A method and device for predicting critical flow velocity of shell-and-tube heat exchanger flow bullet instability

Abstract

The application provides a shell-and-tube heat exchanger flow bullet instability critical flow velocity prediction method and device, including the following steps: obtaining historical structure parameters and process parameters of the shell-and-tube heat exchanger; analyzing the historical structure parameters and process parameters to obtain characteristic parameters affecting the shell-and-tube heat exchanger flow bullet instability critical flow velocity; obtaining and preprocessing the historical characteristic parameters and corresponding critical flow velocity measurement values to obtain a historical vector data set; constructing an initial critical flow velocity prediction model according to the historical vector data set, initializing hyperparameters of the initial critical flow velocity prediction model, training the initialized initial critical flow velocity prediction model to obtain an optimal critical flow velocity prediction model; inputting parameters of the shell-and-tube heat exchanger under a target scene into the optimal critical flow velocity prediction model for prediction to obtain the shell-and-tube heat exchanger flow bullet instability critical flow velocity. The application realizes accurate prediction of the shell-and-tube heat exchanger flow bullet instability critical flow velocity.

Description

Technical Field

[0001] This invention relates to the field of flow-induced vibration in shell-and-tube heat exchangers, and in particular to a method and apparatus for predicting the critical flow velocity for helical instability in shell-and-tube heat exchangers. Background Technology

[0002] Shell-and-tube heat exchangers are widely used in industries such as chemical and power generation due to their excellent adaptability and low manufacturing cost. However, the heat exchange tubes are often subjected to cross-flow scouring by the shell-side fluid, which may cause flow-induced vibration problems in the tube bundle. Lithotropic instability, as the most dangerous flow-induced vibration mechanism, must be avoided in the design of shell-and-tube heat exchangers. For calculating the critical velocity of fluid elastic instability, the most widely used model in engineering is the CONNORS model, which can be used to quickly estimate the critical velocity of shell-and-tube heat exchangers, but the evaluation results are overly conservative.

[0003] The critical velocity of a shell-and-tube heat exchanger is closely related to parameters such as the tube arrangement, tube diameter ratio, and liquid density. Machine learning can extract patterns and relationships between various factors from a large number of relevant parameters of shell-and-tube heat exchangers, and predict the critical velocity, reducing the conservatism of the prediction and saving design costs. Therefore, machine learning models can be used to predict the critical velocity, improving the accuracy of predicting the heliotropic instability critical velocity of shell-and-tube heat exchangers, thereby saving design costs. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for predicting the critical velocity of helical instability in shell-and-tube heat exchangers, thereby achieving accurate prediction of the critical velocity of helical instability in shell-and-tube heat exchangers, solving the problem of overly conservative results in traditional methods, and saving the design cost of heat exchangers.

[0005] To achieve the above objectives, the present invention provides a method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger, comprising the following steps:

[0006] Obtain historical structural and process parameters of shell-and-tube heat exchangers;

[0007] Analysis of historical structural and process parameters yields characteristic parameters that influence the critical flow velocity for helical instability in shell-and-tube heat exchangers.

[0008] Historical feature parameters and corresponding critical flow velocity measurements are acquired and preprocessed to obtain a historical vector dataset.

[0009] An initial critical velocity prediction model is constructed based on a historical vector dataset. The hyperparameters of the initial critical velocity prediction model are initialized. The initialized initial critical velocity prediction model is then trained to obtain the optimal critical velocity prediction model.

[0010] The parameters of the shell-and-tube heat exchanger in the target scenario are input into the optimal critical velocity prediction model for prediction, and the critical velocity for helical instability of the shell-and-tube heat exchanger is obtained.

[0011] Preferably, an initial critical velocity prediction model based on a Gaussian kernel support vector machine is constructed based on a historical vector dataset, including:

[0012] Based on historical vector datasets, a critical flow velocity prediction model based on Gaussian kernel support vector machine regression is established.

[0013]

[0014] In the formula, y j , Denotes the Lagrange multiplier, k(x) j x i ) represents the Gaussian kernel function, a represents the correction coefficient, and x j x i Let N represent the feature parameters at time i and time j in the historical vector dataset, and let N represent the dimension of the feature parameters.

[0015] k(x j x i )=exp(-γ||x j -x i || 2 );

[0016] In the formula, γ represents the kernel function decay coefficient;

[0017] The minimum optimization algorithm is used to solve for the optimal parameters of the critical flow velocity prediction model based on Gaussian kernel support vector machine regression.

[0018]

[0019] In the formula, z i The i-th critical flow velocity measurement value in the vector dataset is used to perform global optimization to solve for the optimal parameters, thus obtaining the optimal parameters of the prediction model.

[0020] Preferably, the characteristic parameters include tube bundle arrangement, section diameter ratio, mass per unit tube length, tube diameter, tube frequency, liquid density, and damping ratio.

[0021] Preferably, the steps for the critical velocity prediction model to predict the critical velocity include:

[0022] The acquired dataset is mapped to a high-dimensional feature space, and each feature parameter forms its own feature vector;

[0023] Each feature vector is used as a node. Each time the model is optimized, the value of the node is also optimized. Based on the optimization of the node value, the system adaptively adjusts the relationship between nodes and the weight of nodes to form node links. The critical flow velocity is used as the central node. The central node and nodes, and node links, form a feature tree.

[0024] The node relationships obtained by processing the feature tree are transmitted to the decision layer of the prediction model to obtain the critical flow velocity and the corresponding feature parameter values.

[0025] Preferably, a loss function is set to optimize the prediction model, including

[0026] The critical velocity is calculated by using the values ​​of the corresponding characteristic parameters and the classical formula for calculating the critical velocity of helical instability in shell-and-tube heat exchangers.

[0027] A loss function is established by combining the predicted and calculated critical flow rates for each critical flow rate, and the prediction model is optimized using this loss function.

[0028] Preferably, the constraints of the prediction model during each optimization are as follows:

[0029]

[0030] Among them, U shell Indicates the shell-side flow velocity. Indicates the minimum flow velocity in the shell side. The maximum flow velocity in the shell side is represented by t, and the wall thickness of the heat exchange tube is represented by t. min A represents the minimum wall thickness of the heat exchange tube, and A represents the heat exchange area. min Indicates the minimum heat transfer area;

[0031] The objective function is: critical flow velocity U cr Process flow rate U operating .

[0032] A device for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger, comprising:

[0033] The first data acquisition module is used to acquire the historical structural and process parameters of the shell-and-tube heat exchanger.

[0034] The analysis module is used to analyze historical structural and process parameters to obtain characteristic parameters that affect the critical flow velocity of helical instability in shell-and-tube heat exchangers;

[0035] The second data acquisition module is used to acquire and preprocess historical feature parameters and corresponding critical flow velocity measurements to obtain a historical vector dataset.

[0036] The model building module is used to build an initial critical velocity prediction model based on historical vector datasets, initialize the hyperparameters of the initial critical velocity prediction model, train the initialized initial critical velocity prediction model, and obtain the optimal critical velocity prediction model.

[0037] The prediction module is used to input the parameters of the shell-and-tube heat exchanger in the target scenario into the optimal critical velocity prediction model for prediction, so as to obtain the critical velocity for helical instability of the shell-and-tube heat exchanger.

[0038] Therefore, this invention employs the aforementioned method and apparatus for predicting the critical flow velocity of helical instability in shell-and-tube heat exchangers. The method for predicting the critical flow velocity of helical instability in shell-and-tube heat exchangers is based on the vector machine learning approach. Vector machine models have strong generalization ability and robustness, making them suitable for datasets with limited samples. By collecting experimental parameters and critical flow velocities of the heat exchange tubes in the shell-and-tube heat exchanger to form the original dataset, and generating processed data through data cleaning and preprocessing, the hyperparameters are adjusted and cross-validated to obtain the best-performing model. This achieves accurate prediction of the critical flow velocity of helical instability in shell-and-tube heat exchangers, solving the problem of overly conservative results in traditional methods and saving on heat exchanger design costs. Attached Figure Description

[0039] Figure 1 This is a flowchart of a method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger according to the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0042] Example 1

[0043] like Figure 1 As shown, a method for predicting the critical velocity for helical instability in a shell-and-tube heat exchanger includes the following steps:

[0044] Obtain historical structural and process parameters of shell-and-tube heat exchangers;

[0045] Analysis of historical structural and process parameters yields characteristic parameters affecting the critical velocity of helical instability in shell-and-tube heat exchangers. These characteristic parameters include tube bundle arrangement, tube diameter ratio, mass per unit tube length, tube diameter, tube frequency, liquid density, and damping ratio.

[0046] Historical feature parameters and corresponding critical flow velocity measurements are acquired and preprocessed to obtain a historical vector dataset.

[0047] Collect characteristic parameters of shell-and-tube heat exchangers and critical flow velocities of heat exchange tubes;

[0048] The characteristic parameters of the shell-and-tube heat exchanger and the critical flow velocity of the heat exchange tubes are converted into a table and recorded as the original dataset. Each row in the original dataset is a sample data, and each column is a feature number.

[0049] Data cleaning and preprocessing

[0050] Raw datasets are usually unsuitable for machine learning and require cleaning and preprocessing.

[0051] Data cleaning includes removing duplicate samples, eliminating outliers, and filling missing values ​​with the mean of all data in the corresponding feature attribute column. This results in a complete and usable dataset.

[0052] The relationships between the parameters in the original dataset are not obvious, so the parameters need to be rearranged and combined by referring to empirical formulas (CONNORS) to enhance the correlation between the parameters.

[0053] The mass per unit length (M), pipe diameter (D), liquid density (ρ), and damping ratio (ζ) are combined to form the mass damping parameter (MDP). Its expression is as follows:

[0054] MDP=2πζM / ρD 2 ;

[0055] The critical flow velocity (U), pipe frequency (f), and pipe diameter (D) are used to construct the reduced flow velocity (U). r ), as a variable. Its expression is as follows:

[0056]

[0057] The mass damping parameters and reduced flow velocity are taken as logarithms with base 10.

[0058] MDP2 = log 10 MDP;

[0059] U r 2 = log 10 U r ;

[0060] This method enhances the correlation between various parameters.

[0061] The nodal diameter ratio and mass damping parameters are normalized.

[0062] The tube bundle arrangement is encoded and converted into numerical data.

[0063] Since the original data on the arrangement of shell-and-tube heat exchangers was text-based, it was encoded to convert it into numerical data. The arrangement characteristics of the tube bundles were changed from 1 column to 4 columns, with "1-0-0-1" representing a triangular arrangement; "0-1-0-0" representing a corner triangular arrangement; "0-0-1-0" representing a square arrangement; and "0-0-0-1" representing a corner square arrangement.

[0064] The parameters obtained from the processed data set DA2 include the heat exchanger tube arrangement (A), tube diameter ratio (P), mass damping parameter MDP2, and reduced flow velocity U. r .

[0065] Step 3, Dataset Partitioning

[0066] The processed dataset obtained in step 2 is divided by random sampling. 80% of the data is designated as the training set (Train) and used for training machine learning models and tuning parameters. The remaining 20% ​​is designated as the test set (Test) and used for model prediction and evaluation.

[0067] An initial critical velocity prediction model is constructed based on a historical vector dataset. The hyperparameters of the initial critical velocity prediction model are initialized. The initialized initial critical velocity prediction model is then trained to obtain the optimal critical velocity prediction model.

[0068] The parameters of the shell-and-tube heat exchanger in the target scenario are input into the optimal critical velocity prediction model for prediction, and the critical velocity for helical instability of the shell-and-tube heat exchanger is obtained.

[0069] An initial critical velocity prediction model based on a Gaussian kernel support vector machine is constructed using a historical vector dataset, including:

[0070] Based on historical vector datasets, a critical flow velocity prediction model based on Gaussian kernel support vector machine regression is established.

[0071]

[0072] In the formula, y j , Denotes the Lagrange multiplier, k(x) j x i ) represents the Gaussian kernel function, a represents the correction coefficient, and x j x i Let N represent the feature parameters at time i and time j in the historical vector dataset, and let N represent the dimension of the feature parameters.

[0073] k(x j x i )=exp(-γ||x j -x i ||2 );

[0074] In the formula, γ represents the kernel function decay coefficient;

[0075] The minimum optimization algorithm is used to solve for the optimal parameters of the critical flow velocity prediction model based on Gaussian kernel support vector machine regression.

[0076]

[0077] In the formula, z i The i-th critical flow velocity measurement value in the vector dataset is used to perform global optimization to solve for the optimal parameters, thus obtaining the optimal parameters of the prediction model.

[0078] The steps involved in predicting the critical velocity using a critical velocity prediction model include:

[0079] The acquired dataset is mapped to a high-dimensional feature space, and each feature parameter forms its own feature vector;

[0080] Each feature vector is used as a node. Each time the model is optimized, the value of the node is also optimized. Based on the optimization of the node value, the system adaptively adjusts the relationship between nodes and the weight of nodes to form node links. The critical flow velocity is used as the central node. The central node and nodes, and node links, form a feature tree.

[0081] The node relationships obtained by processing the feature tree are transmitted to the decision layer of the prediction model to obtain the critical flow velocity and the corresponding feature parameter values.

[0082] Set a loss function to optimize the prediction model, including

[0083] The critical velocity is calculated by using the values ​​of the corresponding characteristic parameters and the classical formula for calculating the critical velocity of helical instability in shell-and-tube heat exchangers.

[0084] The formula for calculating the critical velocity for flexural instability in a classic shell-and-tube heat exchanger is as follows:

[0085]

[0086] In the formula, δ s The mass damping parameter is dimensionless and is calculated using the following formula:

[0087]

[0088] b represents the exponent; d o Indicates the outer diameter of the heat exchange tube; f n The natural frequency of the heat exchanger tube is expressed in Hz; K. cρ represents the proportionality coefficient, determined based on the relationship between the heat exchanger tube arrangement, tube diameter ratio, and mass damping parameters; s represents the center-to-center distance of the heat exchanger tubes; m represents the mass per unit tube length; δ represents the logarithmic decay rate of the heat exchanger tubes, which is dimensionless; ρ o The density of the shell-side fluid is expressed in kg / m³. 3 ;

[0089]

[0090] Where E represents the elastic modulus and I represents the moment of inertia of the cross section. A represents the cross-sectional area, A = π(D) 4 -(D-2t) 2 / 4), L represents the span distance, t represents the pipe wall thickness, ρ tube This indicates the density of the pipe material.

[0091] A loss function is established by combining the predicted and calculated critical flow rates for each critical flow rate, and the prediction model is optimized using this loss function.

[0092] During each optimization, the constraints of the prediction model are:

[0093]

[0094] Among them, U shell Indicates the shell-side flow velocity. Indicates the minimum flow velocity in the shell side. The maximum flow velocity in the shell side is represented by t, and the wall thickness of the heat exchange tube is represented by t. min A represents the minimum wall thickness of the heat exchange tube, and A represents the heat exchange area. min Indicates the minimum heat transfer area;

[0095] The objective function is: critical flow velocity U cr Process flow rate U operating .

[0096] A device for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger, comprising:

[0097] The first data acquisition module is used to acquire the historical structural and process parameters of the shell-and-tube heat exchanger.

[0098] The analysis module is used to analyze historical structural and process parameters to obtain characteristic parameters that affect the critical flow velocity of helical instability in shell-and-tube heat exchangers;

[0099] The second data acquisition module is used to acquire and preprocess historical feature parameters and corresponding critical flow velocity measurements to obtain a historical vector dataset.

[0100] The model building module is used to build an initial critical velocity prediction model based on historical vector datasets, initialize the hyperparameters of the initial critical velocity prediction model, train the initialized initial critical velocity prediction model, and obtain the optimal critical velocity prediction model.

[0101] The prediction module is used to input the parameters of the shell-and-tube heat exchanger in the target scenario into the optimal critical velocity prediction model for prediction, so as to obtain the critical velocity for helical instability of the shell-and-tube heat exchanger.

[0102] Therefore, the present invention adopts the above-mentioned method and device for predicting the critical velocity of helical instability in shell-and-tube heat exchangers, so as to achieve accurate prediction of the critical velocity of helical instability in shell-and-tube heat exchangers, solve the problem that the results of traditional methods are too conservative, and save the design cost of heat exchangers.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger, characterized in that, Includes the following steps: Obtain historical structural and process parameters of shell-and-tube heat exchangers; Analysis of historical structural and process parameters yields characteristic parameters that influence the critical flow velocity for helical instability in shell-and-tube heat exchangers. Historical feature parameters and corresponding critical flow velocity measurements are acquired and preprocessed to obtain a historical vector dataset. An initial critical velocity prediction model is constructed based on a historical vector dataset. The hyperparameters of the initial critical velocity prediction model are initialized. The initialized initial critical velocity prediction model is then trained to obtain the optimal critical velocity prediction model. An initial critical velocity prediction model based on a Gaussian kernel support vector machine is constructed using a historical vector dataset, including: Based on historical vector datasets, a critical flow velocity prediction model based on Gaussian kernel support vector machine regression is established. In the formula, y j , Denotes the Lagrange multiplier, k(x) j x i ) represents the Gaussian kernel function, a represents the correction coefficient, and x represents the Gaussian kernel function. j x i Let N represent the feature parameters at time i and time j in the historical vector dataset, and let N represent the dimension of the feature parameters. k(x j ,x i )=exp(-γ||x j -x i || 2 ); In the formula, γ represents the kernel function decay coefficient; The minimum optimization algorithm is used to solve for the optimal parameters of the critical flow velocity prediction model based on Gaussian kernel support vector machine regression. In the formula, z i The i-th critical flow velocity measurement value in the vector dataset is used to perform global optimization to solve for the optimal parameters, thereby obtaining the optimal parameters of the prediction model. During each optimization, the constraints of the prediction model are: Among them, U shell Indicates the shell-side flow velocity. Indicates the minimum flow velocity in the shell side. The maximum flow velocity in the shell side is represented by t, and the wall thickness of the heat exchange tube is represented by t. min A represents the minimum wall thickness of the heat exchange tube, and A represents the heat exchange area. min Indicates the minimum heat transfer area; The objective function is: critical flow velocity U cr Process flow rate U operating ; The parameters of the shell-and-tube heat exchanger in the target scenario are input into the optimal critical velocity prediction model for prediction, and the critical velocity for helical instability of the shell-and-tube heat exchanger is obtained.

2. The method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger according to claim 1, characterized in that, The characteristic parameters include tube bundle arrangement, section diameter ratio, mass per unit tube length, tube diameter, tube frequency, liquid density, and damping ratio.

3. The method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger according to claim 1, characterized in that, The steps involved in predicting the critical velocity using a critical velocity prediction model include: The acquired dataset is mapped to a high-dimensional feature space, and each feature parameter forms its own feature vector; Each feature vector is used as a node. Each time the model is optimized, the value of the node is also optimized. Based on the optimization of the node value, the system adaptively adjusts the relationship between nodes and the weight of nodes to form node links. The critical flow velocity is used as the central node. The central node and nodes, and node links, form a feature tree. The node relationships obtained by processing the feature tree are transmitted to the decision layer of the prediction model to obtain the critical flow velocity and the corresponding feature parameter values.

4. The method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger according to claim 1, characterized in that, Set a loss function to optimize the prediction model, including The critical velocity is calculated by using the values ​​of the corresponding characteristic parameters and the classical formula for calculating the critical velocity of helical instability in shell-and-tube heat exchangers. A loss function is established by combining the predicted and calculated critical flow rates for each critical flow rate, and the prediction model is optimized using this loss function.

5. A device for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger, characterized in that, A method for predicting the critical velocity of helical instability in a shell-and-tube heat exchanger as described in any one of claims 1-4, comprising: The first data acquisition module is used to acquire the historical structural and process parameters of the shell-and-tube heat exchanger. The analysis module is used to analyze historical structural and process parameters to obtain characteristic parameters that affect the critical flow velocity of helical instability in shell-and-tube heat exchangers; The second data acquisition module is used to acquire and preprocess historical feature parameters and corresponding critical flow velocity measurements to obtain a historical vector dataset. The model building module is used to build an initial critical velocity prediction model based on historical vector datasets, initialize the hyperparameters of the initial critical velocity prediction model, train the initialized initial critical velocity prediction model, and obtain the optimal critical velocity prediction model. The prediction module is used to input the parameters of the shell-and-tube heat exchanger in the target scenario into the optimal critical velocity prediction model for prediction, so as to obtain the critical velocity for helical instability of the shell-and-tube heat exchanger.

Images