Grate cooler working condition prediction method combining intrinsic orthogonal decomposition and deep learning

Abstract

The invention discloses a grate cooler working condition prediction method combining intrinsic orthogonal decomposition and deep learning, and belongs to the technical field of digital twinning. The prediction method comprises the following steps: simulating a grate cooler flow field through a CFD technology to obtain a large amount of sample data, and combining an intrinsic orthogonal decomposition method with a multilayer back propagation neural network to predict the working condition of a grate cooler; therefore, rapid prediction of the flow field of the grate cooler is achieved, the problems of high cost and long time of CFD simulation calculation are solved, meanwhile, prediction data and actual sensor detection values are in butt joint through a coupling filtering correction algorithm, the problem that errors exist between simulation prediction values and field detection values is solved, and the prediction accuracy of the flow field of the grate cooler is improved. And actual deployment and guidance optimization production of the reduced order and digital twin model of the grate cooler in a factory are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital twins, and specifically, relates to a method for predicting a grate cooler operating condition by combining intrinsic orthogonal decomposition and deep learning. Background Art

[0002] Grate cooler is a large cooling and heat exchange equipment in the cement industry. In the cement production process, grate cooler is a key equipment for high-temperature clinker cooling and heat recovery. Its stable and efficient operation is of great significance for cement enterprises to improve clinker quality, reduce consumption and save energy. The heat exchange between the clinker and the cooling air inside it directly affects the quality of cement and the energy consumption of the clinker burning system. In order to improve the step-by-step utilization efficiency of heat in the grate cooler, it is necessary to fully understand the temperature field and flow field distribution inside the grate cooler. The grate cooler is essentially a large heat exchanger. The high-temperature clinker and the low-temperature cooling air exchange heat inside. Its main function is to cool the high-temperature clinker discharged from the rotary kiln, recover the heat at the same time, and supply it to the rotary kiln and decomposition furnace in the form of secondary air and tertiary air respectively, so as to make full use of the heat in the grate cooler and improve the thermal efficiency and clinker quality of the entire burning system.

[0003] In the actual application of cement production, grate coolers face problems such as equipment wear caused by long-term operation, variability of working conditions, and improper maintenance. The current technical level is limited and it is impossible to deploy monitoring equipment such as thermocouples at each key position of the grate cooler to accurately measure the temperature of the clinker in real time. Many domestic cement companies still rely mainly on the experience and intuition of staff when adjusting the key operating parameters of the grate cooler. This practice lacks scientific theoretical support and has obvious subjective and lag problems. This unstable control level can easily lead to "undercooling" or "overcooling" of the clinker during the cooling process, affecting the quality of the clinker. The internal structure and heat exchange of the grate cooler are relatively complex. The temperature of the clinker is high, and the temperature distribution and flow conditions inside the grate cooler are difficult to obtain directly.

[0004] To improve the heat utilization efficiency of the grate cooler, it is necessary to fully understand the temperature field and flow field distribution inside the grate cooler. In order to improve the calculation speed of numerical simulation and realize the real-time prediction of the full flow field data, the grate cooler simulation model must be reduced in order to establish a digital twin model with sufficient accuracy and computational efficiency. How to use the features obtained by POD (Properorthogonal decomposition) to accurately predict and represent the flow field under specified working conditions is a key technical issue in realizing grate cooler model reduction and digital twin. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting the operating conditions of a grate cooler by combining intrinsic orthogonal decomposition and deep learning, so as to solve the problem in the prior art that the internal structure and heat exchange conditions of the grate cooler are relatively complex, the clinker temperature is high, and the temperature distribution and flow conditions inside the grate cooler are difficult to directly obtain.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A method for predicting the working condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning comprises the following steps:

[0008] S1 Establish a numerical simulation model. Based on CFD simulation software such as Fluent, the gas-solid two-phase flow behavior of the grate cooler is simulated based on the porous medium model. The boundary conditions, material parameters, particle parameters, etc. are set according to the actual working parameters of the grate cooler, and the calculation is submitted to realize the transient simulation of the grate cooler under different wind speed conditions. The model can achieve: (1) heat exchange simulation between high-temperature clinker particles and cooling air; (2) simulation analysis of the internal flow field parameters of the grate cooler.

[0009] S2 Verification and adjustment of simulation results. Verify the simulation results based on the actual production data of the plant. By comparing the simulation results with the actual production data, adjust the simulation model, verify the accuracy of the simulation parameters and model construction, and improve the accuracy and reliability of the model.

[0010] S3 outputs simulation sample data. After the simulation model is verified, the grate cooler simulation model with different input parameters is submitted in batches to explore the influence of different parameters on the flow field and temperature field of the grate cooler system to obtain different sample spaces.

[0011] S4 intrinsic orthogonal decomposition of sample data. Through POD decomposition, the fundamental modal coefficient matrix and a small number of POD fundamental modes can be used to characterize most of the information of the original sample, so that the high-dimensional flow field data can be reduced and mapped to a low-dimensional orthogonal fundamental mode space, thereby analyzing the main characteristics of the flow field temperature field.

[0012] S5 model deployment and flow field prediction. The basic mode eigenvalue coefficients are quickly predicted through the BPNN model, and then reverse reconstruction is performed based on POD to obtain the predicted value of the flow field temperature field, achieving millisecond-level input and output.

[0013] S6 sensor deployment and prediction data correction. In the actual production process of the factory, due to the interweaving of various complex factors, there is inevitably a deviation between the predicted data and the actual situation on site. The simulation results and the actual measurement data are fused using a filtering algorithm, and the data of the local observation point is combined with the full-field prediction data to update the full-field flow field prediction data to make it closer to the actual situation of the actual physical field. After the actual measured data is corrected, the predicted flow field and temperature field data of the grate cooler are obtained, and the internal flow field of the grate cooler is mastered. These data are used as an important basis for optimizing decision support and adjusting operating conditions to ensure that the grate cooler operates under optimal conditions.

[0014] Beneficial effects of the present invention:

[0015] 1. The present invention simulates the grate cooler flow field through CFD technology to obtain a large amount of sample data, and then combines the intrinsic orthogonal decomposition method with the multi-layer back propagation neural network to achieve rapid prediction of the grate cooler flow field, solving the problems of high cost and long time of CFD simulation calculation. At the same time, the coupling filter correction algorithm connects the predicted data with the actual sensor detection value, solves the problem of error between the simulation prediction value and the on-site detection value, and realizes the actual deployment of grate cooler reduction and digital twin model in the factory and guides optimized production.

[0016] 2. The present invention combines the POD data reduction method and the BPNN neural network model to build a digital twin model of the grate cooler, which effectively removes the redundant features of the data, greatly improves the efficiency of the flow field and temperature field calculation, and can achieve millisecond-level operating parameter input and flow field data output, so that the model has sufficient calculation accuracy and generalization ability.

[0017] 3. The present invention integrates and corrects the real-time monitoring data of sensor deployment and the predicted data of the BPNN neural network, and can obtain the flow field and temperature field data of the grate cooler corrected based on the measured data, so that the predicted data is closer to the real physical field, which serves as an important basis for optimization decision support and operation condition adjustment, and provides optimization suggestions and early warning information. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below in conjunction with the accompanying drawings.

[0019] Figure 1 It is a flow chart of a method for predicting the working condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to the present invention;

[0020] Figure 2 Schematic diagram of neural network structure;

[0021] Figure 3 Schematic diagram of the model training process. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] The core of the present invention is to use a neural network as a proxy model to build an approximate model from operating variables to flow field data, and realize the establishment of a data-driven grate cooler reduced-order model and digital twin method. Through this method, it is only necessary to use CFD software to simulate the grate cooler flow field temperature field under dozens of different operating conditions, and then perform intrinsic orthogonal decomposition on the sample data to obtain all the basic modes and corresponding eigenvalues ​​of the grate cooler in the entire operating space, and establish a multi-layer neural network model from operating parameters to basic mode coefficients through a back-propagation neural network to achieve rapid prediction of the flow field. The present invention simulates the grate cooler flow field through CFD technology, obtains a large amount of sample data, and then combines the intrinsic orthogonal decomposition method with the multi-layer back-propagation neural network to achieve rapid prediction of the grate cooler flow field.

[0024] Specifically, Figure 1 As shown, the method for predicting the working condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning described in the present invention comprises the following steps:

[0025] S1. CFD simulation of grate cooler. The numerical simulation module establishes a transient simulation model of the grate cooler based on the porous media model to simulate the gas-solid two-phase flow behavior of the grate cooler. The simulation model takes into account the structural characteristics, material characteristics and other factors of the grate cooler. Through averaging and statistical processing, the stacked cement clinker layer can be approximated as a porous medium with large porosity. The flow and heat transfer of cooling air between high-temperature cement clinker is similar to the flow heat transfer of gas flowing through porous media, realizing the non-thermodynamic equilibrium heat exchange between cooling air and clinker, better predicting the gas-solid flow in larger-scale industrial equipment, reflecting its internal flow field law, and comparing with engineering data to verify the accuracy.

[0026] Based on the representative element volume (REV) hypothesis and combined with the volume averaging method, the expression of the macroscopic continuity equation of porous media is as follows:

[0027]

[0028] The porous media model in Fluent adds a flow direction resistance based on empirical assumptions to the defined porous media region, which is equivalent to adding a momentum source term to the standard momentum equation, which can be expressed as

[0029]

[0030] In the formula, S i represents the source term of the momentum equation; i represents the three directions of x, y, and z; the first and second terms on the right side of the formula represent the viscous drag term and the inertial drag term respectively; D and C are the matrices of the viscous drag coefficient and the inertial drag coefficient respectively; μ and ρ represent the viscosity and density of the fluid respectively; is the velocity vector and velocity magnitude of the fluid in the j direction.

[0031] The fluid energy equation of the porous media internal heat transfer model is as follows:

[0032]

[0033] The solid phase energy equation is as follows:

[0034]

[0035] Where, T f represents the average temperature of the fluid, T s represents the average temperature of the solid, h v represents the unit volume convection heat transfer coefficient between the porous medium solid phase and the fluid, ε represents the porosity, and λ s , f Denote the thermal conductivity of solid and fluid, q f ,q s They represent the internal heat source of fluid and the internal heat source of solid respectively.

[0036] S2. Set up multi-parameter input to obtain sample space. Using the Design point function in ANSYS, batch submit grate cooler simulation models with different input parameters to explore the effects of different parameters such as cooling air velocity, cooling air temperature, different particle sizes of clinker particles, and porosity of clinker particles on the flow field of the grate cooler system to obtain different sample spaces.

[0037] S3. Intrinsic orthogonal decomposition of sample data. Through the POD orthogonal decomposition method, based on the sample variance maximization theory, all the basic modal flow fields and corresponding eigenvalues ​​of the grate cooler in the entire operating space are obtained. Intrinsic orthogonal decomposition or principal component analysis is a method of vector data statistical analysis that can reduce the order of high-dimensional flow field data and map it to a low-dimensional orthogonal basic modal space, thereby extracting the main features of the flow field and its corresponding basic modal coefficients. In essence, it is to maximize the sample variance in each dimension after reducing the order of the flow field to a low dimension. First, the samples (flow field data) calculated by simulation need to be standardized. Let the original data be x i , where i = 1, 2, 3, ..., r, sample data x iis an n-dimensional vector (n is related to the number of grids divided during CFD calculations), and r is the number of samples.

[0038]

[0039] Formula (1) is to obtain the mean of the sample, and formula (2) is to subtract the mean from the original data. Data standardization is conducive to eliminating the differences between sample data features and avoiding the influence of individual discrete values ​​on the analysis results. The covariance matrix of the standardized data can be obtained:

[0040]

[0041] By solving the eigenvalues ​​of the nXn-order covariance matrix, the first m-order eigenvalues ​​can be recorded as λ 1 , 2 , …, λ m , the corresponding fundamental mode eigenvector can be recorded as ξ 1 , 2 ,…,ξ m The selection of m value is determined by the proportion of variance values ​​of different fundamental modes to the total variance value, ensuring that the characteristic components contained in the fundamental mode account for more than 95% of the entire sample space. Then the original sample data can be approximately expressed as:

[0042] X=U λ ξ(8)

[0043] In this way, the fundamental mode coefficient matrix and a small number of POD fundamental modes can be used to represent most of the information of the original sample.

[0044] S4. Train the BPNN neural network. Build a neural network (BPNN) and train a multi-input-output mapping model between the operating parameter space and the eigenvalue coefficients of the flow field base mode to achieve a better fitting effect. The reduced-order eigenvalue matrix corresponding to the grate cooler flow field under different input parameters cannot be estimated manually. The back propagation neural network with multiple input parameters and multiple output results has the characteristics of strong nonlinear mapping ability, self-learning weights and good generalization ability. It is suitable as a proxy model to train the mapping relationship between input parameters and the POD base coefficient matrix. The sample data obtained from the CFD simulation of the grate cooler is relatively small (generally dozens or hundreds of groups, which are small sample data). The shallow neural network has sufficient nonlinear mapping and generalization capabilities. The BPNN model with two hidden layers can be selected to take into account stability and feature description capabilities. The neural network structure is as follows: Figure 2 shown.

[0045] The neural network establishes a mapping from r-dimensional input to m-dimensional output, where the input is, k corresponds to the number of input parameters of the training sample, the number of hidden layer neurons is p and q respectively, and the output is y = [y1, y2, ..., ym ] T , m corresponds to the number of POD basic modes after dimensionality reduction. ω1, ω2, ω3 are weight matrices between layers, and their sizes are ,×k, q×p and m×q respectively. The BPNN model updates the weight matrix by back-propagating the error, and increases the nonlinear mapping ability and stability of the model through different activation functions, thereby completing the model training. Figure 3 As shown, the specific training steps are as follows:

[0046] 1) Divide the samples into training set and validation set for training and result verification;

[0047] 2) Use orthogonal initialization of weight matrix. Weight initialization affects the training effect and convergence speed of neural network. By initializing the weight matrix to an orthogonal matrix, the information transmission is maintained and the ladder and vanishing problems in training are reduced. The weight matrix W is orthogonal, that is, W T W = I;

[0048] 3) Calculate the input and output of the first hidden layer, and use the "Sigmoid" activation function to increase the nonlinear expression ability of the model, smooth the gradient, and avoid jumping output values;

[0049] 4) Calculate the input and output of the second hidden layer, and use the "Leaky Relu" activation function to accelerate training convergence, which solves the problem of gradient disappearance during back propagation, that is, if a negative number is input, the gradient is completely zero, and also avoids the problem of neuron saturation;

[0050] 5) Calculate the reverse error of each layer and update the weight matrix;

[0051] 6) Calculate the prediction error of the validation set to evaluate the appropriate number of training rounds until the model has sufficient computational accuracy and generalization ability.

[0052] S5. Model deployment and flow field prediction. The reduced-order model generated by this method is relatively small in size, fast in calculation, low in memory usage, and does not rely on foreign CFD software environments. After configuring a small amount of open source Python libraries, the digital twin model can be deployed on factory industrial control equipment. For a new set of actual factory parameter combinations Z j , the fundamental mode eigenvalue coefficient can be quickly predicted through the BPNN model Then, based on POD, reverse reconstruction is performed to obtain the predicted flow field value. Achieve millisecond-level input and output;

[0053] S6. Sensor deployment and data correction. Sensor detection is performed at local points of the grate cooler to obtain local observation data. The prediction data is updated across the field using distance-based correction and Gaussian function-based filtering correction algorithms. Assume that n sensors are deployed and the measured value of each sensor is T i (i=1,2,3…n), the predicted flow field value at the same position is Then for any data point x in the predicted flow field k , and its corrected value is:

[0054]

[0055] Among them, d i For data point x k The distance from each sensor position; σ is the correction coefficient, which is related to the specific structural parameters of the grate cooler and is obtained by fitting according to the specific working conditions of different grate coolers. This algorithm systematically considers the correction of the predicted data by all sensor measurement points, and can effectively improve the accuracy of the flow field prediction value on the basis of retaining the real characteristics of the flow field, making the predicted data closer to the data of the real physical field. After the measured data is corrected, the predicted flow field and temperature field data of the grate cooler are obtained, and the internal flow field of the grate cooler is mastered. These data are used as an important basis for optimizing decision support and adjusting operating conditions to ensure that the grate cooler operates under the optimal working conditions.

[0056] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

Claims

1. A method for predicting the working condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning, characterized in that: The steps include: S1. Establish a numerical simulation model: simulate the gas-solid two-phase flow behavior of the grate cooler, submit the calculation according to the actual working parameters of the grate cooler, and realize the transient simulation of the grate cooler under different wind speed conditions; S2. Verification and adjustment of simulation results: By comparing the simulation results with the actual production data, the simulation model is adjusted to verify the accuracy of the simulation parameters and model building; S3. Output simulation sample data: Batch submit grate cooler simulation models with different input parameters to explore the influence of different parameters on the flow field and temperature field of the grate cooler system to obtain different sample spaces; S4. Intrinsic orthogonal decomposition of sample data: Through the POD orthogonal decomposition method, based on the sample variance maximization theory, all the basic modal flow fields and corresponding eigenvalues ​​of the grate cooler in the entire working space are obtained; S5. Build a neural network and train a multi-input-output mapping model between the working condition parameter space and the basic modal eigenvalue coefficients of the flow field to achieve a better fitting effect; S6, model deployment and flow field prediction; S7, sensor deployment and data correction: by performing sensor detection at local points of the grate cooler, local observation data is obtained, and the prediction data is updated throughout the field using distance-based correction and Gaussian function-based filtering correction algorithms.

2. The method for predicting the operating condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to claim 1 is characterized in that: In step S1, a numerical simulation model is established based on the porous medium model.

3. The method for predicting the operating condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to claim 2 is characterized in that: Based on the representation volume element hypothesis and combined with the volume averaging method, the expression of the macroscopic continuity equation of the porous medium is: The porous media model in Fluent adds a flow direction resistance based on empirical assumptions to the defined porous media region, which is equivalent to adding a momentum source term to the standard momentum equation, which can be expressed as In the formula, S i represents the source term of the momentum equation; i represents the three directions of x, y, and z; the first and second terms on the right side of the formula represent the viscous drag term and the inertial drag term respectively; D and C are the matrices of the viscous drag coefficient and the inertial drag coefficient respectively; μ and ρ represent the viscosity and density of the fluid respectively; is the velocity vector and velocity magnitude of the fluid in the j direction; The fluid energy equation of the porous media internal heat transfer model is as follows: The solid phase energy equation is as follows: Where, T f represents the average temperature of the fluid, T s represents the average temperature of the solid, h v represents the unit volume convection heat transfer coefficient between the porous medium solid phase and the fluid, ε represents the porosity, and λ s , f Denote the thermal conductivity of solid and fluid, q f ,q s They represent the internal heat source of fluid and the internal heat source of solid respectively.

4. The method for predicting the operating condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to claim 1 is characterized in that: The neural network uses a BPNN model with two hidden layers. The neural network establishes a mapping from r-dimensional input to m-dimensional output, where the input is x = [x1, x2, ..., x k ] T , k corresponds to the number of input parameters of the training sample, the number of neurons in the hidden layer are p and q respectively, and the output is y = [y1, y2, ..., y m ] T , m corresponds to the number of POD basis modes after dimensionality reduction; ω1, ω2, ω3 are the weight matrices between layers, and their sizes are p×k, q×p, and m×q, respectively.

5. The method for predicting the operating condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to claim 4 is characterized in that: The specific training steps of the neural network model are: Divide the samples into training set and validation set; The information transfer is maintained by initializing the weight matrix as an orthogonal matrix; Calculate the input and output of the first hidden layer, and use the "Sigmoid" activation function to increase the nonlinear expression ability of the model; Calculate the input and output of the second hidden layer, and use the "Leaky Relu" activation function to accelerate training convergence; Calculate the reverse error of each layer and update the weight matrix; Calculate the prediction error of the validation set to evaluate the appropriate number of training rounds until the model has sufficient computational accuracy and generalization ability.

6. The method for predicting the operating condition of a grate cooler combining intrinsic orthogonal decomposition and deep learning according to claim 1 is characterized in that: In step S7, it is assumed that n sensors are arranged and the measured value of each sensor is T i (i=1,2,3…n), the predicted flow field value at the same position is Then for any data point x in the predicted flow field k , and its corrected value is: Among them, d i For data point x k The distance from each sensor position; σ is the correction factor.

Images