Hot continuous rolling process fault diagnosis method and device based on multilayer dynamic causal diagram

Abstract

The invention provides a hot continuous rolling process fault diagnosis method and device based on a multilayer dynamic causal diagram, and relates to the technical field of engineering process monitoring. The method comprises the following steps: according to a historical key variable data set, performing causal relationship analysis by using a dynamic causal entropy algorithm to obtain a multi-layer dynamic causal graph, a causal entropy data set and a quality-related causal entropy data set; training a causal entropy prediction model to be trained by using the causal entropy data set and the quality-related causal entropy data set to obtain a causal entropy prediction model, a fault detection threshold and a quality-related fault detection threshold; according to the actual data, performing fault prediction by using a causal entropy prediction model to obtain fault prediction data; and performing fault analysis based on the fault detection threshold and the quality-related fault detection threshold to obtain a fault detection result and a fault tracing path. The hot continuous rolling process fault diagnosis method is based on the multi-layer dynamic causal diagram and is high in judgment precision and high in interpretability.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering process monitoring, and in particular to a hot rolling process fault diagnosis method and device based on a multi-layer dynamic cause-effect graph. Background Art

[0002] Modern large-scale industrial processes are generally characterized by complex processes, numerous steps, coupled variables, and dynamic operation. The hot strip rolling process, for example, typically includes multiple key steps, including the heating furnace, roughing rolling, hot run-off rollers and shears, finishing rolling, laminar cooling, and coiling, forming a highly dynamic, multi-stage process. Hot strip rolling processes produce products with thicknesses ranging from 0.8–12.7 mm and widths from 700–1550 mm. The production process is characterized by a rapid pace, continuous material flow, and drastic state changes. Accurately and timely monitoring of process status, identifying potential anomalies, and effectively locating faults are crucial for ensuring product quality stability and production safety. Data-driven process monitoring methods (such as multivariate statistical analysis, machine learning, and deep learning) have become an important research direction for hot strip rolling fault diagnosis. However, these methods focus on pattern recognition and anomaly detection, lacking the ability to model spatiotemporal causal structures. In particular, they lack the ability to explain faults when they occur, presenting significant challenges for their application in industrial fields.

[0003] Methods based on mathematical models and mechanistic modeling can leverage process physics knowledge to provide highly interpretable analytical results on the causes of variable changes and fault propagation paths. However, as industrial systems expand in scale, become more coupled, and undergo more complex dynamics, constructing complete and accurate mechanistic models becomes increasingly difficult, limiting the application of traditional methods in modern process industries. The rational application of mechanistic knowledge in data-driven fault diagnosis methods would greatly improve the interpretability of fault diagnosis results. Consequently, data-driven methods struggle to achieve joint modeling and interpretable analysis of the spatial hierarchical structure and causal relationships between variables in large-scale industrial processes.

[0004] In the existing technology, there is a lack of a hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph with high discrimination accuracy and strong interpretability. Summary of the Invention

[0005] To address the technical problem of existing data-driven approaches that make it difficult to jointly model and interpret the spatial hierarchical structure and causal relationships between variables in large-scale industrial processes, the present invention provides a method and device for diagnosing hot rolling process faults based on a multi-layer dynamic causal graph. The technical solution is as follows:

[0006] In one aspect, a method for diagnosing hot rolling process faults based on a multi-layer dynamic causal graph is provided. The method is implemented by a hot rolling process fault diagnosis device, and the method comprises:

[0007] Obtain historical data of the hot strip rolling process; based on key variables of the hot strip rolling process, construct a historical key variable data set based on the historical data;

[0008] Based on the physical meaning and spatial location of key variables and the historical key variable data set, a dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set;

[0009] Based on the dynamic causal characteristics and spatiotemporal characteristics of key variables, a causal entropy prediction model to be trained is constructed according to the model structure of a three-dimensional convolutional neural network;

[0010] The causal entropy dataset and the quality-related causal entropy dataset are used to train the causal entropy prediction model to be trained, and the causal entropy prediction model, the fault detection threshold, and the quality-related fault detection threshold are obtained;

[0011] Acquire actual data of the hot strip rolling process; use the causal entropy prediction model to perform fault prediction based on the actual data to obtain fault prediction data;

[0012] Based on fault detection thresholds and quality-related fault detection thresholds, fault analysis is performed according to actual data, fault prediction data, and multi-layer dynamic cause-and-effect diagrams to obtain fault detection results and fault tracing paths.

[0013] On the other hand, a hot rolling process fault diagnosis device based on a multi-layer dynamic causal graph is provided. The device is applied to a hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph. The device includes:

[0014] A data set construction module is used to obtain historical data of the hot strip rolling process; based on the key variables of the hot strip rolling process, a historical key variable data set is constructed according to the historical data;

[0015] The causal graph construction module is used to perform causal relationship analysis based on the physical meaning and spatial location of key variables and the historical key variable data set using a dynamic causal entropy algorithm to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set;

[0016] The model building module is used to build a causal entropy prediction model to be trained based on the dynamic causal characteristics and spatiotemporal characteristics of key variables and the model structure of a three-dimensional convolutional neural network;

[0017] A model training module is used to train the causal entropy prediction model to be trained using the causal entropy dataset and the quality-related causal entropy dataset, and obtain the causal entropy prediction model, the fault detection threshold, and the quality-related fault detection threshold;

[0018] The fault prediction module is used to obtain actual data of the hot strip rolling process; based on the actual data, the causal entropy prediction model is used to perform fault prediction and obtain fault prediction data;

[0019] The fault tracing module is used to perform fault analysis based on fault detection thresholds and quality-related fault detection thresholds, actual data, fault prediction data, and multi-layer dynamic cause-and-effect graphs to obtain fault detection results and fault tracing paths.

[0020] On the other hand, a hot rolling process fault diagnosis device is provided, which includes: a processor; a memory, wherein computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, any one of the above-mentioned hot rolling process fault diagnosis methods based on the multi-layer dynamic causal graph is implemented.

[0021] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned hot rolling process fault diagnosis methods based on a multi-layer dynamic causal graph.

[0022] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0023] This paper proposes a fault diagnosis method for hot rolling processes based on a multi-layer dynamic causal graph. This method combines a causal entropy algorithm with a process mechanism model of hot strip rolling to automatically mine and structure the causal relationships between variables. It constructs a multi-layer causal graph with physical semantics and a hierarchical structure, improving the rationality and interpretability of causal modeling.

[0024] The sliding time window mechanism is used in combination with dynamic causal entropy calculation to achieve dynamic update of the causal graph structure over time, build a multi-layer dynamic causal graph, capture the time-varying characteristics of the causal relationship between key variables, and adapt to behavioral changes under complex working conditions.

[0025] A causal entropy prediction model is constructed by integrating three-dimensional convolution and channel attention mechanisms. This model extracts deep spatiotemporal correlation features and causal coupling features between variables from multi-layer dynamic causal graphs, enhancing the model's detection capability and accuracy for quality-related faults.

[0026] This method combines the detection results of the causal entropy prediction model with the multi-layer dynamic causal graph structure to construct a causal propagation path identification mechanism based on causal relationships. This enables cross-level and explainable fault traceability, providing accurate fault location and cause analysis support for on-site personnel. This method, based on a multi-layer dynamic causal graph, provides high accuracy and strong interpretability for hot rolling process fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 This is a flow chart of a hot rolling process fault diagnosis method based on a multi-layer dynamic cause-effect graph provided by an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of a process flow of hot strip rolling provided by an embodiment of the present invention;

[0030] Figure 3 A multi-layer dynamic cause-and-effect diagram of a strip hot rolling process provided by an embodiment of the present invention;

[0031] Figure 4 This is a schematic diagram of the structure of a 3D-SECNN model that integrates three-dimensional convolution and channel attention mechanisms, provided by an embodiment of the present invention;

[0032] Figure 5 This is a block diagram of a hot rolling process fault diagnosis device based on a multi-layer dynamic causal graph provided by an embodiment of the present invention;

[0033] Figure 6 It is a structural schematic diagram of a hot rolling process fault diagnosis device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0035] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0036] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0037] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0038] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0039] The embodiment of the present invention provides a hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph. The method can be implemented by a hot rolling process fault diagnosis device, which can be a terminal or a server. Figure 1 The flowchart of the hot rolling process fault diagnosis method based on the multi-layer dynamic cause-effect graph is shown. The processing flow of the method may include the following steps:

[0040] S1. Obtain historical data of the hot strip rolling process; based on key variables of the hot strip rolling process, construct a historical key variable data set according to the historical data.

[0041] Among them, the key variables include the key variables of the frame, the key variables of the strip and the loop tension between the frames; the key variables of the frame include rolling force, rolling force difference, bending roll force, balance force, roll gap, rolling speed, and rolling power; the key variables of the strip include the thickness and convexity of the strip.

[0042] In a feasible embodiment, the strip hot rolling process is as follows Figure 2 As shown, the historical data obtained in the present invention includes sensor variable data, performance index data, equipment information data, etc. collected on site. These data fully reflect the evolution of equipment and processes during the hot rolling process, and there is a potential temporal causal relationship between the variables.

[0043] However, the variables collected on-site in the hot rolling industry are diverse, including numerous auxiliary variables that are irrelevant to the target task or redundant. Incorporating all of these into the model would increase computational complexity and interfere with the accuracy of causal modeling. Therefore, a preliminary screening of historical variables was conducted, combining the hot rolling process mechanism with expert knowledge. Variables with strong correlations with fault propagation and quality were retained. Key variables selected for the finishing process included rolling force, rolling force difference, bending roll force, balancing force, roll gap, rolling speed, rolling power and thickness, crown, and inter-stand looper tension.

[0044] The above variables can fully characterize the dynamic characteristics and quality evolution trend of the rolling process, which will help improve the accuracy of subsequent causal relationship modeling and fault detection. According to the application requirements and data set size of the causal entropy prediction model, the historical key variable data set is divided into training sets in proportion. , validation set and test set .

[0045] S2. Based on the physical meaning and spatial location of key variables and the historical key variable data set, a dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set.

[0046] Optionally, based on the physical meaning and spatial location of key variables and according to historical key variable datasets, a dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain a multi-layer dynamic causal graph, a causal entropy dataset, and a quality-related causal entropy dataset, including:

[0047] Based on the historical key variable data set, the CMI-DTE algorithm is used to model the causal relationship and obtain a causal diagram;

[0048] Based on the physical meaning and spatial location of key variables, the causal diagram is divided into layers to obtain a multi-layer causal diagram;

[0049] Based on the preset sliding window and the historical key variable data set, the causal relationship is quantified using the causal entropy algorithm to obtain the dynamic connection weights of the multi-layer causal graph;

[0050] Construct a multi-layer dynamic causal graph based on the multi-layer causal graph and dynamic connection weights;

[0051] According to the multi-layer dynamic causal graph, a causal entropy dataset is obtained;

[0052] The quality-related data sets of key variables are constructed based on the causal entropy data set to obtain the quality-related causal entropy data set.

[0053] In a feasible implementation, the causal relationship between variables is modeled using the Causal Mutual Information-Dynamic Transfer Entropy (CMI-DTE) algorithm, which is calculated using the following formula (1):

[0054] (1);

[0055] in, and represents any two key variables in the key variable dataset, represents the causal entropy, which is calculated by the following formula (2):

[0056] (2);

[0057] in, and Represents the weight coefficient, generally set to 0.5; express right The dynamic mutual information of is calculated by the following formula (3):

[0058] (3);

[0059] in, Represents the source variable in the past The value of the moment, Represents the target variable At the current moment The value of Represents the target variable in the past A historical sequence of moments, Represents the joint or conditional probability distribution of each variable; Represents the source variable Past information about the target variable The dynamic transfer entropy of is calculated by the following formula (4):

[0060] (4);

[0061] The conditional entropy is calculated by the following formula (5):

[0062] (5);

[0063] The joint entropy is calculated by the following formula (6):

[0064] (6);

[0065] like , considering the source variable is the target variable dependent variable; if , considering the source variable is the target variable The outcome variable; if ,It is considered that there is no causal relationship between the source variable and the target variable.,According to the above method, directed edges are added between process,variables to indicate the causal relationship between them.

[0066] According to the above steps, the mechanism knowledge is used to further improve the causal diagram, and the key variables in the diagram are divided into process variable layer, physical system layer and control decision layer according to physical meaning and spatial position to form a multi-layer causal diagram.

[0067] Set the sliding window size to ,According to the causal entropy algorithm, the causal relationship between the key variables in each time window is quantified, and used as the dynamic connection weight of the multi-layer causal graph to construct a multi-layer dynamic causal graph that changes with time.

[0068] The schematic diagram of the multi-layer dynamic cause-effect diagram of the hot strip rolling process is as follows: Figure 3 As shown. The historical data set is calculated by the dynamic causal entropy algorithm to form a multi-layer dynamic causal graph; the multi-layer dynamic causal graph is obtained by visualizing the causal entropy matrix; therefore, the causal entropy data set can be obtained based on the multi-layer dynamic causal graph, which is divided into the causal entropy training set. , causal entropy validation set and the causal entropy test set The causal entropy calculated with the quality variables constitutes the quality-related causal entropy training set , validation set and test set .

[0069] S3. Based on the dynamic causal characteristics and spatiotemporal characteristics of key variables, a causal entropy prediction model to be trained is constructed according to the model structure of a three-dimensional convolutional neural network.

[0070] In a feasible implementation, the present invention proposes a causal entropy prediction model for the hot strip rolling process, namely a three-dimensional compression-excitation convolutional neural network model (3D-Squeeze-and-Excitation Convolutional Neural Network, 3D-SECNN), which mines the dynamic causal characteristics of process variables.

[0071] The model input is the causal relationship matrix that evolves over time from the causal entropy dataset, which can be represented as a tensor sequence ,in, Indicates the number of channels, represents the time step, Indicates height, Indicates width, the schematic diagram of the 3D-SECNN model is as follows Figure 4 shown.

[0072] S4. Use the causal entropy dataset and the quality-related causal entropy dataset to train the causal entropy prediction model to be trained, and obtain the causal entropy prediction model, the fault detection threshold, and the quality-related fault detection threshold.

[0073] Among them, the causal entropy prediction model includes a three-dimensional convolutional layer, a SE attention module, a maximum pooling layer and a fully connected layer.

[0074] In a feasible implementation method, the size of the causal relationship between nodes on a multi-layer dynamic causal graph changes continuously over time. By considering the size of the causal relationship on the multi-layer causal graph at each moment as a causal relationship matrix, a tensor sequence that evolves over time can be obtained, which is called dynamic causal entropy.

[0075] The three-dimensional causal entropy is divided into different batches along the time axis and input into the designed 3D-SECNN model to realize model training, and based on the validation set Prediction, calculation of error statistical characteristics, and adaptive design of threshold. The fault detection threshold is calculated by the following formula (7):

[0076] (7);

[0077] in, and Represents the validation set Prediction error distance The mean and standard deviation of Represents the empirical coefficient. Validation set Prediction error distance Calculated by the following formula (8):

[0078] (8);

[0079] The quality-related fault detection threshold is calculated by the following formula (9):

[0080] (9);

[0081] in, and Represents the quality-related causal entropy prediction error distance The mean and standard deviation of Represents the empirical coefficient. Quality-related causal entropy validation set Prediction error distance Calculated by the following formula (10):

[0082] (10);

[0083] in, Represents the corresponding quality-related causal entropy matrix Elements in .

[0084] S5. Acquire actual data of the strip hot rolling process; perform fault prediction using a causal entropy prediction model based on the actual data to obtain fault prediction data.

[0085] Optionally, based on actual data, a causal entropy prediction model is used to perform fault prediction to obtain fault prediction data, including:

[0086] Based on actual data, the dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain the actual causal entropy data set;

[0087] According to the actual causal entropy data set, feature extraction is performed through a three-dimensional convolution layer to obtain the first data feature;

[0088] Based on the SE attention module, channel calibration is performed according to the first data feature to obtain the second data feature;

[0089] Based on the second data feature, the spatiotemporal evolution feature of the causal relationship is extracted through the maximum pooling layer to obtain the third data feature;

[0090] According to the third data feature, a fully connected layer is used to perform dimensionality reduction mapping to obtain fault prediction data.

[0091] In one feasible implementation, a 3D-Convolutional Neural Network (3D-CNN) is used to extract the structural patterns between nodes in the multi-layer causal graph, and cross-temporal convolution is used to extract the change pattern of the multi-layer causal graph at consecutive time steps. This process is calculated by the following formula (11):

[0092] (11);

[0093] in, Indicates channel The input causal entropy of represents the 3D convolution kernel, represents a three-dimensional convolution operation, Indicates the total number of channels, represents the bias term, represents a nonlinear activation function, Represents the output features.

[0094] The SE attention module dynamically enhances the attention to the causal relationship of key variables, and the global average pooling of the output features is calculated by the following formula (12):

[0095] (12);

[0096] in, Indicates channel The global average pooling result of .

[0097] Generate channel weights, calculated by the following formula (13):

[0098] (13);

[0099] in, and Represent the weight matrices of the two-layer fully connected network, Represents the output channel attention weight vector.

[0100] Recalibrate the features and calculate them using the following formula (14):

[0101] (14);

[0102] in, Represents the features after recalibration.

[0103] The spatiotemporal evolution features of causal relationships are extracted through the maximum pooling layer and calculated by the following formula (15):

[0104] (15);

[0105] in, Represents the output features of the maximum pooling layer.

[0106] The deep features are mapped to the prediction output through the fully connected layer, which is calculated by the following equations (16) and (17):

[0107] (16);

[0108] (17);

[0109] in, and represents the fully connected weight matrix, express The flattened deep feature vector, represents the first layer output, Indicates the operation of converting vector into matrix, Indicates the prediction result.

[0110] S6. Based on the fault detection threshold and the quality-related fault detection threshold, perform fault analysis according to the actual data, fault prediction data and the multi-layer dynamic cause-and-effect diagram to obtain the fault detection results and the fault tracing path.

[0111] Optionally, based on the fault detection threshold and the quality-related fault detection threshold, a fault analysis is performed according to actual data, fault prediction data, and a multi-layer dynamic cause-and-effect graph to obtain fault detection results and fault tracing paths, including:

[0112] Based on the Frobenius norm, statistics are calculated according to actual data and fault prediction data to obtain detection statistics and quality-related detection statistics;

[0113] Based on the fault detection threshold and the quality-related fault detection threshold, fault verification is performed according to the detection statistics and the quality-related detection statistics to obtain a fault detection result;

[0114] Based on the multi-layer dynamic causal graph, abnormal variables are traced according to detection statistics and quality-related detection statistics to obtain the fault tracing path.

[0115] In a feasible implementation, the causal entropy prediction model can predict the causal relationship matrix of the next moment online, and by constructing global statistics and quality-related statistics to compare the predicted value with the actual value, detect abnormal deviations, and thus accurately identify the time when the fault occurs and the corresponding key variable nodes.

[0116] Under normal operating conditions, the causal entropy distribution between key variables remains relatively stable in terms of probability distribution. However, when a fault occurs, the causal strength between specific variables will fluctuate significantly, reflected in the prediction error significantly deviating from the normal fluctuation range. By performing a statistical significance test on the errors of each element in the causal relationship matrix, abnormal units are identified element by element, and a multi-layer causal anomaly graph is dynamically constructed based on the level and connection relationship of the variables. Furthermore, by using the directed dependency paths between variables in the multi-layer dynamic causal graph to track the connectivity between abnormal variables, the possible paths of fault propagation between different levels can be inferred, achieving interpretable cross-level fault tracing analysis.

[0117] Optionally, based on the fault detection threshold and the quality-related fault detection threshold, fault verification is performed according to the detection statistics and the quality-related detection statistics to obtain a fault detection result, including:

[0118] Perform fault detection on the detection statistics according to the fault detection threshold to obtain a fault detection result;

[0119] When the fault verification result is a fault, a quality-related fault verification is performed on the quality-related test statistic according to the quality-related fault detection threshold to obtain a quality verification result;

[0120] The fault detection result is obtained based on the fault verification result and the quality verification result.

[0121] In a feasible implementation, based on real-time prediction of the next moment causal entropy , the detection statistic is constructed using the Frobenius norm and is calculated by the following formula (18):

[0122] (18);

[0123] in, express Moment OK The causal entropy matrix elements of the columns, correspondingly, Represents its predicted value at the next moment.

[0124] if , it is considered that a fault occurs at this moment, otherwise it is normal.

[0125] If a fault occurs, it is further determined whether a quality-related fault occurs. The quality-related detection statistic is constructed using the Frobenius norm and is calculated using the following formula (19):

[0126] (19);

[0127] in, Represents the real-time prediction of the quality-related causal entropy matrix at the next moment, represents the quality-related causal entropy matrix measured in real time, Represents the width of the quality-related causal entropy matrix.

[0128] This paper proposes a fault diagnosis method for hot rolling processes based on a multi-layer dynamic causal graph. This method combines a causal entropy algorithm with a process mechanism model of hot strip rolling to automatically mine and structure the causal relationships between variables. It constructs a multi-layer causal graph with physical semantics and a hierarchical structure, improving the rationality and interpretability of causal modeling.

[0129] The sliding time window mechanism is used in combination with dynamic causal entropy calculation to achieve dynamic update of the causal graph structure over time, build a multi-layer dynamic causal graph, capture the time-varying characteristics of the causal relationship between key variables, and adapt to behavioral changes under complex working conditions.

[0130] A causal entropy prediction model is constructed by integrating three-dimensional convolution and channel attention mechanisms. This model extracts deep spatiotemporal correlation features and causal coupling features between variables from multi-layer dynamic causal graphs, enhancing the model's detection capability and accuracy for quality-related faults.

[0131] This method combines the detection results of the causal entropy prediction model with the multi-layer dynamic causal graph structure to construct a causal propagation path identification mechanism based on causal relationships. This enables cross-level and explainable fault traceability, providing accurate fault location and cause analysis support for on-site personnel. This method, based on a multi-layer dynamic causal graph, provides high accuracy and strong interpretability for hot rolling process fault diagnosis.

[0132] Figure 5 This is a block diagram of a hot rolling process fault diagnosis device based on a multi-layer dynamic causal graph according to an exemplary embodiment. The device is used in a hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph. Figure 5The device includes a data set construction module 510, a causal graph construction module 520, a model construction module 530, a model training module 540, a fault prediction module 550 and a fault tracing module 560.

[0133] The data set construction module 510 is used to obtain historical data of the hot strip rolling process; based on the key variables of the hot strip rolling process, a historical key variable data set is constructed according to the historical data;

[0134] A causal graph construction module 520 is configured to perform causal relationship analysis based on the physical meaning and spatial location of key variables and the historical key variable dataset using a dynamic causal entropy algorithm to obtain a multi-layer dynamic causal graph, a causal entropy dataset, and a quality-related causal entropy dataset;

[0135] A model building module 530 is used to build a causal entropy prediction model to be trained based on the dynamic causal characteristics and spatiotemporal characteristics of key variables and the model structure of a three-dimensional convolutional neural network;

[0136] A model training module 540 is configured to train a causal entropy prediction model to be trained using a causal entropy dataset and a quality-related causal entropy dataset to obtain a causal entropy prediction model, a fault detection threshold, and a quality-related fault detection threshold;

[0137] The fault prediction module 550 is used to obtain actual data of the hot strip rolling process; based on the actual data, a causal entropy prediction model is used to perform fault prediction to obtain fault prediction data;

[0138] The fault tracing module 560 is used to perform fault analysis based on the fault detection threshold and the quality-related fault detection threshold, according to the actual data, the fault prediction data and the multi-layer dynamic cause-and-effect diagram, to obtain the fault detection results and the fault tracing path.

[0139] Among them, the key variables include the key variables of the frame, the key variables of the strip and the loop tension between the frames; the key variables of the frame include rolling force, rolling force difference, bending roll force, balance force, roll gap, rolling speed, and rolling power; the key variables of the strip include the thickness and convexity of the strip.

[0140] Optionally, the causal graph construction module 520 is further configured to:

[0141] Based on the historical key variable data set, the CMI-DTE algorithm is used to model the causal relationship and obtain a causal diagram;

[0142] Based on the physical meaning and spatial location of key variables, the causal diagram is divided into layers to obtain a multi-layer causal diagram;

[0143] Based on the preset sliding window and the historical key variable data set, the causal relationship is quantified using the causal entropy algorithm to obtain the dynamic connection weights of the multi-layer causal graph;

[0144] Construct a multi-layer dynamic causal graph based on the multi-layer causal graph and dynamic connection weights;

[0145] According to the multi-layer dynamic causal graph, a causal entropy dataset is obtained;

[0146] The quality-related data sets of key variables are constructed based on the causal entropy data set to obtain the quality-related causal entropy data set.

[0147] Among them, the causal entropy prediction model includes a three-dimensional convolutional layer, a SE attention module, a maximum pooling layer and a fully connected layer.

[0148] Optionally, the fault prediction module 550 is further configured to:

[0149] Based on actual data, the dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain the actual causal entropy data set;

[0150] According to the actual causal entropy data set, feature extraction is performed through a three-dimensional convolution layer to obtain the first data feature;

[0151] Based on the SE attention module, channel calibration is performed according to the first data feature to obtain the second data feature;

[0152] Based on the second data feature, the spatiotemporal evolution feature of the causal relationship is extracted through the maximum pooling layer to obtain the third data feature;

[0153] According to the third data feature, a fully connected layer is used to perform dimensionality reduction mapping to obtain fault prediction data.

[0154] Optionally, the fault tracing module 560 is further configured to:

[0155] Based on the Frobenius norm, statistics are calculated according to actual data and fault prediction data to obtain detection statistics and quality-related detection statistics;

[0156] Based on the fault detection threshold and the quality-related fault detection threshold, fault verification is performed according to the detection statistics and the quality-related detection statistics to obtain a fault detection result;

[0157] Based on the multi-layer dynamic causal graph, abnormal variables are traced according to detection statistics and quality-related detection statistics to obtain the fault tracing path.

[0158] Optionally, the fault tracing module 560 is further configured to:

[0159] Perform fault detection on the detection statistics according to the fault detection threshold to obtain a fault detection result;

[0160] When the fault verification result is a fault, a quality-related fault verification is performed on the quality-related test statistic according to the quality-related fault detection threshold to obtain a quality verification result;

[0161] The fault detection result is obtained based on the fault verification result and the quality verification result.

[0162] This paper proposes a fault diagnosis method for hot rolling processes based on a multi-layer dynamic causal graph. This method combines a causal entropy algorithm with a process mechanism model of hot strip rolling to automatically mine and structure the causal relationships between variables. It constructs a multi-layer causal graph with physical semantics and a hierarchical structure, improving the rationality and interpretability of causal modeling.

[0163] The sliding time window mechanism is used in combination with dynamic causal entropy calculation to achieve dynamic update of the causal graph structure over time, build a multi-layer dynamic causal graph, capture the time-varying characteristics of the causal relationship between key variables, and adapt to behavioral changes under complex working conditions.

[0164] A causal entropy prediction model is constructed by integrating three-dimensional convolution and channel attention mechanisms. This model extracts deep spatiotemporal correlation features and causal coupling features between variables from multi-layer dynamic causal graphs, enhancing the model's detection capability and accuracy for quality-related faults.

[0165] This method combines the detection results of the causal entropy prediction model with the multi-layer dynamic causal graph structure to construct a causal propagation path identification mechanism based on causal relationships. This enables cross-level and explainable fault traceability, providing accurate fault location and cause analysis support for on-site personnel. This method, based on a multi-layer dynamic causal graph, provides high accuracy and strong interpretability for hot rolling process fault diagnosis.

[0166] Figure 6 FIG. 1 is a schematic diagram of a hot rolling process fault diagnosis device provided by an embodiment of the present invention. Figure 6 As shown, the hot rolling process fault diagnosis equipment may include the above Figure 5 The hot rolling process fault diagnosis device based on the multi-layer dynamic causal graph is shown. Optionally, the hot rolling process fault diagnosis device 610 may include a first processor 2001.

[0167] Optionally, the hot rolling process fault diagnosis device 610 may further include a memory 2002 and a transceiver 2003 .

[0168] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0169] The following combination Figure 6 The various components of the hot rolling process fault diagnosis device 610 are described in detail:

[0170] The first processor 2001 is the control center of the hot rolling process fault diagnosis device 610 and can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0171] Optionally, the first processor 2001 may execute various functions of the hot rolling process fault diagnosis device 610 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002 .

[0172] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 6 CPU0 and CPU1 are shown in FIG.

[0173] In a specific implementation, as an embodiment, the hot rolling process fault diagnosis device 610 may also include multiple processors, such as Figure 6 1 and 2. The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). A processor herein can refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0174] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0175] Alternatively, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and capable of being accessed by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or exist independently and accessed through the interface circuit ( Figure 6 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0176] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0177] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 6 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0178] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently and be connected to the hot rolling process fault diagnosis device 610 through the interface circuit ( Figure 6 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0179] It should be noted that Figure 6 The structure of the hot rolling process fault diagnosis device 610 shown in the figure does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0180] In addition, the technical effects of the hot rolling process fault diagnosis device 610 can refer to the technical effects of the hot rolling process fault diagnosis method based on the multi-layer dynamic causal graph described in the above method embodiment, and will not be repeated here.

[0181] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc.

[0182] It should also be understood that the memory in the embodiments of the present invention may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0183] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0184] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0185] In this disclosure, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of a, b, or c" can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0186] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0187] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0188] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0189] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of the device or unit, which can be electrical, mechanical or other forms.

[0190] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0191] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0192] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical disks.

[0193] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph, characterized in that: The method comprises: Obtain historical data of the hot strip rolling process; based on key variables of the hot strip rolling process, construct a historical key variable data set based on the historical data; Based on the physical meaning and spatial location of key variables and the historical key variable data set, a dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set; Based on the dynamic causal characteristics and spatiotemporal characteristics of key variables, a causal entropy prediction model to be trained is constructed according to the model structure of a three-dimensional convolutional neural network; The causal entropy dataset and the quality-related causal entropy dataset are used to train the causal entropy prediction model to be trained, and the causal entropy prediction model, the fault detection threshold, and the quality-related fault detection threshold are obtained; Acquire actual data of the hot strip rolling process; use the causal entropy prediction model to perform fault prediction based on the actual data to obtain fault prediction data; Based on fault detection thresholds and quality-related fault detection thresholds, fault analysis is performed according to actual data, fault prediction data, and multi-layer dynamic cause-and-effect diagrams to obtain fault detection results and fault tracing paths.

2. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 1 is characterized in that: The key variables include the key variables of the frame, the key variables of the strip and the loop tension between the frames; the key variables of the frame include rolling force, rolling force difference, bending roll force, balancing force, roll gap, rolling speed, and rolling power; the key variables of the strip include the thickness and convexity of the strip.

3. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 1 is characterized in that: Based on the physical meaning and spatial location of key variables and the historical key variable data set, a dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set, including: Based on the historical key variable data set, the CMI-DTE algorithm is used to model the causal relationship and obtain a causal diagram; Based on the physical meaning and spatial location of key variables, the causal diagram is divided into layers to obtain a multi-layer causal diagram; Based on the preset sliding window and the historical key variable data set, the causal relationship is quantified using the causal entropy algorithm to obtain the dynamic connection weights of the multi-layer causal graph; Construct a multi-layer dynamic causal graph based on the multi-layer causal graph and dynamic connection weights; According to the multi-layer dynamic causal graph, a causal entropy dataset is obtained; The quality-related data sets of key variables are constructed based on the causal entropy data set to obtain the quality-related causal entropy data set.

4. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 1 is characterized in that: The causal entropy prediction model includes a three-dimensional convolutional layer, an SE attention module, a maximum pooling layer and a fully connected layer.

5. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 4 is characterized in that: The method of using the causal entropy prediction model to perform fault prediction based on actual data to obtain fault prediction data includes: Based on actual data, the dynamic causal entropy algorithm is used to perform causal relationship analysis to obtain the actual causal entropy data set; According to the actual causal entropy data set, feature extraction is performed through a three-dimensional convolution layer to obtain the first data feature; Based on the SE attention module, channel calibration is performed according to the first data feature to obtain the second data feature; Based on the second data feature, the spatiotemporal evolution feature of the causal relationship is extracted through the maximum pooling layer to obtain the third data feature; According to the third data feature, a fully connected layer is used to perform dimensionality reduction mapping to obtain fault prediction data.

6. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 1 is characterized in that: The method of performing fault analysis based on the fault detection threshold and the quality-related fault detection threshold according to actual data, fault prediction data, and a multi-layer dynamic cause-and-effect graph to obtain fault detection results and fault tracing paths includes: Based on the Frobenius norm, statistics are calculated according to actual data and fault prediction data to obtain detection statistics and quality-related detection statistics; Based on the fault detection threshold and the quality-related fault detection threshold, fault verification is performed according to the detection statistics and the quality-related detection statistics to obtain a fault detection result; Based on the multi-layer dynamic causal graph, abnormal variables are traced according to detection statistics and quality-related detection statistics to obtain the fault tracing path.

7. The hot rolling process fault diagnosis method based on multi-layer dynamic causal graph according to claim 6 is characterized in that: The method of performing fault detection based on the fault detection threshold and the quality-related fault detection threshold and the detection statistic and the quality-related detection statistic to obtain a fault detection result includes: Perform fault detection on the detection statistics according to the fault detection threshold to obtain a fault detection result; When the fault verification result is a fault, a quality-related fault verification is performed on the quality-related test statistic according to the quality-related fault detection threshold to obtain a quality verification result; The fault detection result is obtained based on the fault verification result and the quality verification result.

8. A hot rolling process fault diagnosis device based on a multi-layer dynamic causal graph, wherein the hot rolling process fault diagnosis device based on a multi-layer dynamic causal graph is used to implement the hot rolling process fault diagnosis method based on a multi-layer dynamic causal graph according to any one of claims 1 to 7, characterized in that: The device comprises: A data set construction module is used to obtain historical data of the hot strip rolling process; based on the key variables of the hot strip rolling process, a historical key variable data set is constructed according to the historical data; The causal graph construction module is used to perform causal relationship analysis based on the physical meaning and spatial location of key variables and the historical key variable data set using a dynamic causal entropy algorithm to obtain a multi-layer dynamic causal graph, a causal entropy data set, and a quality-related causal entropy data set; The model building module is used to build a causal entropy prediction model to be trained based on the dynamic causal characteristics and spatiotemporal characteristics of key variables and the model structure of a three-dimensional convolutional neural network; A model training module is used to train the causal entropy prediction model to be trained using the causal entropy dataset and the quality-related causal entropy dataset, and obtain the causal entropy prediction model, the fault detection threshold, and the quality-related fault detection threshold; The fault prediction module is used to obtain actual data of the hot strip rolling process; based on the actual data, the causal entropy prediction model is used to perform fault prediction and obtain fault prediction data; The fault tracing module is used to perform fault analysis based on fault detection thresholds and quality-related fault detection thresholds, actual data, fault prediction data, and multi-layer dynamic cause-and-effect graphs to obtain fault detection results and fault tracing paths.

9. A hot rolling process fault diagnosis device, characterized in that: The hot rolling process fault diagnosis equipment includes: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images