Metallurgical equipment twin HMI control system based on multi-source data

Abstract

The invention belongs to the technical field of metallurgical equipment intelligent control and digital twinning, and provides a metallurgical equipment twinning HMI control system based on multi-source data, and the system comprises a multi-source data collection module which is used for collecting the multi-source data; the data fusion processing module is used for performing data fusion processing on the multi-source data by adopting a space-time alignment algorithm and a feature extraction network; the digital twinborn modeling module is used for constructing a metallurgical equipment digital twinborn model by utilizing an improved deep belief network algorithm based on the fusion processing data; the HMI interaction module is used for dynamically presenting the fusion processing data and the operation state data of the metallurgical equipment digital twinning model based on the output of the digital twinning modeling module; and the intelligent control module is used for performing closed-loop control on the metallurgical equipment by adopting a self-adaptive predictive control algorithm. According to the invention, full-process intelligent management of the metallurgical equipment can be realized, and the operation efficiency, the control precision and the reliability of the metallurgical equipment are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent control of metallurgical equipment and digital twin technology, and particularly relates to a metallurgical equipment twin HMI control system based on multi-source data. BACKGROUND

[0002] In the metallurgical industry, metallurgical equipment (such as raw material yards, pelletizing, sintering, converters, electric furnaces, refining furnaces, rolling mills, continuous casting machines, etc.) has complex structure and variable operating conditions, and its stable and efficient operation is crucial to product quality and production efficiency.

[0003] Traditional metallurgical process equipment control and monitoring methods have many shortcomings: on the one hand, control relies heavily on human experience or simple automation programs, making it difficult to cope with complex and variable production conditions. When faced with multi-source interference (such as fluctuations in raw material composition, changes in environmental temperature, etc.), control accuracy and response speed are not satisfactory, which can lead to unstable product quality. On the other hand, monitoring methods are scattered, and various equipment operating data (such as temperature, pressure, speed, etc.), environmental data (such as workshop humidity, dust concentration, etc.), and production process data (such as material transportation progress, process connection, etc.) are not effectively integrated, making it difficult for maintenance personnel to fully understand the equipment status through intuitive means, resulting in delayed fault warning and diagnosis, increasing the risk of equipment downtime and maintenance costs.

[0004] Although digital twin technology has been applied in some industrial fields, it can map physical entities to virtual digital models to simulate, analyze and optimize the entities. However, in the context of metallurgical equipment HMI control, existing solutions have the following shortcomings: first, multi-source data fusion is low, and equipment operating data, environmental data, process data, etc. from the entire metallurgical production process are not fully integrated, making it difficult to build an accurate and practical production-oriented equipment twin model. Second, the HMI interaction design lacks specificity and does not consider the operational needs of maintenance personnel in complex metallurgical conditions, resulting in cluttered interface display information or missing key information, which is not conducive to quick decision-making and precise control. Third, the coordination mechanism between the control system and the twin model is not perfect, making it difficult to dynamically and intelligently optimize the control of metallurgical equipment based on real-time simulation and analysis results of the twin model, and failing to fully realize the value of digital twin technology in efficient operation and quality control of metallurgical equipment.

[0005] Therefore, it is necessary to provide a metallurgical equipment twin HMI control system based on multi-source data, which optimizes HMI to achieve intelligent control and efficient interaction, thereby improving the stability, controllability and intelligent level of the production process of metallurgical equipment. SUMMARY

[0006] The present application provides a metallurgical equipment twin HMI control system based on multi-source data, through the cooperative work of each module, realizes the whole-process intelligent management of the metallurgical equipment from data acquisition, fusion processing, digital twin modeling, human-computer interaction to intelligent control, fault diagnosis and self-healing, and linkage with enterprise management system, significantly improves the operation efficiency, control accuracy and reliability of the metallurgical equipment, reduces the equipment failure rate and operation and maintenance cost, and provides strong technical support for the intelligent production of metallurgical enterprises.

[0007] The present application provides a metallurgical equipment twin HMI control system based on multi-source data, comprising:

[0008] A multi-source data acquisition module is used to acquire multi-source data by using a configured sensor assembly, an environment monitoring terminal and a process parameter acquisition assembly; the multi-source data includes metallurgical equipment body operation data, production environment data and whole-process process data;

[0009] A data fusion processing module is used to perform data fusion processing on the multi-source data by using a space-time alignment algorithm and a feature extraction network to obtain fusion processing data;

[0010] A digital twin modeling module is used to construct a metallurgical equipment digital twin model based on the fusion processing data by using an improved deep belief network algorithm; the improved deep belief network algorithm introduces a metallurgical process constraint regularization term to constrain the model training process, so as to improve the compliance of the model to the metallurgical process rules and the prediction accuracy;

[0011] An HMI interaction module is used to dynamically present the fusion processing data and the running state data of the metallurgical equipment digital twin model based on the output of the digital twin modeling module;

[0012] An intelligent control module is used to perform closed-loop control on the metallurgical equipment by using an adaptive predictive control algorithm based on the fusion processing data and the running state data of the metallurgical equipment digital twin model in combination with the real-time feedback of the multi-source data.

[0013] Further, the sensor assembly includes a temperature sensor, a pressure sensor and a vibration sensor; the temperature sensor is a distributed optical fiber temperature measuring device, which is arranged in a spiral around the high temperature area of the metallurgical equipment smelting furnace and is densely arranged in the easy erosion area of the metallurgical equipment smelting furnace lining;

[0014] The metallurgical equipment body operation data includes the real-time temperature, pressure and vibration frequency of the key components of the metallurgical equipment; the production environment data includes the workshop temperature and humidity, dust concentration and harmful gas content; and the whole-process process data includes the material composition, conveying speed and energy consumption of each process.

[0015] Further, the data fusion processing on the multi-source data by using the space-time alignment algorithm and the feature extraction network includes:

[0016] The multi-source data is cleaned, denoised and time-space registered by using a space-time alignment algorithm, and the multi-source data is periodically synchronized and corrected based on a designed period matching factor to obtain corrected data;

[0017] The dynamic weight distribution mechanism is introduced for weight fusion processing of the corrected data to obtain basic fusion processing data; the dynamic weight distribution mechanism dynamically adjusts the data fusion weight of each data source according to the real-time reliability of the data source of the multi-source data, and automatically reduces the weight proportion when an abnormal fluctuation occurs in a certain data source;

[0018] Based on the constructed feature extraction network, the basic fusion processing data is subjected to feature extraction to obtain a key feature vector reflecting the running state of the metallurgical equipment, and the fusion processing data is composed.

[0019] Further, an improved deep belief network algorithm is used to construct a metallurgical equipment digital twin model, including: in the training process of the improved deep belief network algorithm, a metallurgical process constraint regularization term is introduced for constraint, and the model parameters are corrected based on a metallurgical process expert rule base to construct the metallurgical equipment digital twin model; the metallurgical process expert rule base contains metallurgical equipment overload protection logic and material adaptation process interval rules, and is configured to be dynamically updated.

[0020] Further, the digital twin modeling module further includes a multi-physical field coupling correction unit, which is configured to perform the following operations:

[0021] A physical mechanism model is constructed based on thermodynamic-hydrodynamic-structural mechanics coupling equations, and temperature, flow rate and stress monitoring values in the fusion processing data are received in real time;

[0022] When the output of the metallurgical equipment digital twin model deviates from the simulation result of the physical mechanism model by more than a preset threshold, a parameter recalibration mechanism is triggered; the parameter recalibration mechanism corrects the deep belief network weight through a back propagation algorithm with a physical field loss function, and outputs an optimized metallurgical equipment digital twin model.

[0023] Further, the HMI interaction module includes an augmented reality assisted operation and maintenance unit;

[0024] The augmented reality assisted operation and maintenance unit is configured to superimpose and display the internal structure of the metallurgical equipment in the form of augmented reality in the real space based on the metallurgical equipment digital twin model; the internal structure of the metallurgical equipment includes at least one of a rolling mill gear box transmission structure and a smelting furnace gas flow channel;

[0025] The augmented reality assisted operation and maintenance unit is further configured to enable the operation and maintenance personnel to intuitively view the running state of the hidden components of the metallurgical equipment through the augmented reality display device; the running state of the hidden components includes at least one of temperature and vibration distribution.

[0026] The augmented reality auxiliary operation and maintenance unit is further configured to realize spatial registration between the augmented reality display and the actual position of the metallurgical equipment.

[0027] Further, the adaptive predictive control algorithm of the intelligent control module is configured to:

[0028] The device operation trend prediction model is constructed, the long short-term memory network is adopted for time series analysis of the multi-source data, and the prediction period is adjusted according to the rhythm of the metallurgical process;

[0029] The optimal control scheme is generated in combination with the equipment response results under different control strategies simulated by the metallurgical equipment digital twin model, wherein the equipment response results include product thickness changes after adjusting the rolling mill speed.

[0030] Further, it further includes a fault diagnosis and self-healing control module; the fault diagnosis and self-healing control module is configured to: based on the equipment operation state data output by the metallurgical equipment digital twin model, adopt an improved isolated forest algorithm fused with the characteristics of the metallurgical equipment fault samples for fault identification; after identifying the fault, automatically trigger the self-healing control strategy, and the self-healing control strategy includes switching redundant equipment and / or adjusting process parameters for compensation.

[0031] Further, the fault diagnosis and self-healing control module includes a voiceprint feature analysis submodule, which is used to perform the following operations:

[0032] Based on the high-heat-resistant acoustic sensors deployed at key positions such as bearings and gearboxes of the metallurgical equipment, the voiceprint signal data of the equipment operation is collected;

[0033] Based on the Mel-frequency cepstral coefficients extracted from the voiceprint signal data, a voiceprint-vibration multi-modal fault identification model is constructed in combination with the improved isolated forest algorithm;

[0034] Based on the real-time dynamic monitoring of the voiceprint features by the voiceprint-vibration multi-modal fault identification model, if the voiceprint features deviate from the normal range and meet the fault judgment conditions, a fault warning signal is generated.

[0035] Further, it further includes a production management interaction module, and the production management interaction module is configured to interact with the production management system of the metallurgical enterprise; the production management interaction module is configured to: support synchronizing the multi-source data collected by the multi-source data acquisition module and the analysis results of the twin modeling module output by the digital twin modeling module to the production management system of the metallurgical enterprise; realize the linkage between the device control function of the metallurgical equipment twin HMI control system based on the multi-source data and the production planning function of the production management system of the metallurgical enterprise through the production management interaction module; receive the production task instructions of the production management system of the metallurgical enterprise, and the production task instructions include target product specifications, delivery cycles and energy consumption indicators;

[0036] The process parameter self-optimization engine is configured to: based on a dynamic knowledge graph constructed based on a metallurgical process expert rule base, parse production task instructions, generate multiple candidate process paths;Call the metallurgical equipment digital twin model, simulate each candidate process path, and predict product quality, equipment wear and tear and comprehensive energy consumption under each path;Based on a pre-set multi-objective optimization function, filter out the optimal process parameter combination from the candidate paths;The optimal process parameter combination is issued as the initial control parameter to the intelligent control module, and in the production process, fine-tuning is performed according to real-time feedback data to realize dynamic self-optimization of process parameters;Wherein, the dynamic knowledge graph is learned by graph neural network, and its nodes include equipment state, process parameters, material properties, environmental factors, product quality and energy consumption, and the edges represent the causal relationship and influence weight between nodes;The weight of the dynamic knowledge graph is continuously updated based on the deviation between the simulation results of the twin model and the actual production data through an online learning mechanism.

[0037] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0038] 1. The multi-source data acquisition module realizes high-precision acquisition of full-dimensional data of metallurgical equipment through the collaborative arrangement of multiple types of sensors and monitoring terminals, laying a solid data foundation for subsequent data processing and model construction;The differential arrangement of the distributed optical fiber temperature measurement device in the high-temperature zone and the easy-erosion area effectively improves the pertinence and reliability of temperature data acquisition, ensuring the accuracy of the metallurgical equipment operation data.

[0039] 2. The spatio-temporal alignment algorithm and dynamic weight distribution mechanism adopted by the data fusion processing module can effectively solve the inconsistency of multi-source data in time and space, and dynamically adjust the weight according to the real-time reliability of the data source, significantly improving the quality of the fusion processing data;The feature extraction network further extracts key feature vectors from the basic fusion data, providing high-quality input features for the construction of the digital twin model.

[0040] 3. The digital twin modeling module introduces a metallurgical process constraint regularization term and a metallurgical process expert rule base, combined with a multi-physics field coupling correction unit, greatly improving the accuracy and reliability of the digital twin model;The improved deep belief network algorithm corrects the simulation results and actual data, and the expert rule base is optimized twice, so that the model can more accurately reflect the actual operation state of the metallurgical equipment;The multi-physics field coupling correction unit calibrates the digital twin model based on the physical mechanism model, further enhancing the physical consistency of the model and ensuring the credibility of the model output.

[0041] 4. The HMI interaction module as the core of human-computer interaction, through the bidirectional interaction with the digital twin modeling module, realizes the dynamic and intuitive presentation of the fusion processing data and the twin body running state data, provides comprehensive and real-time equipment running information for the operator, and enhances the reality auxiliary operation and maintenance unit visualizes the running state of the internal structure and hidden components of the equipment through the augmented reality technology, greatly improves the efficiency and accuracy of the operation and maintenance work.

[0042] 5. The intelligent control module adopts the adaptive predictive control algorithm, combines the real-time feedback of multi-source data, can realize the precise closed-loop control of the metallurgical equipment, simulates the equipment response results under different control strategies through the digital twin model, generates the optimal control scheme, effectively improves the control precision and stability of the metallurgical equipment, and the fault diagnosis and self-healing control module combines the improved isolated forest algorithm and the voiceprint feature analysis technology, realizes the early identification and rapid self-healing of the metallurgical equipment fault, the improved isolated forest algorithm is integrated into the fault sample feature, improves the accuracy of fault identification, the voiceprint-vibration multi-modal fault identification model realizes the early warning of the fault through the real-time monitoring of the voiceprint signal, the self-healing control strategy can quickly take measures such as switching redundant equipment or adjusting process parameters when the fault occurs, and the influence of the fault on production is minimized, the production management interaction module is seamlessly connected with the production management system of the metallurgical enterprise, realizes the deep integration of the system and the enterprise production management system, the synchronization of multi-source data and the analysis results of the twin model, and the linkage of the equipment control function and the production plan function, effectively improves the production management efficiency and the coordination level of the metallurgical enterprise, shortens the linkage response time, and ensures the smooth execution of the production plan.

[0043] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and attained by the structure particularly pointed out in the written description and the accompanying drawings.

[0044] The technical solutions of the present application will be further described in detail below by means of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, for explaining the present application, and do not constitute a limitation on the present application. In the drawings:

[0046] Figure 1 It is a structure schematic diagram of the metallurgical equipment twin HMI control system based on multi-source data;

[0047] Figure 2 It is a method step schematic diagram for carrying out data fusion processing on multi-source data by adopting the space-time alignment algorithm and the feature extraction network.

[0048] Figure 3 A configuration schematic diagram of the augmented reality auxiliary operation and maintenance unit. DETAILED DESCRIPTION

[0049] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to explain and illustrate the present application, and are not used to limit the present application.

[0050] The present application provides a metallurgical equipment twin HMI control system based on multi-source data, as shown in the figure, the core is to build a closed-loop control system integrating data perception, model driving, man-machine collaboration, intelligent decision-making and execution, the metallurgical equipment twin HMI control system comprises: Figure 1 A multi-source data acquisition module is used to acquire multi-source data by using a configured sensor assembly, an environment monitoring terminal and a process parameter acquisition assembly; the multi-source data includes metallurgical equipment body operation data, production environment data and whole-process process data;

[0051] A data fusion processing module is used to perform data fusion processing on the multi-source data by using a space-time alignment algorithm and a feature extraction network to obtain fusion processing data;

[0052] A digital twin modeling module is used to construct a metallurgical equipment digital twin model based on the fusion processing data by using an improved deep belief network algorithm; the improved deep belief network algorithm introduces a metallurgical process constraint regularization term to constrain the model training process, so as to improve the compliance of the model to the metallurgical process rules and the prediction accuracy;

[0053] An HMI interaction module is used to dynamically present the fusion processing data and the running state data of the metallurgical equipment digital twin model based on the output of the digital twin modeling module;

[0054] An intelligent control module is used to perform closed-loop control on the metallurgical equipment by using an adaptive predictive control algorithm based on the fusion processing data and the running state data of the metallurgical equipment digital twin model in combination with real-time feedback of the multi-source data.

[0055] The working principle of the above technical solution is as follows: in order to realize the metallurgical equipment twin HMI control system based on multi-source data, first, the sensor assembly deployed by the multi-source data acquisition module captures the running parameters such as temperature, pressure and speed of the metallurgical equipment body in real time, the environment monitoring terminal synchronously acquires production environment data such as temperature, humidity and dust concentration in the workshop, and the process parameter acquisition assembly records whole-process process data from raw material input to finished product output, ensuring comprehensive coverage of multi-dimensional data;

[0056]

[0057] ​Multi-source data refers to a heterogeneous data set from different physical locations, different types of sensors and different business systems. In this system, it refers to a comprehensive data set that includes metallurgical equipment ontology operation data (such as temperature, pressure, vibration), production environment data (such as workshop temperature and humidity, dust concentration) and full-process process data (such as material composition, conveying speed, process energy consumption);

[0058] Then, the data fusion processing module uses the space-time alignment algorithm to calibrate the time axis and match the spatial position of data from different sources and different time scales. For the periodic characteristics of the metallurgical production process (such as rolling rhythm and smelting period), a period matching factor is designed to correct the data periodically to solve the data misplacement problem caused by period fluctuations. Then, the feature extraction network is used to mine the deep association features contained in the data and remove redundant information to form structured fusion processing data.

[0059] Subsequently, the digital twin modeling module is based on the fusion processing data and uses an improved deep belief network algorithm to construct a digital twin model. This algorithm introduces a metallurgical process constraint regularization term based on the traditional deep belief network. During the model training process, parameter combinations that do not meet the metallurgical process rules are penalized. This regularization term converts expert knowledge (such as "the smelting temperature of a certain alloy must be between 1500-1600℃") into a mathematical penalty term. When the model prediction result violates these hard process rules, it will be severely punished, forcing the model to learn the internal mapping relationship that meets the physical laws and process requirements. This can improve the compliance of the model, making the constructed digital twin model more accurately reflect the process requirements in actual production and improve the prediction accuracy of equipment operating status and process indicators.

[0060] The HMI interaction module visually presents the fusion-processed data in the form of visual charts, three-dimensional dynamic simulations, etc., and simultaneously displays the running state of the digital twin in real time, providing a clear and comprehensive monitoring interface for operators.

[0061] Finally, the intelligent control module integrates fusion processing data and twin running state data, combines real-time feedback information from multi-source data, and continuously optimizes control strategies through adaptive predictive control algorithms. Control instructions are issued to the actuators of metallurgical equipment based on the current operating conditions and future trends, achieving full-process automated management from data acquisition, fusion processing, model construction, state display to closed-loop control, effectively improving the operating stability and production efficiency of metallurgical equipment.

[0062] The beneficial effects of the above technical solution are: by using the scheme provided in this embodiment, the metallurgical equipment body, production environment and full-process process data can be comprehensively perceived through the multi-source data acquisition module; by integrating the deep correlation characteristics of data of different dimensions, the quality and usability of the data are significantly improved, laying a solid data foundation for building a high-precision digital twin model; the adaptability of the model to complex metallurgical process scenarios is enhanced, the control strategy can be automatically adjusted according to the changes in the equipment operation state, manual intervention is effectively reduced, and the stability of the equipment operation and the continuity of the production process are improved.

[0063] In one embodiment, the sensor assembly includes a temperature sensor, a pressure sensor and a vibration sensor; the temperature sensor is a distributed optical fiber temperature measurement device, which is arranged in a spiral around the high temperature area of the metallurgical equipment smelting furnace and is densely arranged in the easy erosion area of the furnace lining of the metallurgical equipment smelting furnace;

[0064] The metallurgical equipment body operation data includes real-time temperature, pressure and vibration frequency of key components of the metallurgical equipment; the production environment data includes workshop temperature and humidity, dust concentration and harmful gas content; the full-process process data includes material composition, conveying speed and energy consumption of each process.

[0065] The working principle of the above technical solution is: through the sensor assembly composed of a temperature sensor (distributed optical fiber temperature measurement device, arranged in a spiral around the high temperature area of the smelting furnace and densely arranged in the easy erosion area of the furnace lining), a pressure sensor and a vibration sensor, real-time metallurgical equipment body operation data such as real-time temperature, pressure and vibration frequency of key components of the metallurgical equipment are collected, and production environment data such as workshop temperature and humidity, dust concentration and harmful gas content, and full-process process data such as material composition, conveying speed and energy consumption of each process are obtained. These multi-dimensional data are understood and analyzed, so as to realize the monitoring of the running state of the metallurgical equipment and related work assistance.

[0066] The beneficial effects of the above technical solution are: by using the scheme provided in this embodiment, the metallurgical equipment body, production environment and full-process process data can be comprehensively perceived through the multi-source data acquisition module; by integrating the deep correlation characteristics of data of different dimensions, the quality and usability of the data are significantly improved, laying a solid data foundation for building a high-precision digital twin model; the adaptability of the model to complex metallurgical process scenarios is enhanced, the control strategy can be automatically adjusted according to the changes in the equipment operation state, manual intervention is effectively reduced, and the stability of the equipment operation and the continuity of the production process are improved.

[0067] In one embodiment, a spatio-temporal alignment algorithm and a feature extraction network are used for data fusion processing of multi-source data, as shown in Figure 2 , including:

[0068] The spatio-temporal alignment algorithm is used to clean, denoise and spatio-temporally register the multi-source data, and based on the designed period matching factor, the multi-source data is corrected in period synchronization to obtain corrected data;

[0069] The dynamic weight distribution mechanism is introduced to weight fusion processing on the corrected data to obtain basic fusion processing data; the dynamic weight distribution mechanism dynamically adjusts the data fusion weight of each data source according to the real-time reliability of the data source of the multi-source data, and automatically reduces the weight proportion when an abnormal fluctuation occurs in a data source;

[0070] Based on the constructed feature extraction network, the key feature vector reflecting the running state of the metallurgical equipment is obtained by performing feature extraction on the basic fusion processing data, and the fusion processing data is composed.

[0071] The working principle of the above technical solution is as follows: first, the multi-source data is cleaned, denoised and time-space registered by using a time-space alignment algorithm, and the period matching factor is combined to complete the period synchronization correction, so as to ensure the consistency and accuracy of the data in time and space dimensions; the time-space alignment algorithm is a data preprocessing technology, which aims to solve the inconsistency of multi-source data in time and space dimensions, and the core steps are: time alignment (through interpolation, resampling and other methods, the data of different sampling frequencies are unified to the same time axis) and space registration (through coordinate transformation, the sensor data from different physical positions are mapped to a unified device space coordinate system);

[0072] Then, a dynamic weight distribution mechanism is introduced, and the fusion weight of each data source is dynamically adjusted according to the real-time reliability, and the weight proportion is automatically reduced when an abnormal fluctuation occurs in a data source, so as to weight fusion on the corrected data to obtain basic fusion processing data; finally, the basic fusion processing data is deeply mined by the constructed feature extraction network, and the key feature vector reflecting the running state of the metallurgical equipment is extracted, so as to compose the final fusion processing data.

[0073] The beneficial effects of the above technical solution are as follows: by combining the time-space alignment algorithm and the period matching factor, the time-space misalignment problem caused by the difference in sensor sampling frequency, transmission delay and periodic fluctuation of equipment running in the collection process of multi-source data can be effectively solved, and the time consistency and spatial correlation of the data can be significantly improved.

[0074] In one embodiment, an improved deep belief network algorithm is used to construct a metallurgical equipment digital twin model, which includes:

[0075] A basic metallurgical equipment digital twin model is constructed by using a deep belief network algorithm;

[0076] In the training process of the basic metallurgical equipment digital twin model, a metallurgical process constraint regularization term is introduced, and its mathematical expression is:

[0077] l z = ε * ∑ | θ - θ0 | 2

[0078] In the above formula, l z represents the regularization loss, ε represents the regularization coefficient, controls the regularization strength, the larger the value, the stronger the regularization effect; ∑ represents the summation symbol, the regularization terms of all parameters of the model are accumulated; |θ-θ0| 2 represents the square of the Euclidean distance of the parameter θ and the reference value θ0.

[0079] The parameter of the digital twin model of the basic metallurgical equipment is secondarily corrected by introducing a metallurgical process expert rule base to obtain a digital twin model of the metallurgical equipment; the metallurgical process expert rule base includes but is not limited to a metallurgical equipment overload protection logic and a material adaptation process interval; the metallurgical process expert rule base is configured to be dynamically updated by an incremental update algorithm based on a graph database Neo4j; wherein, the knowledge graph is integrated with a metallurgical equipment fault tree in the construction process; the update frequency of the metallurgical process expert rule base is associated with the fault handling closed loop cycle of the metallurgical equipment;

[0080] In actual operation, first, the expert rule base constructed by experts in the metallurgical field according to long-term production experience and process knowledge is structured to determine the reasonable value range of each process parameter, the determination threshold of the equipment operating state, and the operation constraint conditions under different working conditions, etc., to form specific constraint boundary data; when the improved deep belief network algorithm is used to train the digital twin model of the basic metallurgical equipment, when the intermediate layer parameters or output results of the model exceed the constraint boundary set by the expert rule base in the back propagation process, a corresponding penalty value is applied to the out-of-bound parameters through a pre-set penalty mechanism, and the penalty value will increase with the increase of the out-of-bound degree of the parameters, prompting the model to continuously adjust the parameters in the subsequent iteration training, so as to converge to the reasonable range specified by the expert rule base, thereby deeply integrating the expert experience into the model training process in a quantifiable way.

[0081] The working principle of the above technical solution is as follows: first, a basic model framework is built by using an improved deep belief network algorithm, and a plurality of source information such as historical operation data of metallurgical equipment and real-time collected data of sensors is inputted for preliminary training, so that the model has the simulation capability for the basic operation state of the equipment; the metallurgical process constraint regular term is an additional term introduced in the loss function of the machine learning model (such as the deep belief network), which quantifies the hard process rules in the metallurgical field (such as the allowable range of temperature, pressure and composition) into a mathematical penalty term. When the prediction result or internal parameter of the model violates these rules, the regular term will increase the loss value, thereby guiding the model to learn the solution that meets the process requirements during the model training process, and improving the physical rationality and compliance of the model; then, a metallurgical process expert rule library is introduced for secondary optimization, the rule library is constructed based on the expert experience and process standards in the metallurgical field, and covers the current, temperature safety threshold in different working conditions in the overload protection logic, and the composition ratio and reaction time in the material adaptation process interval and other core rules. By converting these rules into constraint conditions recognizable by the model, the parameters of the first digital twin model are adjusted in a targeted manner, so that the model can not only reflect the operation law of the physical layer of the equipment, but also meet the process requirements in actual production; the application of the knowledge graph technology provides support for the dynamic update of the expert rule library. When the knowledge graph is constructed, the fault tree analysis results are integrated, and the knowledge such as equipment fault type, cause and treatment measures is stored in the form of a structured graph. When the metallurgical equipment fails and completes the closed-loop processing, the knowledge graph update process is automatically triggered according to the fault processing period, and the related process parameters and protection logic in the expert rule library are updated synchronously, so that the rule library always keeps consistent with the actual operation state of the equipment and the process optimization direction, thereby enabling the final constructed digital twin model of the metallurgical equipment to accurately simulate the operation state of the equipment under complex working conditions, and providing reliable model support for the HMI control system to realize real-time monitoring of the equipment state, fault warning and process optimization.

[0082] The beneficial effects of the above technical solution are as follows: by using the scheme provided in the embodiment, the simulation accuracy and process adaptability of the digital twin model for the operation state of the metallurgical equipment are significantly improved by combining the improved deep belief network algorithm with the metallurgical process expert rule library.

[0083] In one embodiment, the digital twin modeling module further includes a multi-physical field coupling correction unit, which is configured to perform the following operations:

[0084] A physical mechanism model is constructed based on thermodynamic-hydrodynamic-structural mechanics coupling equations, and temperature, flow rate and stress monitoring values in the fusion processing data are received in real time;

[0085] When the output of the digital twin model of the metallurgical equipment deviates from the simulation result of the physical mechanism model by more than a preset threshold, a parameter recalibration mechanism is triggered; the parameter recalibration mechanism corrects the weights of the deep belief network by using a physical field loss function through a back propagation algorithm, and outputs a calibrated digital twin model of the metallurgical equipment; the physical field loss function measures the degree of coincidence between the model output and the physical law by calculating the sum of the weighted square errors between the model prediction value and the physical actual value, and minimizing the loss during training can improve the physical consistency of the model; the expression of the physical field loss function is:

[0086]

[0087] In the above formula, l phy represents a physical loss function for quantifying the error between the model output and the physical actual data; ∑ i represents the sum of all samples i; k i represents the weight coefficient of sample i, used to adjust the importance of different samples in loss calculation; represents the square of the norm, where is the prediction output vector of the deep belief network model for sample i, is the physical observation data vector of sample i.

[0088] The working principle of the above technical solution is as follows: first, the multi-physical field coupling correction unit integrates the related parameters of thermodynamics, fluid mechanics and structural mechanics according to the key physical characteristics in the operation process of the metallurgical equipment, and constructs a coupling equation that can reflect the real physical behavior of the equipment, which is used as the core framework of the physical mechanism model; the multi-physical field coupling correction unit is a functional module for calibrating and improving the precision of the digital twin model, which is constructed based on the physical laws of thermodynamics, fluid mechanics, structural mechanics, etc., and reflects the physical mechanism model of the real physical behavior of the equipment. When the output of the data-driven digital twin model deviates significantly from the simulation results of the physical mechanism model, the unit will trigger a calibration mechanism to correct the parameters of the data-driven model in reverse using the observation data of the physical field, and ensure that the digital twin model is consistent with the physical law; during system operation, the unit continuously receives real-time monitoring data such as temperature field distribution, fluid flow rate change and structural stress from the fusion processing module, and uses these data as input conditions for the physical mechanism model for simulation operation; when the deviation between the output of the digital twin model (such as the temperature prediction value of the key components, the structural deformation amount, etc.) and the simulation results of the physical mechanism model exceeds the set threshold value, the parameter re-calibration mechanism is immediately started; at this time, the system calls the back propagation algorithm to dynamically adjust the weights of each layer of the deep belief network with the physical field loss function as the optimization target; during the weight correction process, the physical field loss function calculates the weighted sum of squared errors between the prediction output of the digital twin model and the simulation results of the physical mechanism model, accurately quantifying the deviation between the model output and the physical law, wherein the sample weight coefficient is dynamically allocated according to the reliability of the data, the importance of the monitoring points and other factors, to ensure that the data of the key physical process occupies a higher weight in the loss calculation; through continuous iterative training, the value of the physical field loss function is minimized, and the weight parameters of the deep belief network gradually converge to the optimal solution, and finally the calibrated digital twin model of the metallurgical equipment is output, thereby effectively improving the precision of the model in describing the physical behavior of the equipment, and ensuring that the digital twin model can accurately reflect the real running state of the metallurgical equipment under the complex multi-physical field coupling effect;

[0089] In a specific implementation, in a digital twin system of a hot rolling production line of a certain steel enterprise, a temperature field monitoring unit receives temperature data (sampling frequency 1 Hz) returned by 128 thermocouples in a slab heating furnace, flow rate change data of a circulating airflow speed sensor array (32 monitoring points) in the furnace, and stress values collected by strain gauges (64 measuring points) of a furnace body load-bearing structure. These data are used as input to drive a physical mechanism model (including a coupled solver of radiation heat transfer, convection heat transfer, and elastoplastic mechanics equations). The slab temperature field distribution cloud map, the furnace roller thermal deformation amount (unit: mm), and the furnace lining structure stress value (unit: MPa) are output every 5 minutes. When the deviation between the slab discharge temperature predicted by the digital twin model (set threshold ± 5℃) and the measured value of the infrared temperature measuring instrument exceeds 8℃ for 3 consecutive sampling periods, or the maximum deformation amount of the furnace roller (threshold 0.3mm) deviates from the measured value of the laser displacement meter by 0.05mm, the system automatically triggers parameter recalibration. At this time, the back propagation algorithm based on the TensorFlow framework is called to adjust the synaptic weights of an 8-layer deep belief network (input layer 128 neurons, feature layer 256-128-64 decreasing, output layer 32 neurons) with the physical field loss function as the optimization target. The sample data of the slab center temperature measuring point and the furnace roller bearing seat stress monitoring point are iteratively trained 1000 times by the Adam optimizer (learning rate 0.001, decay rate 0.9) to make the physical field loss function value converge from the initial 0.85 to below 0.02. The calibrated model finally outputs the slab temperature prediction deviation within ± 3℃, and the furnace roller deformation prediction accuracy is improved to 92%, ensuring that the digital twin system can accurately reproduce the multi-physical field coupling behavior of the 2250mm wide and thick slab heating furnace under the conditions of coal gas flow fluctuation (± 15%) and slab size switching (thickness 200-300mm).

[0090] The technical scheme has the beneficial effects that: by adopting the scheme provided in the embodiment, a multi-physical field coupled physical mechanism model is constructed, and when the deviation between the output of the digital twin model and the simulation result of the physical mechanism model exceeds the threshold, a back propagation algorithm based on a physical field loss function (sum of weighted square errors) is started to correct the weights of a deep belief network, the model is dynamically calibrated, the accuracy of the description of the physical behavior of the metallurgical equipment under the complex multi-physical field coupling is improved, and it is ensured that the digital twin model can accurately reflect the real running state of the equipment.

[0091] In one embodiment, as shown in Figure 3 The HMI interaction module includes an augmented reality assisted operation and maintenance unit;

[0092] The augmented reality assisted operation and maintenance unit is configured to superimpose and display the internal structure of the metallurgical equipment in the form of augmented reality in the real space based on the digital twin model of the metallurgical equipment. The internal structure of the metallurgical equipment includes at least one of a rolling mill gear box transmission structure and a smelting furnace airflow channel.

[0093] The augmented reality auxiliary operation and maintenance unit is further configured to enable the operation and maintenance personnel to intuitively view the running state of hidden components of the metallurgical equipment through the augmented reality display device; the running state of the hidden components includes at least one of temperature and vibration distribution;

[0094] The augmented reality auxiliary operation and maintenance unit is further configured to realize spatial registration between the augmented reality display and the actual position of the metallurgical equipment.

[0095] The working principle of the above technical solution is that: the working principle of the technical solution is based on a digital twin model of the metallurgical equipment, and through the augmented reality auxiliary operation and maintenance unit, the internal structure of the metallurgical equipment based on the model (such as the transmission structure of the rolling mill gear box, the airflow channel of the smelting furnace, etc.) is displayed in the form of augmented reality in the real space, so that the operation and maintenance personnel can intuitively view the running state (such as temperature, vibration distribution, etc.) of the hidden components (such as internal transmission structure, airflow channel, etc.) through the augmented reality display device;

[0096] The augmented reality auxiliary operation and maintenance unit is a subsystem in the HMI interaction module, which uses augmented reality (AR) technology to accurately superimpose (display) the digital twin model of the metallurgical equipment (especially the real-time running state of its internal structure and hidden components, such as temperature, vibration distribution) in the form of a virtual image on the real equipment seen by the operation and maintenance personnel through the AR display device (such as AR glasses), realizing intuitive operation and maintenance guidance of virtual and real fusion.

[0097] At the same time, the augmented reality auxiliary operation and maintenance unit also completes the spatial registration of the augmented reality display and the actual position of the metallurgical equipment, ensuring that the position of the augmented reality display accurately corresponds to the real position of the metallurgical equipment, so that the augmented reality information seen by the operation and maintenance personnel can be accurately associated with the actual equipment part, thereby more accurately and effectively performing operation and maintenance work.

[0098] The beneficial effects of the above technical solution are: by adopting the scheme provided in the embodiment, the operation and maintenance personnel can break through the observation limitations of internal structure and hidden component running state in the traditional operation and maintenance mode, and intuitively and clearly grasp the real-time situation of key components without disassembling the equipment; this visualization method not only greatly reduces the cognitive difficulty of the operation and maintenance personnel on the internal structure of the equipment, reduces the misjudgment risk caused by unclear information, but also significantly improves the efficiency and accuracy of operation and maintenance work.

[0099] In one embodiment, the adaptive predictive control algorithm of the intelligent control module is configured to:

[0100] The equipment running trend prediction model is constructed, and a long short-term memory network is used for time series analysis of multi-source data, and the prediction period is adjusted according to the metallurgical process rhythm;

[0101] Generate an optimal control scheme in combination with the equipment response results under different control strategies simulated by the digital twin model of the metallurgical equipment, wherein the equipment response results include product thickness changes after adjusting the rolling mill speed.

[0102] The working principle of the above technical solution is that the adaptive predictive control algorithm is a closed-loop control strategy that combines the predictive ability of a prediction model (LSTM) for future system behavior and an optimization algorithm. The algorithm can dynamically evaluate the effects of multiple control strategies based on the current system state and predicted future trends, and select the scheme that optimally meets the control objectives (such as product quality and energy consumption) from among them, achieving forward-looking and adaptive intelligent control.

[0103] The intelligent control module first preprocesses the multi-source data during the operation of the metallurgical equipment, including data cleaning, outlier removal, and feature extraction, to ensure the quality of the data input into the long short-term memory network. The long short-term memory network, with its good modeling ability for time series data, can capture the hidden dynamic change rules in multi-source data and accurately predict the running trends of the equipment at different time nodes in the future, such as the load fluctuations of the rolling mill and the temperature changes of the smelting furnace, by learning from historical operation data. The dynamic adjustment of the prediction period closely follows the rhythm of the metallurgical process. When the process is in the high-speed rolling stage, the prediction period is shortened to achieve more real-time trend tracking, while in the low rhythm period of stable equipment operation, the prediction period is appropriately extended to reduce the consumption of computing resources. At the same time, the digital twin model simulates the responses of the equipment under different control strategies based on the current equipment state and predicted running trends. For example, when it is predicted that the product thickness may deviate, the changes in product thickness corresponding to multiple control schemes such as adjusting the rolling mill speed and changing the rolling force are simulated. By comparing the response results under different schemes, such as product quality indicators, equipment energy consumption, and production efficiency, the most comprehensive optimal control scheme is selected, and the metallurgical equipment is driven to perform the corresponding control actions, achieving precise regulation and control of the equipment operating state.

[0104] The beneficial effects of the above technical solution are that the scheme provided in this embodiment can achieve accurate prediction of the running trend of the metallurgical equipment by means of deep mining and processing of multi-source data by the intelligent control module, combined with the time series prediction ability of the long short-term memory network, and provide a reliable basis for the formulation of equipment control strategies.

[0105] In one embodiment, a fault diagnosis and self-healing control module is further included; the fault diagnosis and self-healing control module is configured to: based on the equipment operation state data output by the metallurgical equipment digital twin model, adopt an improved isolation forest algorithm for fault identification; the improved isolation forest algorithm is integrated with metallurgical equipment fault sample features, including vibration frequency feature spectrum when the bearing is faulty; after identifying the fault, automatically trigger a self-healing control strategy, which includes switching redundant equipment and / or adjusting process parameters for compensation;

[0106] The metallurgical equipment fault sample features are extracted based on a constructed metallurgical equipment fault sample library, which uses a transfer learning technology to perform cross-line transfer training on historical fault data of the same type of metallurgical equipment, so as to expand the sample coverage and improve the model generalization ability; the metallurgical equipment fault sample library is periodically updated, and the update content includes new real-time fault data, failure mode correction and feature weight optimization.

[0107] The working principle of the above technical solution is as follows: first, the fault diagnosis and self-healing control module continuously receives the equipment operation state data output by the metallurgical equipment digital twin model in real time, which covers parameters in multiple dimensions such as vibration, temperature, pressure, and current; then, the module calls the improved isolation forest algorithm to process the received data, since the algorithm has integrated the metallurgical equipment fault sample features, especially the key information such as the vibration frequency feature spectrum when the bearing is faulty, it can quickly and accurately identify abnormal patterns from normal data, thereby achieving accurate positioning and type judgment of potential faults; once a fault is identified, the module immediately selects the appropriate self-healing control measures according to the preset self-healing control strategy library, combined with the fault type, severity, and current production conditions; if the fault involves key equipment and there is a redundant configuration, the system will quickly switch to the redundant equipment to ensure that the production process does not stop; if the fault can be compensated by adjusting the process parameters, the relevant process parameters such as smelting temperature, raw material ratio, and running speed are adjusted in real time to offset the impact of the fault on equipment performance and product quality; at the same time, the metallurgical equipment fault sample library plays an important supporting role in the background, which uses a transfer learning technology to perform cross-line transfer training on historical fault data of the same type of metallurgical equipment, effectively solving the problem of insufficient fault samples on a single production line, greatly expanding the sample coverage, and improving the generalization recognition ability of the model for faults under different production lines and different conditions; in addition, the sample library is updated according to the set period, continuously incorporating new real-time fault data, correcting existing failure modes, and optimizing feature weights, so that the improved isolation forest algorithm can always identify faults based on the latest and most comprehensive fault sample features, further ensuring the accuracy of fault diagnosis and the effectiveness of self-healing control strategies.

[0108] The beneficial effects of the above technical solution are: by using the scheme provided in this embodiment, the whole-process closed-loop management from equipment running state monitoring, accurate fault identification to automatic triggering of self-healing control is realized, the running reliability and stability of the metallurgical equipment are significantly improved, the unplanned downtime is reduced, and the continuous and efficient production is ensured.

[0109] In one embodiment, the fault diagnosis and self-healing control module includes a voiceprint feature analysis submodule, which is configured to perform the following operations:

[0110] Based on the high-heat-resistant acoustic sensors deployed at key positions of the metallurgical equipment such as bearings and gearboxes, the voiceprint signal data of the equipment running is collected;

[0111] Based on the Mel frequency cepstrum coefficients extracted from the voiceprint signal data, a voiceprint-vibration multi-modal fault identification model is constructed by combining an improved isolated forest algorithm;

[0112] Based on the real-time dynamic monitoring of the voiceprint features by the voiceprint-vibration multi-modal fault identification model, if the voiceprint features deviate from the normal range and meet the fault determination conditions, a fault warning signal is generated.

[0113] The working principle of the above technical solution is: by deploying high-heat-resistant acoustic sensors at key positions of the metallurgical equipment such as bearings and gearboxes, the voiceprint signal data during the equipment running is continuously collected to provide raw data input for fault diagnosis; the collected voiceprint signal data is processed to extract Mel frequency cepstrum coefficients that can effectively represent the equipment state, and the Mel frequency cepstrum coefficients are combined with an improved isolated forest algorithm to construct a voiceprint-vibration multi-modal fault identification model, which can fuse voiceprint feature information and improve the recognition ability of the equipment fault mode; the voiceprint-vibration multi-modal fault identification model is a composite fault diagnosis model that fuses acoustic signals and vibration signals, which extracts voiceprint features (such as Mel frequency cepstrum coefficients MFCC) and vibration signal features during the equipment running, and uses improved isolated forest algorithms to analyze the data of the two modalities at the same time to improve the recognition sensitivity and accuracy of early and weak faults; the constructed voiceprint-vibration multi-modal fault identification model is used to perform real-time dynamic monitoring of the voiceprint features during the equipment running, and when the voiceprint features deviate from the pre-set normal range and the degree of deviation meets the pre-set fault determination conditions, the voiceprint feature analysis submodule of the fault diagnosis and self-healing control module generates a fault warning signal to timely remind relevant personnel to handle or trigger the subsequent self-healing control process.

[0114] The beneficial effects of the above technical solution are: by using the scheme provided in this embodiment, the scheme provided in this embodiment, the key parts of the metallurgical equipment are continuously monitored by high-heat-resistant acoustic sensors, which can stably obtain voiceprint signals in a complex industrial environment with high temperature and high vibration, and lay a reliable data foundation for fault diagnosis.

[0115] In one embodiment, a production management interaction module is further included, which is configured to interact with the metallurgical enterprise production management system; the production management interaction module is configured to: support synchronization of the multi-source data collected by the multi-source data collection module and the analysis results of the twinborn model output by the digital twin modeling module to the metallurgical enterprise production management system; realize linkage between the equipment control function of the metallurgical equipment twinborn HMI control system based on multi-source data and the production planning function of the metallurgical enterprise production management system through the production management interaction module; receive the production task instruction of the metallurgical enterprise production management system, and the production task instruction includes the target product specification, the delivery cycle and the energy consumption index;

[0116] The process parameter self-optimization engine is further included, which is configured to: based on the dynamic knowledge graph constructed based on the metallurgical process expert rule base, analyze the production task instruction, and generate a plurality of candidate process paths; call the metallurgical equipment digital twin model, simulate each candidate process path, and predict the product quality, equipment wear and tear and comprehensive energy consumption under each path; based on a preset multi-objective optimization function, filter out the optimal process parameter combination from the candidate paths; issue the optimal process parameter combination as the initial control parameter to the intelligent control module, and fine-tune according to the real-time feedback data during the production process to realize dynamic self-optimization of the process parameters; wherein the dynamic knowledge graph is learned by a graph neural network, the nodes thereof include equipment state, process parameters, material properties, environmental factors, product quality and energy consumption, and the edges represent the causal relationship and influence weight between the nodes; the weight of the dynamic knowledge graph is continuously updated based on the deviation between the simulation results of the twinborn model and the actual production data through an online learning mechanism; the objectives of the multi-objective optimization function include quality, cost, efficiency and energy consumption.

[0117] The working principle of the technical solution is as follows: the production management interaction module first establishes a standardized data communication link with the metallurgical enterprise production management system, uses an OPC UA or MQTT industrial communication protocol to ensure the real-time and reliability of data transmission; the production management interaction module is a standardized communication channel for data exchange and function collaboration between the system and the external system (metallurgical enterprise production management system), supports bidirectional data transmission, can synchronize real-time data and analysis results of the system to the external system, can also receive control instructions or production plans of the external system, and converts them into internal operations of the system, realizing cross-system linkage; in the data synchronization process, the production management interaction module performs format conversion and standardization processing on the equipment operation parameters, fault warning information collected by the multi-source data acquisition module, and the equipment performance simulation data and life prediction results generated by the digital twin modeling module, so as to meet the data receiving requirements of the metallurgical enterprise production management system; when the metallurgical enterprise production management system issues a new production plan instruction, the production management interaction module can receive and analyze the instruction in real time, combines the current running state of the equipment, capacity bottleneck and other information in the twin model analysis result, performs feasibility checking on the production plan, and feeds back the checking result to the metallurgical enterprise production management system; if the checking is passed, the production management interaction module decomposes the production plan into specific equipment control parameters, and transmits them to the execution module of the twin HMI control system, realizing dynamic matching of equipment control and production plan; if the checking finds that the production plan conflicts with the actual capacity of the equipment, the production management interaction module sends an adjustment suggestion to the metallurgical enterprise production management system, assists the production management personnel in optimizing the production plan, so as to realize close linkage of the equipment control function and the production plan function, and improve the overall production scheduling efficiency and resource utilization rate of the metallurgical enterprise;

[0118] The process parameter self-optimization engine, when analyzing production task instructions, can accurately locate key process parameters that affect the specifications of target products through the association between nodes in the dynamic knowledge graph. For example, in the rolling process, it focuses on the mapping relationship between core parameters such as rolling temperature, reduction rate, and rolling speed and quality indicators such as product thickness and hardness. When calling the digital twin model for simulation, finite element analysis and multi-physical field coupling algorithms are used to simulate the stress distribution of equipment components, temperature field changes, and material flow states under different process paths, accurately predicting product microstructure performance, surface quality defects, and other detailed indicators. Meanwhile, combining with the wear model constructed based on historical operation data of the equipment, the bearing wear and roll fatigue life of each candidate path are calculated, and based on real-time data collected by energy consumption monitoring sensors, an energy consumption calculation model is established to accurately estimate the comprehensive energy consumption of electricity, water resources, and gas. When constructing the multi-objective optimization function, the weight coefficients of each target are dynamically adjusted according to the strategic goals of the metallurgical enterprise and the current market demand. For example, when the enterprise faces delivery cycle pressure, the efficiency target weight is appropriately increased, and when environmental protection policies are tightened, the energy consumption target weight is increased. In the process of selecting the optimal process parameter combination, an improved genetic algorithm is used to quickly converge to the Pareto optimal solution set through adaptive crossover and mutation operators and elite preservation strategies, and the parameter combination with the best comprehensive performance is selected. During production, the process parameter self-optimization engine receives real-time equipment operation data (such as motor current and hydraulic system pressure), material detection data (such as composition analysis results and temperature measured values), and product online detection data (such as laser thickness gauge and X-ray flaw detection results) from the intelligent control module. Through the online learning mechanism of the dynamic knowledge graph, the influence weights between process parameters and product quality and energy consumption are continuously corrected. When product quality fluctuations or energy consumption exceed the threshold, the parameter fine-tuning mechanism is automatically triggered. For example, when the product thickness deviation exceeds ±0.05mm, the rolling speed or reduction amount is immediately adjusted to ensure that the production process is always in an optimal control state. Ultimately, the metallurgical production is transformed from experience-driven to data-driven intelligent transformation, significantly reducing process debugging time in the production process, reducing waste caused by unreasonable parameters, and maximizing energy consumption costs per product, enhancing the core competitiveness of enterprises in market competition.

[0119] The beneficial effects of the technical scheme are as follows: by using the scheme provided in the embodiment, an efficient collaborative communication link can be constructed between the production management interaction module and the metallurgical enterprise production management system, seamless integration and real-time interaction of multi-source heterogeneous data can be realized by means of a standardized data processing mechanism, and accurate docking of equipment operation data, simulation data and production plan information can be ensured; this integration process effectively breaks the data barrier between the equipment control layer and the production management layer in traditional metallurgical production, and avoids the production scheduling lag problem caused by information islands; the process parameter self-optimization engine realizes accurate positioning and multi-dimensional influence analysis of key process parameters through the deep integration of dynamic knowledge graph and digital twin simulation, and provides comprehensive data support for multi-objective optimization decision-making; the combination of real-time data feedback and dynamic knowledge graph online learning mechanism enables the system to have the ability of continuous self-optimization, can respond to various fluctuations in the production process in a timely manner, automatically trigger parameter fine-tuning, ensure that the production is always in a stable and efficient optimal state, and change the traditional mode of relying on manual experience to adjust process parameters.

[0120] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A twin HMI control system for metallurgical equipment based on multi-source data, characterized in that, The method comprises the following steps: A multi-source data acquisition module is used to collect multi-source data by using a configured sensor assembly, an environmental monitoring terminal and a process parameter acquisition assembly; The multi-source data includes metallurgical equipment body operation data, production environment data and full-process process data; A data fusion processing module is used to perform data fusion processing on the multi-source data by using a time-space alignment algorithm and a feature extraction network to obtain fusion processing data; A digital twin modeling module is used to construct a metallurgical equipment digital twin model based on the fusion processing data by using an improved deep belief network algorithm; the improved deep belief network algorithm introduces a metallurgical process constraint regularization term to constrain the model training process to improve the compliance of the model to the metallurgical process rules and the prediction accuracy; An HMI interaction module is used to dynamically present the fusion processing data and the running state data of the metallurgical equipment digital twin model based on the output of the digital twin modeling module; An intelligent control module is used to perform closed-loop control on the metallurgical equipment by using an adaptive predictive control algorithm based on the fusion processing data and the running state data of the metallurgical equipment digital twin model in combination with real-time feedback of the multi-source data.

2. The multi-source data based metallurgical plant twin HMI control system of claim 1, wherein, The sensor assembly includes temperature sensors, pressure sensors and vibration sensors; the temperature sensors are distributed optical fiber temperature measurement devices arranged in a spiral around the high-temperature area of the metallurgical equipment smelting furnace and densely arranged in the easy-erosion area of the metallurgical equipment smelting furnace lining; The metallurgical equipment body operation data includes real-time temperature, pressure and vibration frequency of key components of the metallurgical equipment; the production environment data includes workshop temperature and humidity, dust concentration and harmful gas content; the full-process process data includes material composition, conveying speed and energy consumption of each process.

3. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, The time-space alignment algorithm and the feature extraction network are used to perform data fusion processing on the multi-source data, including: The time-space alignment algorithm is used to clean, denoise, time-space register and periodically synchronize and correct the multi-source data based on a designed period matching factor to obtain corrected data; A dynamic weight distribution mechanism is introduced to perform weight fusion processing on the corrected data to obtain basic fusion processing data; the dynamic weight distribution mechanism dynamically adjusts the data fusion weight of each data source according to the real-time reliability of the data source of the multi-source data, and automatically reduces the weight proportion when an abnormal fluctuation occurs in a certain data source; The feature extraction network is used to extract features from the basic fusion processing data to obtain a key feature vector reflecting the running state of the metallurgical equipment to form the fusion processing data.

4. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, The improved deep belief network algorithm is used to construct the metallurgical equipment digital twin model, including: The improved deep belief network algorithm introduces a metallurgical process constraint regularization term to constrain the training process and modifies the model parameters based on a metallurgical process expert rule base to construct the metallurgical equipment digital twin model; the metallurgical process expert rule base includes metallurgical equipment overload protection logic and material adaptation process interval rules and is configured to be dynamically updated.

5. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, The digital twin modeling module further comprises a multi-physical field coupling correction unit, which is used to perform the following operations: A physical mechanism model is constructed based on thermodynamic-hydrodynamic-structural mechanics coupling equations, and real-time receiving and fusion processing data of temperature, flow rate, and stress monitoring values; When the output of the metallurgical equipment digital twin model deviates from the simulation result of the physical mechanism model by more than a preset threshold, a parameter re-calibration mechanism is triggered; The parameter re-calibration mechanism corrects the deep belief network weight by a physical field loss function through a back propagation algorithm, and outputs an optimized metallurgical equipment digital twin model.

6. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, The HMI interaction module includes an augmented reality auxiliary operation and maintenance unit; The augmented reality auxiliary operation and maintenance unit is configured to superimpose and display the internal structure of the metallurgical equipment in the form of augmented reality on the real space based on the metallurgical equipment digital twin model; the internal structure of the metallurgical equipment includes at least one of a rolling mill gear box transmission structure and a smelting furnace gas flow channel; The augmented reality auxiliary operation and maintenance unit is further configured to enable the operation and maintenance personnel to intuitively view the running state of the hidden components of the metallurgical equipment through the augmented reality display device; the running state of the hidden components includes at least one of temperature and vibration distribution; The augmented reality auxiliary operation and maintenance unit is further configured to realize spatial registration between the augmented reality display and the actual position of the metallurgical equipment.

7. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, The adaptive predictive control algorithm of the intelligent control module is configured to: Construct a device operation trend prediction model, use a long short-term memory network to perform time series analysis on multi-source data, and adjust the prediction period according to the metallurgical process rhythm; Generate an optimal control scheme in combination with the device response results under different control strategies simulated by the metallurgical equipment digital twin model, wherein the device response results include product thickness changes after adjusting the rolling mill speed.

8. The multi-source data based twin HMI control system for metallurgical equipment of claim 1, wherein, Further comprising a fault diagnosis and self-healing control module; the fault diagnosis and self-healing control module is configured to: based on the device running state data output by the metallurgical equipment digital twin model, use an improved isolated forest algorithm incorporating metallurgical equipment fault sample features for fault identification; automatically trigger a self-healing control strategy after identifying the fault, which includes switching redundant devices and / or adjusting process parameters for compensation.

9. The multi-source data based twin HMI control system for metallurgical equipment of claim 8, wherein, The fault diagnosis and self-healing control module includes a voiceprint feature analysis submodule, which is used to perform the following operations: Based on high-heat-resistant acoustic sensors deployed at key positions such as bearings and gear boxes of the metallurgical equipment, collect voiceprint signal data of the device operation; Based on the Mel-frequency cepstral coefficients extracted from the voiceprint signal data, construct a voiceprint-vibration multi-modal fault identification model in combination with the improved isolated forest algorithm; Based on real-time dynamic monitoring of voiceprint features by the voiceprint-vibration multi-modal fault identification model, if the voiceprint features deviate from the normal range and meet the fault judgment conditions, a fault warning signal is generated.

10. The multi-source data based twin HMI control system for metallurgical equipment of claim 4, wherein, Further comprising a production management interaction module, which is configured to interact with the production management system of the metallurgical enterprise; The production management interaction module is configured to: support synchronizing the multi-source data collected by the multi-source data acquisition module and the twin model analysis results output by the digital twin modeling module to the production management system of the metallurgical enterprise; The device control function of the metallurgical equipment twin HMI control system based on multi-source data is linked with the production plan function of the metallurgical enterprise production management system through a production management interaction module; a production task instruction is received from the metallurgical enterprise production management system, and the production task instruction includes a target product specification, a delivery cycle and an energy consumption index; The process parameter self-optimization engine is configured to: based on a dynamic knowledge graph constructed based on a metallurgical process expert rule base, analyze the production task instruction, and generate a plurality of candidate process paths; The metallurgical equipment digital twin model is called to simulate each candidate process path, predict the product quality, equipment wear and tear and comprehensive energy consumption under each path, select an optimal process parameter combination from the candidate paths based on a preset multi-objective optimization function, and issue the optimal process parameter combination as initial control parameters to the intelligent control module, and fine-tune according to real-time feedback data during the production process to realize dynamic self-optimization of the process parameters; wherein the dynamic knowledge graph is learned using a graph neural network, the nodes of which include device state, process parameter, material attribute, environmental factor, product quality and energy consumption, and the edges represent the causal relationship and influence weight between the nodes; the weight of the dynamic knowledge graph is continuously updated based on the deviation between the simulation results of the twin model and the actual production data through an online learning mechanism.

Images