Metallurgy fan multi-physics field coupling and dynamic performance optimization method

Abstract

The invention belongs to the field of industrial automation and intelligent manufacturing, particularly provides a metallurgical fan multi-physics field coupling and dynamic performance optimization method based on digital twinning, and aims to solve the technical problems of performance optimization and fault prediction of a metallurgical fan under complex operation conditions. The method comprises the following steps: acquiring multi-physical field data such as fluid mechanics, thermodynamics and structural mechanics through a distributed sensing network and a real-time data acquisition technology, and performing cleaning, denoising and feature extraction on the data; the method comprises the following steps: constructing a high-precision multi-physics field coupling digital twin model based on a geometric structure and material characteristics of a metallurgical fan, and carrying out model calibration in combination with historical data and real-time sensing data; the dynamic performance of the draught fan under different operation conditions is simulated through the co-simulation technology, the operation parameters of the draught fan are dynamically adjusted through a data-driven optimization algorithm, real-time optimization of the performance is achieved, and the purposes of improving the operation efficiency, reducing energy consumption and prolonging the service life of equipment are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation and intelligent manufacturing, and in particular to a method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan. Background Art

[0002] In the metallurgical industry, fans are key power equipment, and their operating status directly determines the efficiency, stability, and equipment reliability of the production process. However, traditional fan performance optimization and fault management methods are difficult to meet the needs of modern industrial intelligence and efficient operation. Existing technologies usually rely on regular maintenance and offline analysis, such as setting fixed operating parameters based on historical experience. This fixed parameter setting cannot fully adapt to the dynamic changes of fans under different operating conditions, and especially exhibits significant limitations under complex and variable working conditions. In addition, traditional technologies mainly rely on manual inspection and regular maintenance. This lag can easily lead to problems being discovered only after equipment failure, resulting in irreparable losses to production efficiency and economic benefits.

[0003] The core technical challenges faced in optimizing wind turbine performance are determined by the characteristics of wind turbine operation, which involves the complex coupling of multiple physical fields (such as fluid mechanics, thermodynamics, and structural mechanics). For example, fluid impact not only triggers dynamic deformation of the blades, but the deformation of the blades also negatively affects aerodynamic performance. Similarly, the non-uniform distribution of the temperature field can lead to thermal stress concentration, accelerating fatigue damage to key components and thus shortening equipment life. Traditional optimization methods usually model and analyze a single physical field. This one-sided optimization model ignores the complex interactive characteristics of multi-field coupling, resulting in a significant gap between the optimization results and the actual operating conditions.

[0004] With the improvement of the intelligence and automation level of industrial equipment, data-driven technology and intelligent optimization algorithms have shown significant advantages in the field of equipment performance optimization, especially in the research of fan systems. At present, the focus of research is mostly on improving operating efficiency, mostly with a single optimization goal as the core. The research on multi-objective balance optimization between efficiency, energy consumption and life has not yet formed a theoretically complete and practical system framework. At the same time, existing methods are generally based on static optimization models, lacking real-time monitoring and dynamic feedback mechanisms, and are difficult to adapt to the parameter adaptive adjustment requirements under the complex operating conditions of metallurgical fans. These technical bottlenecks significantly limit the practical application of existing technologies in the dynamic performance optimization of metallurgical fans, and cannot meet the industrial needs of efficient and stable operation of equipment.

[0005] To address these issues, digital twin technology has emerged and demonstrated unique advantages in the field of industrial equipment performance optimization. By constructing a virtual model that is completely consistent with the physical system, combined with real-time data acquisition and high-precision simulation analysis, digital twin technology enables full lifecycle management of equipment status. In the field of wind turbine performance optimization, the application of digital twin technology has the following significant features: First, based on high-fidelity digital twin models, it can integrate the dynamic characteristics of fluid mechanics, thermodynamics, and structural mechanics to comprehensively and accurately model the wind turbine's operating status; second, by combining real-time data with intelligent optimization algorithms, it can dynamically adjust wind turbine operating parameters to achieve multi-objective optimization, such as maximizing efficiency, minimizing energy consumption, and extending equipment life.

[0006] Finally, digital twin technology can predict potential failure points and provide timely warnings based on the fusion of real-time monitoring and historical data, effectively ensuring the safety and reliability of wind turbine operations. This dynamic adaptability significantly improves the equipment's ability to respond to complex operating conditions, addressing the limitations of traditional methods in real-time optimization and fault prediction. Summary of the Invention

[0007] This paper addresses the need for efficient operation of metallurgical blower systems under complex operating conditions and the shortcomings of traditional optimization methods. It proposes a multi-physics field coupling and dynamic performance optimization method for metallurgical blowers based on digital twins. This method provides a new approach for performance optimization and fault prediction of metallurgical blowers. By introducing digital twin technology, it enables high-fidelity simulation of the blower's operation, multi-objective dynamic optimization, and fault prediction and early warning.

[0008] According to one aspect of the present invention, a method for multi-physics field coupling and dynamic performance optimization of a metallurgical fan is provided, comprising the following steps:

[0009] Step 1: Based on the distributed sensor network on the metallurgical blower, multi-physics field data is collected in real time. The multi-physics field data includes fluid dynamics, thermodynamics, and structural mechanics data. The collected data is processed, including data cleaning, denoising, interpolation and feature extraction, to form a high-quality basic data set.

[0010] Step 2: Based on the metallurgical blower’s geometry, material properties, and operating boundary conditions, a multi-physics coupled digital twin model is established. This multi-physics coupled digital twin model combines finite element analysis and fluid simulation technology, and is calibrated using historical and real-time data.

[0011] Step 3: Use the multi-physics field coupled digital twin model to perform multi-physics collaborative simulation of the metallurgical fan, analyze the dynamic coupling between fluid mechanics, thermodynamics, and structural mechanics, and simulate the dynamic response characteristics of the fan under different operating conditions;

[0012] Step 4: Based on the collaborative simulation results in step 3, an adaptive particle swarm optimization algorithm is used to dynamically optimize the operating parameters of the metallurgical fan to obtain multi-objective optimization results. The operating parameters include speed, air volume, temperature, and pressure.

[0013] Step 5: By monitoring the operating status of the metallurgical fan in real time, the optimization results are fed back to the control system, and the operating parameters are dynamically adjusted to ensure efficient and stable operation of the fan, and potential faults are predicted and warned.

[0014] S1: A distributed sensor network is deployed on key components of metallurgical fans (such as blades, rotors, and supporting structures) to collect multi-physics field data in real time, including fluid mechanics, thermodynamics, and structural mechanics data. Specific collection content includes:

[0015] Fluid mechanics data: pressure field, velocity field, flow velocity distribution;

[0016] Thermodynamic data: surface temperature distribution and thermal gradient of key components;

[0017] Structural mechanics data: blade stresses, rotor strains, and support point deformations.

[0018] S2 performs outlier detection and denoising based on the multi-physics field data collected in real time by S1. It uses a combination of local outlier factors (LOF) and expert experience to eliminate null values, duplicate values, zero values, data that does not meet process specifications, and abnormal data. The following mainly introduces the specific steps and formulas for LOF outlier elimination:

[0019] First, define the wind turbine related data set as X={x1,x2,...,x n}, each data point x i Contains the following dimensions: wind speed, pressure, temperature, stress;

[0020] Secondly, for the data point x i , find its k-nearest neighbors (k=5 or k=10 by default), and calculate the point x i With each adjacent point x j The Euclidean distance (Euclidean distance) of:

[0021]

[0022] Define the k-neighbor distance as d(x i ),Right now:

[0023] d(x i ) = Euclidean distance from the kth neighbor

[0024] Again, based on the obtained Euclidean distance results, the local reachability density is calculated, which is defined based on the average reachability distance of k-neighboring points. The formula is as follows:

[0025]

[0026] Where: N k (x i ) is point x i k-domain; max(d k (x j ),d(x i ,x j )) for each neighbor point x j , take the distance between the two points and x j The maximum value among the k-th adjacent distances; |N k (x i )| is the number of adjacent points.

[0027] Finally, the local outlier factor, LOF value, is calculated from the local reachable density (LRD). The local outlier factor is the value of point x. j The ratio of the density of its adjacent points to its own density is given by:

[0028]

[0029] Conventional method for determining LOF value:

[0030] If LOF(x i )≈1, point x i The density of is similar to that of its neighborhood, and it is not an outlier;

[0031] If LOF(x i )>1, point x i The density is significantly lower than that of the neighborhood, which is a potential outlier;

[0032] Remove the items that satisfy LOF(x i )>1, generate the cleaned data set X new .

[0033] S3, based on S2, uses a method combining local outlier factor (LOF) and expert experience to generate a new dataset X new Perform feature screening of the Gradient Boosting Decision Tree (GBDT) model. The specific steps are as follows:

[0034] Cleaned dataset X new Contains the following features:

[0035] Xnew ={x1,x2,...,x n}, each x i Includes: wind speed, pressure, temperature, stress, and defines the target variable Y as the key target performance index of the wind turbine, such as operating efficiency and power consumption.

[0036] First, use the GBDT model to fit the relationship between features and target variables:

[0037] y=f(X new )+∈;

[0038] Where: f is the GBDT model, ∈ is the deviation.

[0039] Secondly, calculate the importance score of the feature. Each decision tree of GBDT measures the contribution of the feature to the target by splitting the node, and the importance score of the feature is obtained by accumulating all trees:

[0040]

[0041] Of which: Gain t (x i ) is the feature x used in the t-th tree i As the gain value when splitting the node.

[0042] Finally, sort the importance scores of all features Importance(x i ), set the threshold (such as Importance(x i )>α) Screen out key features, for example, retain the top 5 important features for subsequent modeling;

[0043] Use the selected features to construct the final dataset X last , as input for subsequent digital twin model construction;

[0044] The data is processed by the LOF and GBDT integrated methods in S2 and S3 to ensure data quality and key features, and complete cleaning and denoising preprocessing;

[0045] Preferably, based on the above solution, the digital twin model construction includes the following steps:

[0046] S4, based on the design drawings of the metallurgical fan, define the geometric structure, material properties and operating boundary conditions, and use 3D modeling software to build the fan's geometric model;

[0047] Based on the structural drawings and material properties of the metallurgical fan (such as blade shape, rotor diameter, housing size, density, thermal conductivity, and elastic modulus), a 3D modeling software was used to construct a geometric model of the fan and define key boundary conditions, such as:

[0048] Fluid conditions: inlet velocity and pressure, outlet pressure;

[0049] Thermal conditions: ambient temperature, internal heat transfer coefficient of the fan;

[0050] Structural conditions: support constraints, external vibrations or loads.

[0051] S5, based on the wind turbine geometry model constructed in S4, uses finite element analysis (FEA) and computational fluid dynamics (CFD) techniques to build a high-fidelity multi-physics coupling sub-model, including the following:

[0052] 1. Fluid Dynamics Module: simulates the flow characteristics of airflow in fan blades and channels;

[0053] Computational fluid dynamics (CFD) is used to simulate the airflow distribution. The governing equation (Navier-Stokes equation) is used to describe the distribution of airflow velocity, pressure, and viscous force in the fan. The specific formula is as follows:

[0054]

[0055] Where: ρ is the fluid density, v is the velocity field, p is the pressure field, and μ is the viscosity coefficient.

[0056] 2. Thermodynamics module: Calculates the effects of heat conduction and convection on the temperature distribution of key components;

[0057] The heat conduction and convection effects in the fan are modeled, and the heat conduction formula is used to describe the temperature changes inside the fan components and complete the definition of the convection heat transfer coefficient h:

[0058]

[0059] q=h·(T 表面 -T 环境 );

[0060] Where: T is the temperature distribution, is the thermal diffusivity.

[0061] 3. Structural Mechanics Module: Analyzes the stress distribution and deformation of wind turbine blades and supporting structures;

[0062] To simulate the stress distribution of wind turbine blades and rotors, finite element analysis (FEA) is used, combined with the structural dynamics equations:

[0063]

[0064] Where: M is the mass matrix, C is the damping matrix, K is the stiffness matrix, u is the displacement vector, and F is the external force load.

[0065] S6, based on the three physical field models constructed in S5 above, combines the fluid field, thermal field, and stress field through a dynamic interaction mechanism to achieve multi-physics field model coupling, for example:

[0066] Fluid-structure coupling: The pressure in the flow field acts on the blade, and the deformation of the blade is calculated; the deformation reacts to the flow field, forming feedback, and the pressure and velocity distribution of the fluid field are obtained;

[0067] Thermal and structural coupling: Temperature gradients cause thermal expansion, affecting the stress distribution of components and revealing the dynamic distribution of the temperature field.

[0068] Multi-field joint solution: ANSYS is used to iteratively solve the coupled equations until the system converges and the stress and deformation results of the blades and rotor are obtained.

[0069] S7: In multi-physics simulations, model complexity and computational effort are generally positively correlated with accuracy. However, for real-time applications (such as dynamic optimization and fault prediction of wind turbines), high-fidelity simulation models are computationally expensive and often fail to meet real-time response requirements. Therefore, model order reduction techniques are needed to reduce computational complexity while preserving key dynamic characteristics, enabling real-time control and achieving rapid response.

[0070] Use finite element analysis to build a high-fidelity model of multi-physics field coupling;

[0071] The dimensionality reduction of high-dimensional data is performed based on the singular value decomposition (SVD) algorithm. The formula is as follows:

[0072] A=UΣV T ;

[0073] Among them, U and V are orthogonal matrices, Σ is the singular value matrix; by selecting significant singular values, the reduced-order model A is constructed k , to meet the needs of real-time response:

[0074]

[0075] Among them, k represents the number of selected singular values, and k is the degree threshold used to control dimensionality reduction;

[0076] The Kalman filter algorithm is used to fuse real-time data with historical data, update the model state, and ensure the dynamic and high-precision model calibration. The specific steps include:

[0077] (1) State prediction, based on the system's dynamic model, using prior states:

[0078]

[0079] in, represents the predicted state vector at the current time k, It represents the estimation of the state at time k-1 based on the observation value at time k-1. A and B are the state transfer matrix and control input matrix respectively. u k-1 is the control input vector at time k-1;

[0080] in, represents the predicted state vector at the current time k, It represents the estimation of the state at time k-1 based on the observation value at time k-1. A and B are the state transfer matrix and control input matrix respectively. u k-1 is the control input vector at time k-1;

[0081] Compute the a priori error covariance matrix of the state:

[0082] P k|k-1 =AP k-1|k-1 A T +Q;

[0083] Among them, P k|k-1 represents the predicted error covariance matrix, P k-1|k-1 represents the error covariance matrix after observation update at time k-1, Q represents the process noise covariance matrix, and A is the state transfer matrix, which describes the change law of the system state from time k-1 to time k. T is the transpose of matrix A, which is used to propagate the state error from time k-1 to time k;

[0084] (2) Kalman gain calculation, based on the predicted error covariance P k|k-1 And the observation noise covariance R, calculate the Kalman gain K k :

[0085] K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ;

[0086] Among them, K k represents the Kalman gain matrix (weighing the degree of trust between the predicted value and the observed value, adjusting the weight of the historical data and the new observation data), H represents the observation matrix (mapping the state space variables to the observation space, describing the relationship from the true state to the observed value), and R represents the observation noise covariance matrix (quantifying the noise size in the observation data);

[0087] (3) Status update, based on real-time observation value z k and predicted values Update current status:

[0088]

[0089] in, represents the updated state vector at the current time k (the optimal state of the current system), z k represents the observation vector (actually measured value) at the current time k, represents the predicted state estimate at time k, represents the predicted value of the observed value;

[0090] Updated status Accurately reflect real-time working conditions;

[0091] (4) Error covariance update: Update the error covariance matrix to ensure the dynamic stability of the system calibration:

[0092] P k|k =(IK k H)P k|k-1 ;

[0093] Among them, P k|k represents the updated error covariance matrix, k represents the discrete time step at the current moment, and I represents the identity matrix;

[0094] (5) Calibration feedback and fusion: feed the fused state value as input to the digital twin model, adjust the multi-physics field simulation parameters, and continuously iterate to ensure the high-precision performance of the model in a dynamic environment.

[0095] Based on the above solution, a digital twin model is used to perform multi-physics collaborative simulation of the wind turbine, specifically including:

[0096] (1) Fluid and structural mechanics interaction simulation: simulate the interaction between fluid dynamics and structural mechanics, and analyze the impact of blade deformation on aerodynamic performance;

[0097] (2) Thermodynamic and structural coupling analysis: to obtain the impact of temperature distribution on the thermal stress and life of key components of the fan.

[0098] During the operation of metallurgical fans, there is a complex coupling between the three physical fields of fluid, thermal and structure, which is manifested as follows:

[0099] 1. High-speed airflow creates dynamic pressure on the blades, causing them to deform, further affecting aerodynamic performance.

[0100] 2. Heat conduction and convection cause uneven temperature distribution of fan components, resulting in concentrated thermal stress and affecting equipment life;

[0101] 3. The vibration and fatigue of structural components such as blades and rotors are affected by the combined effects of fluid loads and temperature changes.

[0102] Traditional single-field simulations struggle to capture the interactions between these multiple physical fields. However, multi-physics collaborative simulations can more accurately describe the operating state of metallurgical fans by simultaneously considering the dynamic relationships between fluid, thermal, and structural forces.

[0103] Based on the constructed three types of sub-models of fluid mechanics, thermodynamics and structural mechanics, the sub-model coupling method is as follows:

[0104] Fluid-structure coupling: Fluid pressure acts on the blades, causing them to deform. The deformed blades react to the flow field, changing the velocity and pressure distribution of the airflow as follows:

[0105] Fluid simulation provides pressure distribution on the blade surface;

[0106] Use the pressure distribution as load input for structural mechanics simulation to calculate the deformation of the blade;

[0107] The blade deformation is fed back into the fluid simulation and the solution is iterated until the results converge.

[0108] Thermal-structural coupling: The temperature distribution of components causes thermal expansion, which affects stress distribution and structural strength as follows:

[0109] Thermal simulation provides the temperature distribution of components;

[0110] Use temperature distribution as thermal expansion load input for structural simulation to calculate stress fields and deformations;

[0111] Evaluate fatigue damage in areas of thermal stress concentration.

[0112] Fluid-thermal-structural three-field coupling: The three-field coupling achieves dynamic interaction through a unified solution framework (such as ANSYS), as follows:

[0113] Flow field pressure affects blade deformation and aerodynamic performance;

[0114] Thermal expansion caused by temperature distribution changes the blade shape and thus the flow field;

[0115] The interaction between fluid and thermal fields affects the overall operating performance;

[0116] The following important performance indicators are obtained through coupled simulation calculations between sub-models:

[0117] Fan flow field pressure distribution, velocity field characteristics and aerodynamic performance indicators (such as efficiency and energy consumption);

[0118] Temperature field distribution, thermal load changes of key components, and local overheating areas;

[0119] Structural stress distribution, deformation displacement, possible fatigue damage and maximum stress location.

[0120] Based on the above scheme, the optimization objectives of the metallurgical fan are determined, and the optimization objectives of the metallurgical fan include maximizing operating efficiency, minimizing energy consumption, and uniforming temperature distribution. The optimization objectives are specifically converted into quantifiable performance indicators, and the quantifiable performance indicators include the fan speed, air volume, temperature, and pressure;

[0121] Based on the collaborative simulation results, a multi-objective optimization function f(x) for wind turbine operation is constructed:

[0122] f(x)=α1E(x)+α2C(x)+α3R(x);

[0123] Among them, E(x) is the operating efficiency, C(x) is the energy consumption, R(x) is the equipment life, α1, α2, α3 are weight coefficients;

[0124] Where E(x) represents the operating efficiency of the fan, C(x) represents the energy consumption of the fan, R(x) represents the uniformity of temperature distribution (extending equipment life), α1, α2, and α3 represent weight coefficients used to measure the importance of different objectives, and x represents the parameter set (all fan operating parameters that need to be optimized);

[0125] Adaptive particle swarm optimization algorithm is used to quickly search for the optimal parameter combination. The formula is:

[0126]

[0127] in, represents the velocity of particle i in the t+1 iteration, ω represents the inertia weight (the proportion of the particle retaining its previous velocity), represents the speed of particle i in t iterations, c1 and c2 represent the individual learning factor and the group learning factor respectively, r1 and r2 represent random numbers between [0, 1], and p best represents the historical optimal position of particle i, represents the current position of particle i in t iterations, g best represents the global optimal position among all particles;

[0128] Update fan operating parameters (speed, air volume, temperature, pressure) to achieve dynamic adjustment of operating performance;

[0129] The optimization results are fed back to the simulation module, and the optimization effect is verified through multiple rounds of iterations to ensure the dynamic adjustment capability to adapt to complex working conditions.

[0130] Fault prediction is based on the fusion analysis of machine learning models and digital twin models, and is divided into the following steps:

[0131] First, we use sensor data to train a fault prediction model and build a time series prediction model for the equipment health status H(t):

[0132] H(t)=f(X t )+∈;

[0133] Among them, H(t) is the health status index (such as remaining life, vibration amplitude, etc.), X t The characteristic vector at the current moment (such as temperature, vibration, stress, etc.), ∈ is the prediction error.

[0134] Secondly, the health status is dynamically updated through Kalman filtering by combining real-time sensor data with the simulation results of the digital twin model:

[0135] H(t|t)=H(t|t-1)+K t (Z t -H(t|t-1));

[0136] Among them, H(t|t) is the health state after calibration, K t is the Kalman gain, Z t For real-time measurement data.

[0137] Finally, based on the changing trend of the health status, predict the potential failure time:

[0138] T failure =min{t|H(t)≤H crit};

[0139] Among them, H crit is the preset fault threshold.

[0140] Identify fault characteristics (such as blade cracks or wear, rotor imbalance, thermal stress fatigue, and airflow instability);

[0141] The model used for fault prediction is the Support Vector Machine (SVM), which captures complex fault patterns through nonlinear kernel functions and establishes an early warning mechanism based on the prediction results.

[0142] Preferably, based on the above solution, it is characterized in that:

[0143] Condition monitoring: Real-time updates of wind turbine operating status, including vibration, temperature field, and stress field, are achieved through the digital twin model.

[0144] Parameter adaptive adjustment: Based on the optimization and monitoring results, the fan operating parameters are adjusted in real time to ensure its stability and efficiency under complex working conditions;

[0145] Fault prediction and early warning: By monitoring the changing trends of key parameters, potential fault points can be predicted and alarm signals can be provided in a timely manner.

[0146] The present invention also provides a metallurgical fan multi-physics field coupling and dynamic performance optimization system, which includes:

[0147] Data acquisition and preprocessing unit: Multi-physics field sensors are deployed on key components of metallurgical fans through a distributed sensor network to collect parameters such as fluid dynamics, thermodynamics, and structural mechanics in real time. Data cleaning, denoising, interpolation, and feature extraction are used to form a high-quality basic data set. Anomaly detection algorithms are combined with expert experience to remove data noise, and key features are selected based on the gradient boosting tree algorithm.

[0148] Digital twin model construction unit: Utilizes the geometric structure, material properties, and operating boundary conditions of the metallurgical fan to establish a high-fidelity finite element model with multi-physics field coupling. Based on historical data and real-time sensor data, the singular value decomposition algorithm is used to reduce the dimensionality of high-dimensional simulation data and construct a finite element reduced-order model for rapid response. The model is dynamically calibrated using Kalman filtering to ensure that the model accurately reflects the operating status of the fan under actual operating conditions.

[0149] Multi-physics collaborative simulation unit: Uses digital twin technology to perform multi-physics collaborative simulation of metallurgical fans, simulating the dynamic coupling between fluid mechanics, thermodynamics, and structural mechanics. Analyzes the dynamic response characteristics of the fan under different operating conditions, revealing the mutual influence between blade deformation, temperature distribution, and structural stress.

[0150] Dynamic performance optimization unit: Based on simulation results, it defines optimization objectives such as maximizing fan operating efficiency, minimizing energy consumption, and extending component life, and constructs a multi-objective optimization function. It uses an adaptive particle swarm algorithm or a genetic algorithm to optimize fan operating parameters and verifies the optimization effect through simulation. It implements real-time monitoring and feedback regulation of the fan operating status, and dynamically adjusts control parameters according to changes in operating conditions to ensure that the equipment is in the best working condition.

[0151] The present invention provides a method for multi-physics field coupling and dynamic performance optimization of metallurgical fans, which covers a systematic closed-loop optimization solution for data acquisition, model construction and calibration, collaborative simulation analysis, optimization and feedback control. In particular, by combining the multi-physics field simulation results with real-time monitoring data, a closed-loop process of dynamic adjustment and optimization is formed, which can adapt to different operating conditions and maintain the optimal performance of the fan. The solution integrates intelligent algorithms (such as adaptive optimization and dynamic feedback control) and high-precision digital twin models to realize intelligent management of the equipment throughout its life cycle from design to operation, providing technical support for the digital transformation of the metallurgical industry.

[0152] This method integrates multi-physics field collaborative simulation technologies such as fluid mechanics, thermodynamics, and structural mechanics, and combines intelligent optimization algorithms with real-time fault prediction technology, aiming to solve the performance optimization and equipment health management problems of metallurgical fans under complex operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0153] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without inventive efforts. In the drawings:

[0154] Figure 1 This is a principle block diagram of a metallurgical fan multi-physics field coupling and dynamic performance optimization system based on digital twin according to an embodiment of the present invention;

[0155] Figure 2 This is a flowchart of data collection and preprocessing operations for a method for multi-physics field coupling and dynamic performance optimization of a metallurgical fan based on digital twins according to an embodiment of the present invention;

[0156] Figure 3 This is a flowchart for constructing a multi-physics high-fidelity coupling model of a metallurgical fan multi-physics coupling and dynamic performance optimization method based on digital twin according to an embodiment of the present invention;

[0157] Figure 4 This is a multi-objective optimization flow chart of a method for multi-physics field coupling and dynamic performance optimization of a metallurgical fan based on digital twin according to an embodiment of the present invention;

[0158] Figure 5 4 is an architectural diagram of a multi-channel data transmission method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0159] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0160] It will be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections.

[0161] To simplify the drawings, only the parts relevant to the present invention are schematically shown in each figure. They do not represent the actual structure of the product. Furthermore, to simplify the drawings and facilitate understanding, in some figures, only one of the components with the same structure or function is schematically depicted or labeled. As used herein, "one" not only means "only one" but also "more than one."

[0162] It should be further understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0163] In the embodiments shown in the accompanying drawings, directional indications (such as up, down, left, right, front, and rear) used to explain the structure and movement of various components of the present invention are not absolute but relative. These descriptions are applicable when the components are in the positions shown in the accompanying drawings. If the descriptions of the positions of these components are changed, the directional indications will also change accordingly.

[0164] In addition, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0165] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings and other embodiments can be obtained based on these drawings without inventive work.

[0166] See also Figure 1 As shown, a method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan of the present invention comprises the following steps:

[0167] S1: A distributed sensor network is deployed on key components of metallurgical fans (such as blades, rotors, and supporting structures) to collect multi-physics field data in real time, including fluid mechanics, thermodynamics, and structural mechanics data. Specific collection content includes:

[0168] Fluid mechanics data: pressure field, velocity field, flow velocity distribution;

[0169] Thermodynamic data: surface temperature distribution and thermal gradient of key components;

[0170] Structural mechanics data: blade stresses, rotor strains, and support point deformations.

[0171] S2 performs outlier detection and denoising based on the multi-physics field data collected in real time by S1. It uses a combination of local outlier factors (LOF) and expert experience to eliminate null values, duplicate values, zero values, data that does not meet process specifications, and abnormal data. The following mainly introduces the specific steps and formulas for LOF outlier elimination:

[0172] First, define the wind turbine related data set as X={x1,x2,...,x n}, each data point x i Contains the following dimensions: wind speed, pressure, temperature, stress, etc.;

[0173] Secondly, for the data point x i , find its k-nearest neighbors (k=5 or k=10 by default), and calculate the point x i With each adjacent point x j The Euclidean distance (Euclidean distance) of:

[0174]

[0175] Define the k-neighbor distance as d(x i ),Right now:

[0176] d(x i ) = Euclidean distance from the kth neighbor

[0177] Again, based on the obtained Euclidean distance results, the local reachability density is calculated, which is defined based on the average reachability distance of k-neighboring points. The formula is as follows:

[0178]

[0179] Where: N k (x i ) is point x i k-domain; max(d k (x j ),d(x i ,x j )) for each neighbor point x j , take the distance between the two points and x j The maximum value among the k-th adjacent distances; |N k (x i )| is the number of adjacent points.

[0180] Finally, the local outlier factor, LOF value, is calculated from the local reachable density (LRD). The local outlier factor is the value of point x. j The ratio of the density of its adjacent points to its own density is given by:

[0181]

[0182] Conventional method for determining LOF value:

[0183] If LOF(x i )≈1, point x i The density of is similar to that of its neighborhood, and it is not an outlier;

[0184] If LOF(x i )>1, point x i The density is significantly lower than that of the neighborhood, which is a potential outlier;

[0185] Remove the items that satisfy LOF(x i )>1, generate the cleaned data set X new .

[0186] S3, based on S2, uses a method combining local outlier factor (LOF) and expert experience to generate a new dataset X new Perform feature screening of the Gradient Boosting Decision Tree (GBDT) model. The specific steps are as follows:

[0187] Cleaned dataset X new Contains the following features:

[0188] X new ={x1,x2,...,x n}, each x i Includes: wind speed, pressure, temperature, stress, etc., and defines the target variable Y as the key target performance index of the wind turbine, such as operating efficiency, power consumption, etc.

[0189] First, use the GBDT model to fit the relationship between features and target variables:

[0190] y=f(X new )+∈;

[0191] Where: f is the GBDT model, ∈ is the deviation.

[0192] Secondly, calculate the importance score of the feature. Each decision tree of GBDT measures the contribution of the feature to the target by splitting the node, and the importance score of the feature is obtained by accumulating all trees:

[0193]

[0194] Of which: Gain t (x i ) is the feature x used in the t-th tree i As the gain value when splitting the node.

[0195] Finally, sort the importance scores of all features Importance(x i ), set the threshold (such as Importance(x i )>α) Screen out key features, for example, retain the top 5 important features for subsequent modeling;

[0196] Use the selected features to construct the final dataset X last , as input for subsequent digital twin model construction;

[0197] The data is processed by the LOF and GBDT integrated methods in S2 and S3 to ensure data quality and key features, and complete preprocessing such as cleaning and denoising;

[0198] S4, based on the structural drawings and material properties of the metallurgical fan (such as blade shape, rotor diameter, shell size, density, thermal conductivity and elastic modulus), use 3D modeling software to build a geometric model of the fan and define key boundary conditions, such as:

[0199] Fluid conditions: inlet velocity and pressure, outlet pressure;

[0200] Thermal conditions: ambient temperature, internal heat transfer coefficient of the fan;

[0201] Structural conditions: support constraints, external vibrations or loads.

[0202] S5, based on the wind turbine geometry model constructed in S4, uses finite element analysis (FEA) and computational fluid dynamics (CFD) techniques to build a high-fidelity multi-physics coupling sub-model, including the following:

[0203] 1. Fluid Dynamics Module: simulates the flow characteristics of airflow in fan blades and channels;

[0204] Computational fluid dynamics (CFD) is used to simulate the airflow distribution. The governing equation (Navier-Stokes equation) is used to describe the distribution of airflow velocity, pressure, and viscous force in the fan. The specific formula is as follows:

[0205]

[0206] Where: ρ is the fluid density, v is the velocity field, p is the pressure field, and μ is the viscosity coefficient.

[0207] 2. Thermodynamics module: Calculates the effects of heat conduction and convection on the temperature distribution of key components;

[0208] The heat conduction and convection effects in the fan are modeled, and the heat conduction formula is used to describe the temperature changes inside the fan components and complete the definition of the convection heat transfer coefficient h:

[0209]

[0210] q=h·(T 表面 -T 环境 );

[0211] Where: T is the temperature distribution, is the thermal diffusivity.

[0212] 3. Structural Mechanics Module: Analyzes the stress distribution and deformation of wind turbine blades and supporting structures;

[0213] To simulate the stress distribution of wind turbine blades and rotors, finite element analysis (FEA) is used, combined with the structural dynamics equations:

[0214]

[0215] Where: M is the mass matrix, C is the damping matrix, K is the stiffness matrix, u is the displacement vector, and F is the external force load.

[0216] S6, based on the three physical field models constructed in S5 above, combines the fluid field, thermal field, and stress field through a dynamic interaction mechanism to achieve multi-physics field model coupling, for example:

[0217] Fluid-structure coupling: The pressure in the flow field acts on the blade, and the deformation of the blade is calculated; the deformation reacts to the flow field, forming feedback, and the pressure and velocity distribution of the fluid field are obtained;

[0218] Thermal and structural coupling: Temperature gradients cause thermal expansion, affecting the stress distribution of components and revealing the dynamic distribution of the temperature field.

[0219] Multi-field joint solution: ANSYS is used to iteratively solve the coupled equations until the system converges and the stress and deformation results of the blades and rotor are obtained.

[0220] S7: In multi-physics simulations, model complexity and computational effort are generally positively correlated with accuracy. However, for real-time applications (such as dynamic optimization and fault prediction of wind turbines), high-fidelity simulation models are computationally expensive and often fail to meet real-time response requirements. Therefore, model order reduction techniques are needed to reduce computational complexity while preserving key dynamic characteristics, enabling real-time control and achieving rapid response.

[0221] This method uses the singular value decomposition (SVD) algorithm to reduce the model dimension, extract the main modal features, and construct a digital twin model that can respond quickly.

[0222] The dimensionality reduction of high-dimensional data is performed based on the singular value decomposition (SVD) algorithm. The formula is as follows:

[0223] A=UΣV T ;

[0224] Among them, U and V are orthogonal matrices, Σ is the singular value matrix;

[0225] Retain the first few significant items in the singular value matrix Σ to generate a reduced-order model;

[0226] Embedding the reduced-order model into a real-time control system can respond to external input changes in a short time;

[0227] At this time, the reduced-order digital twin model has real-time response capabilities;

[0228] The high-fidelity simulation model in the initial stage of modeling is built based on static conditions or historical data and may not reflect the current actual working conditions in real time. Based on this problem, to meet the complexity and dynamic changes of the working conditions, dynamic calibration of the model is required. The specific steps are as follows:

[0229] First, real-time sensor data, such as wind speed, temperature, and stress, is obtained, and the state at the next moment is predicted using the reduced-order digital twin model (dynamic model of the wind turbine):

[0230] x k+1|k =Fx k +Gu k ;

[0231] Among them, x k is a state variable, including parameters such as speed, pressure, and temperature.

[0232] Secondly, a Kalman filter is used to fuse the sensor data and the model prediction value to update the estimated state:

[0233] x k|k =x k|k-1 +K k (z k -Hx k|k-1 );

[0234] Among them, the Kalman gain K k To control the weight of the predicted value and the sensor data.

[0235] Finally, based on the fused state, the parameters of the multi-physics field model (such as fluid viscosity coefficient and material elastic modulus) are adjusted in real time, and the calibrated digital twin model ensures that it can reflect the actual operating status of the wind turbine in real time.

[0236] S8, During the operation of metallurgical fans, there is a complex coupling between the three physical fields of fluid, thermal and structure, which is manifested as follows:

[0237] 1. High-speed airflow creates dynamic pressure on the blades, causing them to deform, further affecting aerodynamic performance.

[0238] 2. Heat conduction and convection cause uneven temperature distribution of fan components, resulting in concentrated thermal stress and affecting equipment life;

[0239] 3. The vibration and fatigue of structural components such as blades and rotors are affected by the combined effects of fluid loads and temperature changes.

[0240] Traditional single-field simulations struggle to capture the interactions between these multiple physical fields. However, multi-physics collaborative simulations can more accurately describe the operating state of metallurgical fans by simultaneously considering the dynamic relationships between fluid, thermal, and structural forces.

[0241] S9, based on the three sub-models of fluid mechanics, thermodynamics and structural mechanics constructed in S5, performs sub-model coupling in the following ways:

[0242] 1. Fluid-Structure Coupling: Fluid pressure acts on the blades, causing them to deform. The deformed blades react to the flow field, changing the velocity and pressure distribution of the airflow as follows:

[0243] Fluid simulation provides pressure distribution on the blade surface;

[0244] Use the pressure distribution as load input for structural mechanics simulation to calculate the deformation of the blade;

[0245] The blade deformation is fed back into the fluid simulation and the solution is iterated until the results converge.

[0246] 2. Thermo-structural coupling: The temperature distribution of components causes thermal expansion, which affects stress distribution and structural strength as follows:

[0247] Thermal simulation provides the temperature distribution of components;

[0248] Use temperature distribution as thermal expansion load input for structural simulation to calculate stress fields and deformations;

[0249] Evaluate fatigue damage in areas of thermal stress concentration.

[0250] 3. Fluid-thermal-structural coupling: The three-field coupling achieves dynamic interaction through a unified solution framework (such as ANSYS), as follows:

[0251] Flow field pressure affects blade deformation and aerodynamic performance;

[0252] Thermal expansion caused by temperature distribution changes the blade shape and thus the flow field;

[0253] The interaction between fluid and thermal fields affects the overall operating performance.

[0254] S10, based on S9, obtains the following important performance indicators through coupled simulation calculations between sub-models:

[0255] Fan flow field pressure distribution, velocity field characteristics and aerodynamic performance indicators (such as efficiency and energy consumption);

[0256] Temperature field distribution, thermal load changes of key components, and local overheating areas;

[0257] Structural stress distribution, deformation displacement, possible fatigue damage and maximum stress location.

[0258] S11, the present invention constructs a multi-objective optimization function with the optimization objectives of maximizing the efficiency, minimizing the energy consumption and extending the life of the fan, which are specifically quantified into the following parameters:

[0259] Optimize the matching between speed and air volume;

[0260] Temperature distribution uniformity;

[0261] Minimization of stress levels.

[0262] Dynamic performance optimization is a multi-objective optimization problem. The optimization objective can be defined as the following function:

[0263] F(x)=[f1(x),f2(x),f3(x)];

[0264] Among them, f1(x) is the maximization of operating efficiency, f2(x) is the minimization of energy consumption, and f3(x) is the extension of component life.

[0265] The operating efficiency f1(x) is usually calculated from the fluid dynamics efficiency:

[0266]

[0267] Among them, W output is the effective power generated by the fan, W input is the input mechanical power.

[0268] The energy consumption target f2(x) represents the input power consumed per unit air volume, and the formula is:

[0269]

[0270] Among them, V is the air volume of the fan.

[0271] The life target f3(x) is based on the material fatigue model and can be expressed as the relationship between stress amplitude and number of cycles:

[0272]

[0273] Where N is the fatigue life (number of cycles), which is usually determined based on the SN curve of the material.

[0274] During the optimization process, multiple physical and operational constraints must be met, including fluid dynamics constraints (the relationship between air volume and pressure), thermodynamic constraints (maximum temperature limit), and structural mechanics constraints (maximum stress limit). The specific forms are as follows:

[0275]

[0276] T max ≤T crit ;

[0277] σ max ≤σ yield ;

[0278] Among them, Q is the air volume, C p is the fluid flow coefficient, ΔP is the pressure difference, T crit is the critical temperature allowed by the material, σ yield Material yield strength.

[0279] The optimization objectives and constraints are unified into a mathematical model as shown below:

[0280]

[0281] Among them, x is the variable to be optimized (such as speed, air volume, temperature, pressure, etc.), g i (x) is the constraint condition, and m is the number of constraints.

[0282] The weighted summation method is used to transform the multi-objective optimization problem into a single-objective optimization problem:

[0283] F total (x)=w1f1(x)+w2f2(x)+w3f3(x);

[0284] Among them, w1, w2, and w3 are the weight coefficients of the target.

[0285] In S12, in order to quickly find the optimal solution, the adaptive particle swarm optimization algorithm (APSO) is used to search for parameters, and dynamic optimization is achieved through the following steps:

[0286] Randomly initialize the parameter combination of the particle swarm (such as speed, air volume and pressure);

[0287] Iteratively update particle positions and search for the optimal solution of the multi-objective optimization function;

[0288] The optimization results are fed back to the fan control system to dynamically adjust the operating parameters.

[0289] S13: Use the digital twin model to simulate and verify the optimized operating parameters, such as the effect of energy consumption reduction, efficiency improvement, and life extension. If the simulation results deviate significantly from the expected results, adjust the optimization algorithm and perform a second iteration.

[0290] S14, real-time monitoring relies on multi-physics field data collected by distributed sensor networks to dynamically obtain the changing trends of key parameters (such as vibration, temperature, pressure, stress, etc.) during the operation of metallurgical fans. The data collection frequency is set according to monitoring requirements, and the following key parameters are dynamically monitored:

[0291] Inlet velocity, outlet pressure, and air flow distribution;

[0292] Vibration frequency and amplitude;

[0293] Component surface and internal temperatures;

[0294] Blade stress and fatigue state of key components.

[0295] S15, fault prediction is based on the fusion analysis of machine learning models and digital twin models, and is divided into the following steps:

[0296] First, we use sensor data to train a fault prediction model and build a time series prediction model for the equipment health status H(t):

[0297] H(t)=f(X t )+∈;

[0298] Among them, H(t) is the health status index (such as remaining life, vibration amplitude, etc.), X t The characteristic vector at the current moment (such as temperature, vibration, stress, etc.), ∈ is the prediction error.

[0299] Secondly, the health status is dynamically updated through Kalman filtering by combining real-time sensor data with the simulation results of the digital twin model:

[0300] H(t|t)=H(t|t-1)+K t (Z t -H(t|t-1));

[0301] Among them, H(t|t) is the health state after calibration, K t is the Kalman gain, Z t For real-time measurement data.

[0302] Finally, based on the changing trend of the health status, predict the potential failure time:

[0303] T failure =min{t|H(t)≤H crit};

[0304] Among them, H crit is the preset fault threshold.

[0305] Identify fault characteristics (such as blade cracks or wear, rotor imbalance, thermal stress fatigue, and airflow instability);

[0306] S16: The model used for fault prediction at S15 is a support vector machine (SVM), which captures complex fault patterns through nonlinear kernel functions and establishes a three-level early warning mechanism based on the prediction results.

[0307] It can be seen from the above technical solution that, compared with the prior art, the present invention provides a method for multi-physical field coupling and dynamic performance optimization of metallurgical fans based on digital twins, which has the following beneficial effects:

[0308] This invention achieves comprehensive simulation and real-time optimization of the operating state of metallurgical fans by constructing a high-fidelity coupled model of multiple physical fields, encompassing fluid mechanics, thermodynamics, and structural mechanics. Unlike traditional single-field or weakly coupled analysis, this method utilizes multi-field coupling technology to reveal the dynamic interaction between fluid pressure, temperature field distribution, and structural stress. Combined with digital twin technology, this method enables dynamic adjustment of fan operating parameters (such as speed, air volume, and pressure) through real-time data input. This technology not only improves simulation accuracy but also provides a scientific basis for the efficiency and stability of fan operation under complex operating conditions.

[0309] This paper adopts a high-precision multi-physics field coupling model as the core of the digital twin system and uses the singular value decomposition (SVD) algorithm to reduce the dimension and select the mode of training data. It establishes a finite element reduced-order model that can respond quickly. It also uses the Kalman filter to fuse historical operation data and real sensor measurement data to achieve data denoising and correction, improve data accuracy and model robustness. The high-precision model is combined with advanced filtering technology to provide strong data support and prediction capabilities for wind turbine performance optimization.

[0310] The present invention achieves real-time monitoring of the multi-physical fields of key wind turbine components (such as blades and rotors) through a distributed sensor network, collecting multi-dimensional data including vibration, pressure, temperature, and stress. A Kalman filter is used to fuse real-time sensor data with the prediction results of the digital twin model, eliminating data noise and sensor errors and dynamically updating the model status. Through this feedback mechanism, not only is the operating status of the equipment monitored in real time, but potential failure points of key components can also be predicted and early warnings can be issued, significantly improving the safety and reliability of wind turbine operation.

[0311] The present invention provides a systematic closed-loop optimization solution covering data acquisition, model building and calibration, collaborative simulation analysis, optimization and feedback control. In particular, by combining multi-physics field simulation results with real-time monitoring data, a closed-loop process of dynamic adjustment and optimization is formed, which can adapt to different operating conditions and maintain the optimal performance of the fan. This solution integrates intelligent algorithms (such as adaptive optimization and dynamic feedback control) and high-precision digital twin models to achieve intelligent management of the equipment throughout its life cycle from design to operation, providing technical support for the digital transformation of the metallurgical industry.

[0312] Finally, the method of this application is only a preferred embodiment and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for multi-physical field coupling and dynamic performance optimization of metallurgical fans, characterized in that: The following steps are involved: Step 1: Based on the distributed sensor network on the metallurgical fan, multi-physics field data is collected in real time, and the multi-physics field data includes fluid mechanics, thermodynamics and structural mechanics data; the collected data is processed, including data cleaning, denoising, interpolation and feature extraction, to form a high-quality basic data set; Step 2: Based on the geometric structure, material properties and operating boundary conditions of the metallurgical fan, a multi-physics field coupled digital twin model is established. The multi-physics field coupled digital twin model combines finite element analysis and fluid simulation technology, and is calibrated using historical data and real-time data; Step 3: Use the multi-physics field coupling digital twin model to perform multi-physics field collaborative simulation of the metallurgical fan, analyze the dynamic coupling between fluid mechanics, thermodynamics and structural mechanics, and simulate the dynamic response characteristics of the fan under different operating conditions; Step 4: Based on the collaborative simulation results in step 3, an adaptive particle swarm optimization algorithm is used to dynamically optimize the operating parameters of the metallurgical fan to obtain a multi-objective optimization result, wherein the operating parameters include speed, air volume, temperature and pressure; Step 5: By real-time monitoring of the operating status of the metallurgical fan, the optimization results are fed back to the control system, and the operating parameters are dynamically adjusted to ensure efficient and stable operation of the fan, and to predict and warn of potential faults.

2. A method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan according to claim 1, characterized in that: The digital twin model construction includes the following steps: Based on the design drawings of the metallurgical fan, define the geometric structure, material properties and operating boundary conditions, and use 3D modeling software to build the geometric model of the fan; Use finite element analysis to build a high-fidelity model of multi-physics field coupling; The dimensionality reduction of high-dimensional data is based on the singular value decomposition (SVD) algorithm. The formula is as follows: A=UΣV T ; Among them, U and V are orthogonal matrices, Σ is a singular value matrix; by selecting significant singular values, the reduced-order model A is constructed k , to meet the needs of real-time response: Among them, k represents the number of selected singular values, which is used to control the degree threshold of dimensionality reduction; The Kalman filter algorithm is used to fuse real-time data with historical data, update the model status, and ensure the dynamic and high-precision of model calibration. The specific steps include: (1) State prediction: Based on the dynamic model of the system, using the prior state: in, represents the predicted state vector at the current time k, It represents the estimation of the state at time k-1 based on the observed value at time k-1. A and B are the state transfer matrix and control input matrix respectively. u k-1 is the control input vector at time k-1; Predict the current state Compute the a priori error covariance matrix of the state: P k|k-1 =AP k-1|k-1 From T +Q; Among them, P k|k-1 represents the predicted error covariance matrix, P k-1|k-1 represents the error covariance matrix after observation update at time k-1, Q represents the process noise covariance matrix, A is the state transfer matrix, which describes the change law of the system state from time k-1 to time k, A T is the transpose of matrix A, which is used to propagate the state error from time k-1 to time k; (2) Kalman gain calculation, based on the predicted error covariance P k|k-1 and the observation noise covariance R, calculate the Kalman gain K k : K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ; Among them, K k represents the Kalman gain matrix (weighing the degree of trust between the predicted value and the observed value, adjusting the weight of the historical data and the new observation data), H represents the observation matrix (mapping the state space variables to the observation space, describing the relationship from the true state to the observed value), and R represents the observation noise covariance matrix (quantifying the noise size in the observation data); (3) Status update, based on real-time observation value z k and predicted values Update current status: in, represents the updated state vector at the current time k (the optimal state of the current system), z k represents the observation vector (actually measured value) at the current time k, represents the predicted state estimate at time k, Represents the predicted value of the observed value, the updated state Accurately reflect real-time working conditions; (4) Error covariance update: Update the error covariance matrix to ensure the dynamic stability of system calibration: P k|k =(I-K k H)P k|k-1 ; Among them, P k|k represents the updated error covariance matrix, k represents the discrete time step at the current moment, and I represents the identity matrix; (5) Calibration feedback and fusion: Feedback the fused state value as input to the digital twin model, adjust the multi-physics field simulation parameters, and continuously iterate to ensure the high-precision performance of the model in a dynamic environment.

3. A method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan according to claim 1, characterized in that: The digital twin model is used to perform multi-physics collaborative simulation of the wind turbine, including: (1) Interactive simulation of fluid and structural mechanics: simulate the interaction between fluid dynamics and structural mechanics, and analyze the impact of blade deformation on aerodynamic performance; (2) Thermodynamic and structural coupling analysis: to obtain the impact of temperature distribution on the thermal stress and life of key components of the fan.

4. A method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan according to claim 1, characterized in that: Determine the optimization target of the metallurgical fan, the optimization target of the metallurgical fan includes maximizing operating efficiency, minimizing energy consumption and uniformizing temperature distribution, and specify the optimization target as quantifiable performance indicators, the quantifiable performance indicators include the speed, air volume, temperature and pressure of the fan; The optimization objectives of the metallurgical fan include improving operating efficiency, reducing energy consumption per unit air volume, and equalizing heat loads of key components; the optimization objectives are specifically quantifiable performance indicators, and the quantifiable performance indicators include the speed, air volume, temperature gradient, and pressure distribution of the fan; Based on the collaborative simulation results, a multi-objective optimization function f(x) for wind turbine operation is constructed: f(x)=α1E(x)+α2C(x)+α3R(x); Among them, E(x) is the operating efficiency, C(x) is the energy consumption, R(x) is the equipment life, α1, α2, α3 are weight coefficients; Among them, E(x) represents the operating efficiency of the fan, C(x) represents the energy consumption of the fan, R(x) represents the uniformity of temperature distribution (extending the life of the equipment), α1, α2, and α3 represent weight coefficients used to measure the importance of different objectives, and x represents the parameter set (all fan operating parameters that need to be optimized); Adaptive particle swarm optimization algorithm is used to quickly search for the optimal parameter combination. The formula is: in, represents the velocity of particle i in the t+1 iteration, ω represents the inertia weight (the proportion of the particle retaining its velocity at the previous moment), represents the speed of particle i in t iterations, c1 and c2 represent the individual learning factor and the group learning factor respectively, r1 and r2 represent random numbers between [0, 1], and p best represents the historical optimal position of particle i, represents the current position of particle i in iteration t, g best represents the global optimal position among all particles; Update fan operating parameters (speed, air volume, temperature, pressure) to achieve dynamic adjustment of operating performance; The optimization results are fed back to the simulation module, and the optimization effect is verified through multiple rounds of iterations to ensure the dynamic adjustment capability to adapt to complex working conditions.

5. A method for multi-physical field coupling and dynamic performance optimization of a metallurgical fan as claimed in claim 4, characterized in that: Condition monitoring: The digital twin model is used to update the wind turbine operating status in real time, including vibration, temperature field and stress field; Parameter adaptive adjustment: According to the optimization and monitoring results, the fan operating parameters are adjusted in real time to ensure its stability and efficiency under complex working conditions; Fault prediction and early warning: By monitoring the changing trends of key parameters, potential fault points can be predicted and alarm signals can be provided in a timely manner.

6. A metallurgical fan multi-physics field coupling and dynamic performance optimization system, characterized in that: The system includes: Data acquisition and preprocessing unit: Multi-physics field sensors are arranged on key components of metallurgical fans through distributed sensor networks to collect fluid mechanics, thermodynamics and structural mechanics parameters in real time; data cleaning, denoising, interpolation and feature extraction are used to form a high-quality basic data set; anomaly detection algorithms are combined with expert experience to remove data noise, and key features are screened based on the gradient boosting tree algorithm; Digital twin model construction unit: Use the geometric structure, material properties and operating boundary conditions of the metallurgical fan to establish a high-fidelity finite element model of multi-physical field coupling; Based on historical data and real-time sensor data, combine the singular value decomposition algorithm to reduce the dimension of high-dimensional simulation data, build a finite element reduced-order model for rapid response; Dynamically calibrate the model through Kalman filtering to ensure that the model can accurately reflect the operating status of the fan under actual working conditions; Multi-physics collaborative simulation unit: Use digital twin technology to perform multi-physics collaborative simulation on metallurgical fans, simulate the dynamic coupling between fluid mechanics, thermodynamics and structural mechanics; analyze the dynamic response characteristics of fans under different operating conditions, and reveal the mutual influence between blade deformation, temperature distribution and structural stress; Dynamic performance optimization unit: Based on the simulation results, define the optimization goals of maximizing fan operating efficiency, minimizing energy consumption and extending component life, and construct a multi-objective optimization function; use adaptive particle swarm algorithm or genetic algorithm to optimize fan operating parameters, and verify the optimization effect through simulation; realize real-time monitoring and feedback adjustment of the fan operating status, dynamically adjust control parameters according to changes in operating conditions, and ensure that the equipment is in the best working state.

Images