature dimensionality reduction model of the embodiment of the present invention uses an autoencoder based on transform domain sparse representation, and retains only the most important features through a sparse regularization strategy, aiming to improve the data processing efficiency and anti-interference ability of the model. The use of an autoencoder based on transform domain sparse representation for feature dimensionality reduction enhances the model's ability to capture important features in the input data, while improving the speed and stability of data processing.

[0083] Step S212, input the target parameters into a pre-trained extreme learning machine, calculate the output layer weights of the extreme learning machine using a single-step learning algorithm, and output the prediction results corresponding to the target parameters based on the output layer weights.

[0084] In specific implementation, the embodiment of the present invention obtains the hidden layer output matrix of the extreme learning machine; performs matrix identity calculation on the hidden layer output matrix to determine the output layer weight of the extreme learning machine, thereby determining the prediction result of the target parameter. Among them, the embodiment of the present invention adopts the extreme learning machine algorithm based on matrix identity as the classifier model, and uses the gradient control mechanism to dynamically adjust the learning rate and weight update strategy during the training process to improve the classification accuracy.

[0085] Specifically, the steps to determine the output layer weights are as follows: 1) Input the data into the hidden layer of the extreme learning machine. The data is expressed as follows through nonlinear transformation of the hidden layer:

[0086] In the formula, is the output feature of the hidden layer of the extreme learning machine; is the Sigmoid activation function; It is the data output from the feature dimensionality reduction model as the input of the extreme learning machine.

[0087] The method of initializing the weights and biases of the extreme learning machine used in the embodiment of the present invention is expressed as follows:

[0088]

[0089] In the formula, is the weight of the extreme learning machine, is the bias of the extreme learning machine; It means the mean is 0 and the variance is Normal distribution of Indicated in Uniform distribution within the interval; The preset parameters for initializing the weights of the extreme learning machine, is the preset parameter for initializing the bias of the extreme learning machine. Preferably, Set to 0.1, Set to 0.2.

[0090] 2) Use the single-step learning algorithm to calculate the output layer weights.

[0091] Specifically, a linear or nonlinear transformation is applied to the hidden layer output to match the expected output. The calculation method of the output layer weight is expressed as:

[0092] In the formula, is the output matrix of the hidden layer of the extreme learning machine, is the transpose of the output matrix of the hidden layer of the extreme learning machine; Calculates matrix identities; is the output layer weight of the extreme learning machine; is the target output matrix of the extreme learning machine; is the regularization parameter; is the identity matrix.

[0093] In one embodiment, the matrix identity is used to solve the inverse calculation problem in high dimensions. Suppose the matrix , then the matrix identity calculation method is expressed as:

[0094] Furthermore, the output layer weights are weights adapted to the target parameters, and the extreme learning machine performs classification prediction based on the weights, which can adapt to dynamic changes in data.

[0095] Furthermore, the embodiment of the present invention also designs the learning rate and weight update method of the extreme learning machine during its training process. Specifically, the following steps are included: 1) In each training batch of the extreme learning machine trained with the training sample set, the learning rate is dynamically adjusted according to the training error of the previous batch. The present invention uses an error-based dynamic learning rate adjustment mechanism to adjust the learning rate of the extreme learning machine. The adjustment method is expressed as:

[0096] In the formula, It is The learning rate of the extreme learning machine for the iteration; is the initial learning rate of the extreme learning machine; Indicates the current iteration number; is the learning rate decay coefficient of the extreme learning machine. Preferably, Set to 0.01, Set to 0.95.

[0097] 2) Further, the loss function of the extreme learning machine is calculated, and the parameters of the extreme learning machine are updated using the gradient descent method. The calculation method of the loss function of the extreme learning machine is expressed as:

[0098] In the formula, is the loss function of the extreme learning machine; is the L2 norm; is the L1 norm, which is used to achieve the sparsity of weights and reduce the complexity of the model; is the sparsity regularization function.

[0099] In one embodiment, an adaptive regularization mechanism is used to optimize the hidden layer sparsity control and output weight calculation to avoid model overfitting. Specifically, the weight change regularization term is defined to be adjusted in real time according to the model performance during the training process, thereby adaptively enhancing or weakening the strength of the regularization. The calculation method is expressed as:

[0100] In the formula, For the The regularization term for the weight change at each iteration; is the initial regularization parameter; is the adjustment parameter of the weight change regularization term; is the change in output weight between the t-th iteration and the t-1-th iteration; Represents the sum of squares of weight changes, which is used to measure the magnitude of weight updates. Preferably, Set to 0.3, Set to 0.01.

[0101] Furthermore, the sparsity regularization function is based on an adaptive regularization mechanism and is used to impose sparsity constraints on the output of the hidden layer nodes. The calculation method is expressed as:

[0102] In the formula, is the sparsity regularization function; is the hidden layer output matrix No. The output value corresponding to each node.

[0103] Further, the above steps are repeated until the preset stop iteration condition is met, indicating that the model training is completed. In one embodiment, the preset stop iteration condition is reaching a preset maximum number of iterations, preferably, the preset maximum number of iterations is set to 1000 times.

[0104] The classifier model of the embodiment of the present invention adopts an extreme learning machine based on matrix identity, and uses a gradient control mechanism to dynamically adjust the learning rate and weight update strategy to optimize the classification accuracy and response speed of the model. The multi-scale gradient control strategy of the extreme learning machine significantly improves the learning efficiency and classification performance of the model, especially the adaptability shown in processing complex data environments.

[0105] Step S214, determining the health status of the prebaked anode production equipment based on the prediction results.

[0106] In an embodiment of the present invention, the prediction result includes one of the following health levels: healthy, sub-healthy, mild abnormality, severe abnormality, and failure; the embodiment of the present invention determines the health level in the prediction result as the health status level of the prebaked anode production equipment.

[0107] Furthermore, the embodiment of the present invention also designs a training sample set. Correspondingly, the embodiment of the present invention also provides another artificial intelligence-based health assessment method for prebaked anode production equipment. Figure 3 A flowchart of the method for constructing a training sample set corresponding to this embodiment is shown. Figure 3 As shown, the method comprises the following steps: Step S10, obtaining equipment monitoring samples of prebaked anode production equipment.

[0108] Step S11, annotating data of equipment monitoring samples based on the equipment health status of the prebaked anode production equipment to construct an initial sample set.

[0109] The equipment monitoring samples refer to the monitoring data of the above embodiment, which will not be described here. The embodiment of the present invention constructs an initial sample set by annotating these data. The annotation method of the present invention is manual annotation. In one embodiment, the annotated categories include various health levels of the equipment, specifically: healthy, sub-healthy, mild abnormality, severe abnormality, and failure, 5 categories.

[0110] Step S12, converting the initial sample set into a quantum state, and based on the quantum state of the initial sample set, performing data expansion on the initial sample set to construct a training sample set.

[0111] It is understandable that in the task of the present invention, the collection, acquisition, labeling and preprocessing of training data are time-consuming and labor-intensive, and insufficient training samples can easily lead to poor model generalization ability and affect the accuracy of the model. Existing data enhancement strategies perform poorly in generating virtual samples that are highly matched with measured data, resulting in model training being constrained by a limited amount of real data, which in turn affects the model's ability to generalize to new scenarios.

[0112] In this regard, the present invention maps the data to the high-dimensional Hilbert space of the quantum state, and uses a generative adversarial network algorithm based on quantum computing to generate samples, thereby achieving data expansion. The embodiment of the present invention can realize complex nonlinear data transformation through quantum gate operations and enhance the expressiveness of the data. The generative adversarial network includes two parts: a quantum generator and a discriminator. The quantum generator uses quantum superposition and entanglement characteristics to generate high-dimensional feature data, and the discriminator determines the consistency of the generated data with the real data.

[0113] Specifically, the embodiment of the present invention converts data into quantum state through the following steps: 1) Define the quantum state conversion function; 2) The initial sample set is operated through a generative adversarial network, and the angular parameters of the Bloch sphere of the quantum bit are determined based on the weight matrix and bias term of the generative adversarial network.

[0114] 3) Determine the state of a single quantum bit based on the angle parameter; 4) Based on the state of the quantum bit, the initial sample set is mapped to the quantum state through the quantum state conversion function.

[0115] Among them, the collected monitoring data of the pre-baked anode production equipment is converted into a quantum state through quantum feature mapping, and each monitoring data point is encoded as a set of quantum gate operations of quantum bits. Specifically, the data mapping method is expressed as:

[0116] In the formula, is a quantum state conversion function, which converts the monitoring data of the prebaked anode production equipment into a quantum state; is the quantum state corresponding to the original monitoring data vector, is the original monitoring data vector.

[0117] In one embodiment, the quantum state conversion function is implemented as follows:

[0118] In the formula, is the first parameter of the Bloch sphere of the quantum bit, is the second parameter of the Bloch sphere of the quantum bit; is the number of quantum bits; is an imaginary unit; indicates that the quantum bit is in the ground state of 0, represents the ground state of the quantum bit, Represents the right vector of the quantum state.

[0119] Furthermore, an embodiment of the present invention combines a generative adversarial network to perform data expansion. Correspondingly, based on the weights and biases of the generator of the generative adversarial network, two key angles of the state of a single quantum bit are described to determine the state of each quantum bit.

[0120] Specifically, the first parameter of the Bloch sphere of the quantum bit and the second parameter of the Bloch sphere of the quantum bit determine the state calculation method of each quantum bit, which is expressed as:

[0121]

[0122] In the formula, and is the weight matrix of the generator, and is the bias vector of the generator, is the Sigmoid activation function.

[0123] Furthermore, a generative adversarial network is used for data expansion, wherein an embodiment of the present invention calculates the Hamiltonian of the quantum state through Pauli-X operations and Pauli-Z operations of quantum bits, as well as identity operations; based on the Hamiltonian and quantum gate operations, the quantum state is expanded through a generative adversarial network.

[0124] The specific implementation includes the following steps: 1) The quantum generator of the generative adversarial network generates new quantum states based on the current quantum state by using a sequence of quantum gates. These quantum states represent the expanded feature data, expressed as:

[0125] In the formula, is the quantum gate operation function, is the quantum gate operation function of the iteration; The quantum state corresponding to the monitoring data vector generated by the generator; The number of samples entered for the current batch; Indicates The monitoring data samples are applied to the Hamiltonian of the quantum state, For the The Hamiltonian coefficients corresponding to the monitoring data samples.

[0126] Among them, the weights and bias parameters of the quantum generator and discriminator are initialized randomly, and the initialized parameters must obey a normal distribution with a mean of 0 and a variance of the unit matrix.

[0127] Referring to the above formula, it can be seen that the embodiment of the present invention not only uses the principles of quantum mechanics to enrich the data set and improve the performance of the machine learning model when processing data with complex structures and high correlations. It also combines the Hamiltonian of the quantum state for data expansion. The Hamiltonian is the sum of the energy of the quantum system and includes all possible energy interaction terms of the system to capture the core features of the data and potential interactions.

[0128] In one embodiment, The calculation of the Hamiltonian of the quantum state using a monitoring data sample adopts multiple operations to jointly determine the conversion process of the quantum state. The calculation method is expressed as:

[0129] In the formula, and They are Pauli-X and Pauli-Z operations on qubits, The embodiment of the present invention uses the principle of quantum mechanics to enrich the data set and improve the performance of the machine learning model when processing data with complex structures and high correlation.

[0130] 2) The discriminator evaluates the consistency between the data generated by the quantum generator and the actual monitoring data. The calculation method of its loss function is expressed as:

[0131] In the formula, is the loss function of the discriminator; The quantum state corresponding to the j-th monitoring data vector generated by the generator; is the label of the jth monitoring data, indicating whether the data is real data; is the inverse function of the quantum state conversion function; is the discriminator function.

[0132] In one embodiment, the inverse function of the quantum state conversion function maps the quantum state back to traditional data, expressed as:

[0133]

[0134]

[0135]

[0136] In the formula, is the vector that maps back from the quantum state to traditional data; For quantum state The status value of is the first Pauli matrix, is the second Pauli matrix, is a third Pauli matrix, wherein the first Pauli matrix, the second Pauli matrix and the third Pauli matrix are used to extract data mapped back to the traditional feature space from the quantum state; To take the real part operation, To take the imaginary part operation; represents the inner product; Indicates the absolute value operation.

[0137] The generator and discriminator of the generative adversarial network of the embodiment of the present invention are subjected to data expansion after training. Corresponding to the above steps, during the training process, the embodiment of the present invention adjusts the quantum gate parameters of the quantum generator based on the feedback of the discriminator loss function to optimize the generation process of the quantum state. The adjustment method is expressed as:

[0138] In the formula, For the updated The Hamiltonian coefficient corresponding to the monitoring data sample is is the learning rate of the generative adversarial network, Represents the loss function with respect to parameter The partial derivative of is the parameter adjustment term of the Hamiltonian coefficient. Preferably, Set to 0.01.

[0139] Furthermore, the loss function is about parameter The calculation method of the partial derivative is expressed as:

[0140] In the formula, represents the right vector of the quantum state, Represents the left vector of the quantum state.

[0141] In one embodiment, the parameter adjustment item of the Hamiltonian coefficient is obtained by an adaptive feature optimization mechanism to enhance the capability of the quantum generator, and dynamically adjust the parameters of the quantum gate operation according to the evolution of the quantum state, thereby optimizing the feature representation of the generated data. Specifically, the adaptive adjustment mechanism adjusts the parameters of each operation based on the evolution entropy of the quantum state, thereby better controlling the flow of information and the representation of the feature, and defining an entropy metric to evaluate the degree of disorder of the quantum state, and the calculation method is expressed as:

[0142] In the formula, It is a quantum state The density matrix of represents trace operation; is the entropy measurement function.

[0143] Furthermore, based on the degree of disorder of the quantum state, the parameter adjustment term of the Hamiltonian coefficient is calculated as follows:

[0144] In the formula, is the learning rate of the Hamiltonian. Preferably, Set to 0.03.

[0145] Further, the above steps are repeated until the preset stop iteration condition is met, indicating that the model training is completed. In one embodiment, the preset stop iteration condition is reaching a preset maximum number of iterations, preferably, the preset maximum number of iterations is set to 1000 times.

[0146] In summary, another artificial intelligence-based health assessment method for prebaked anode production equipment provided by the embodiment of the present invention uses a quantum computing-based generative adversarial network for data expansion, and generates virtual data with high consistency with actual monitoring data through quantum state mapping and quantum gate operations, thereby solving the problem of insufficient training data. Moreover, it can achieve effective expansion of training data, improving the generalization ability and prediction accuracy of the model.

[0147] Furthermore, based on the above embodiments, the embodiments of the present invention also provide a health assessment device for prebaked anode production equipment based on artificial intelligence. Figure 4 FIG. 1 shows a schematic diagram of a health assessment device for prebaked anode production equipment based on artificial intelligence provided by an embodiment of the present invention. Figure 4 As shown, the device includes: a data acquisition module 100, used to acquire monitoring data of prebaked anode production equipment; a data processing module 200, used to input the monitoring data into a preset neural network, determine the neural network weights corresponding to the monitoring data, and adjust the neural network weights using fuzzy logic; a feature extraction module 300, used to extract target parameters of the monitoring data based on the adjusted neural network weights; an execution module 400, used to input the target parameters into a pre-trained extreme learning machine, calculate the output layer weights of the extreme learning machine using a single-step learning algorithm, and output the prediction results corresponding to the target parameters based on the output layer weights; an output module 500, used to determine the health status of the prebaked anode production equipment based on the prediction results.

[0148] An artificial intelligence-based health assessment device for prebaked anode production equipment provided in an embodiment of the present invention has the same technical features as an artificial intelligence-based health assessment method for prebaked anode production equipment provided in the above embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0149] Furthermore, based on the above embodiment, the embodiment of the present invention also provides another artificial intelligence-based health assessment device for prebaked anode production equipment. Figure 5The schematic diagram of the structure of another artificial intelligence-based health assessment device for prebaked anode production equipment provided by an embodiment of the present invention is shown. The data processing module 200 is also used to forward propagate the monitoring data through a preset neural network to obtain a neural network output; based on the classification error corresponding to the neural network output and a preset feature importance threshold, fuzzy logic is used to determine the weight adjustment factor corresponding to the neural network output; based on the weight adjustment factor, the initial neural network weight of the neural network is adjusted to obtain an adjusted neural network weight.

[0150] The prediction result includes one of the following health levels: healthy, sub-healthy, mild abnormality, severe abnormality, and failure; the output module 500 is also used to determine the health level in the prediction result as the health status level of the prebaked anode production equipment.

[0151] The execution module 400 is further used to obtain the hidden layer output matrix of the extreme learning machine; perform matrix identity calculation on the hidden layer output matrix to determine the output layer weight of the extreme learning machine.

[0152] The above-mentioned execution module 400 is also used to reduce the dimension of the target parameter through an autoencoder; wherein, the hidden layer of the autoencoder adopts a sparse regularization strategy to activate some neurons of the autoencoder, and the activation state of the neurons is determined based on the dynamic threshold of the target parameter.

[0153] The execution module 400 is also used to calculate the parameter gradient of the autoencoder; and, based on the weight vectors of all layers of the autoencoder, calculate the adjustment vector corresponding to the target parameter; based on the gradient descent of the parameter gradient and the adjustment vector, update the weight and bias of the autoencoder.

[0154] The device also includes a construction module 600, which is used to obtain equipment monitoring samples of prebaked anode production equipment; annotate the equipment monitoring samples based on the equipment health status of the prebaked anode production equipment to construct an initial sample set; convert the initial sample set into a quantum state, and based on the quantum state of the initial sample set, perform data expansion on the initial sample set to construct a training sample set.

[0155] The above-mentioned building module 600 is also used to define a quantum state conversion function; operate on the initial sample set through a generative adversarial network, and determine the angle parameters of the Bloch sphere of the quantum bit based on the weight matrix and bias term of the generative adversarial network; determine the state of a single quantum bit based on the angle parameters; and map the initial sample set to a quantum state through a quantum state conversion function based on the state of the quantum bit.

[0156] The above-mentioned building module 600 is also used to calculate the Hamiltonian of the quantum state through Pauli-X operation and Pauli-Z operation of quantum bits, as well as identity operation; based on the Hamiltonian and quantum gate operation, the quantum state is data expanded through a generative adversarial network.

[0157] An embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above Figures 1 to 3 The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to perform the above Figures 1 to 3 The embodiment of the present invention also provides a structural diagram of an electronic device, such as Figure 6 FIG. 1 is a schematic diagram of the structure of the electronic device, wherein the electronic device includes a processor 61 and a memory 60, the memory 60 stores computer executable instructions that can be executed by the processor 61, and the processor 61 executes the computer executable instructions to implement the above Figures 1 to 3 The method shown. Figure 6 In the illustrated embodiment, the electronic device further includes a bus 62 and a communication interface 63 , wherein the processor 61 , the communication interface 63 and the memory 60 are connected via the bus 62 .

[0158] Among them, the memory 60 may include a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used. The bus 62 can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc., and can also be an AMBA (Advanced Microcontroller Bus Architecture, on-chip bus standard) bus, wherein AMBA defines three types of buses, including APB (Advanced Peripheral Bus) bus, AHB (Advanced High-performance Bus) bus and AXI (Advanced eXtensible Interface) bus. The bus 62 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6Only one bidirectional arrow is used in the diagram, but it does not mean that there is only one bus or one type of bus. The processor 61 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 61 or the instructions in the form of software. The above-mentioned processor 61 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to be executed, or a combination of hardware and software modules in the decoding processor can be executed. The software module can be located in a random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register or other mature storage media in the art. The storage medium is located in the memory, and the processor 61 reads the information in the memory and completes the above-mentioned Figures 1 to 3 Any of the methods shown. A computer program product of a method and apparatus for health assessment of prebaked anode production equipment based on artificial intelligence provided in an embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the method described in the previous method embodiment. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0159] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal connection of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood in specific circumstances. If the function is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program code. Finally, it should be noted that the above embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solution of the present invention, rather than to limit it. The protection scope of the present invention is not limited to this. Although the present invention is described in detail with reference to the above embodiments, those skilled in the art should understand that any technician familiar with the technical field can still modify the technical solution recorded in the above embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solution deviate from the spirit and scope of the technical solution of the embodiment of the present invention, and should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A health assessment method for prebaked anode production equipment based on artificial intelligence, characterized in that: The method comprises: Obtain monitoring data from prebaked anode production equipment; Inputting the monitoring data into a preset neural network, determining the neural network weights corresponding to the monitoring data, and adjusting the neural network weights using fuzzy logic; extracting target parameters of the monitoring data based on the adjusted neural network weights; Inputting the target parameter into a pre-trained extreme learning machine, calculating the output layer weight of the extreme learning machine using a single-step learning algorithm, and outputting a prediction result corresponding to the target parameter based on the output layer weight; Based on the prediction result, the health status of the prebaked anode production equipment is determined.

2. The method according to claim 1, characterized in that The steps of inputting the monitoring data into a preset neural network, determining the neural network weights corresponding to the monitoring data, and adjusting the neural network weights using fuzzy logic include: Performing forward propagation on the monitoring data through a preset neural network to obtain a neural network output; Based on the classification error corresponding to the neural network output and a preset feature importance threshold, fuzzy logic is used to determine the weight adjustment factor corresponding to the neural network output; Based on the weight adjustment factor, the initial neural network weight of the neural network is adjusted to obtain an adjusted neural network weight.

3. The method according to claim 1, characterized in that The prediction result includes one of the following health levels: healthy, sub-healthy, slightly abnormal, seriously abnormal, and fault; The step of determining the health status of the prebaked anode production equipment based on the prediction result comprises: The health level in the prediction result is determined as the health status level of the prebaked anode production equipment.

4. The method according to claim 1, characterized in that The step of calculating the output layer weight of the extreme learning machine using a single-step learning algorithm comprises: Obtaining a hidden layer output matrix of the extreme learning machine; Matrix identity calculation is performed on the hidden layer output matrix to determine the output layer weight of the extreme learning machine.

5. The method according to claim 1, characterized in that The method further comprises: The target parameter is subjected to dimensionality reduction processing by an autoencoder; wherein, the hidden layer of the autoencoder uses a sparse regularization strategy to activate some neurons of the autoencoder, and the activation state of the neurons is determined based on a dynamic threshold of the target parameter.

6. The method according to claim 5, characterized in that The method further comprises: Calculating the parameter gradient of the autoencoder; and, based on the weight vectors of all layers of the autoencoder, calculating the adjustment vector corresponding to the target parameter; Based on the gradient descent of the parameter gradient and the adjustment vector, the weights and biases of the autoencoder are updated.

7. The method according to claim 1, characterized in that The method further comprises: Obtain equipment monitoring samples for prebaked anode production equipment; Based on the equipment health status of the prebaked anode production equipment, data annotation is performed on the equipment monitoring samples to construct an initial sample set; The initial sample set is converted into a quantum state, and based on the quantum state of the initial sample set, data expansion is performed on the initial sample set to construct a training sample set.

8. The method according to claim 7, characterized in that The step of converting the initial sample set into a quantum state comprises: Define quantum state transfer function; Performing operations on the initial sample set through a generative adversarial network, and determining angle parameters of the Bloch sphere of the quantum bit based on a weight matrix and a bias term of the generative adversarial network; determining a state of a single quantum bit based on the angle parameter; Based on the state of the quantum bit, the initial sample set is mapped to a quantum state through the quantum state conversion function.

9. The method according to claim 7, characterized in that: The step of performing data expansion on the initial sample set based on the quantum state of the initial sample set comprises: Calculating the Hamiltonian of the quantum state through Pauli-X and Pauli-Z operations of the quantum bit and identity operations; Based on the Hamiltonian and quantum gate operations, data expansion is performed on the quantum state through a generative adversarial network.

10. An artificial intelligence-based health assessment device for prebaked anode production equipment, characterized in that: The device comprises: A data acquisition module, used to acquire monitoring data of prebaked anode production equipment; A data processing module, used to input the monitoring data into a preset neural network, determine the neural network weights corresponding to the monitoring data, and adjust the neural network weights using fuzzy logic; A feature extraction module, for extracting target parameters of the monitoring data based on the adjusted neural network weights; An execution module, configured to input the target parameter into a pre-trained extreme learning machine, calculate the output layer weight of the extreme learning machine using a single-step learning algorithm, and output a prediction result corresponding to the target parameter based on the output layer weight; An output module is used to determine the health status of the prebaked anode production equipment based on the prediction result.

Images