Plate heat exchanger performance rapid calculation and intelligent design method

Abstract

The invention discloses a plate heat exchanger performance rapid calculation and intelligent design method which comprises the following steps: (1) collecting actual operation data of plate heat exchangers in different industries, and storing the actual operation data in a data isolation form; a cloud platform data analysis center is used for storing an existing Colburnj factor and a Van Ning friction factor f calculation empirical formula, a corresponding application range, an experiment / simulation environment, experiment table data and simulation data, and preprocessing multi-source data; (2) constructing a thermodynamic calculation model containing adaptive j and f calculation of the plate heat exchanger, and realizing adaptive calculation of j and f by using a cross-working-condition weighted transfer learning method; (3) heat exchanger cold and hot side structure parameter collaborative multi-objective optimization based on an improved multi-objective frost ice optimization algorithm; (4) a designer or a user selects a design scheme meeting the actual application requirement by using cosine similarity-TOPSIS according to the actual application scene; according to the method, mutual learning experience among industries is ensured, and the problem of data leakage is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of heat exchanger optimization, and particularly to a method for rapidly calculating the performance of a plate heat exchanger and intelligent design. Background Art

[0002] Heat exchangers, as key devices for realizing heat conversion, are widely used in multiple fields such as energy power, food processing, air-conditioning heat pumps, petrochemical industry, etc. It can mutually convert part of the heat of hot and cold fluids, and is further subdivided into types such as evaporators, condensers, heaters, and coolers. In the context of the current increasingly tense energy situation, it is particularly important to improve the comprehensive performance of heat exchangers. Traditional research on the performance of heat exchangers mainly relies on experimental methods, which are costly and impractical. With the progress of computer technology, although CFD (Computational Fluid Dynamics) simulation provides a new way for the optimization of heat exchangers, its high computational cost and the large amount of computing time required in the face of complex research objects limit its application in rapidly assisting the design of heat exchangers. Moreover, there is no mature heat exchanger design software in China. Commercial heat exchanger design software is all owned by foreign countries, and domestic users are often restricted and warned regarding copyright issues, and there is a risk of being deactivated at any time, which severely restricts the development of China's industrial and commercial sectors. Therefore, there is an urgent need in China to develop a dedicated software for rapidly and efficiently designing heat exchangers.

[0003] When designing heat exchangers, the empirical formulas usually adopted are often obtained based on old experimental data or require manual adjustment of structural parameters for design. This method is not only time-consuming and laborious, but also difficult to adapt to the complex and changeable working conditions in practice. Changing the structural parameters once and then constructing the corresponding experimental bench or CFD simulation model to verify its heat transfer and flow performance are both time-consuming and require a large amount of financial support.

[0004] In addition, taking the heat exchangers in industrial equipment such as reforming furnaces, boilers, cracking furnaces, heating furnaces, steel-making furnaces, waste gas catalytic incinerators, and hot blast furnaces as examples, due to the involvement of national industrial development, the issue of industrial data security must be considered. Coupled with the insufficient degree of industrial intelligence in China, it is difficult to obtain the actual operation data of heat exchangers, resulting in a lack of data. These problems further increase the complexity and challenges of heat exchanger design. Summary of the Invention

[0005] Object of the Invention: The object of the present invention is to provide a method for rapidly calculating the performance of a plate heat exchanger and intelligent design, so as to solve the problems of old empirical formulas, strong subjectivity in manual parameter adjustment, low design accuracy of heat exchangers caused by backward industrial digitalization and small data volume, low design efficiency across software, and the fact that traditional design ignores the issue of industrial data security.

[0006] Technical Solution: A method for rapidly calculating the performance of a plate heat exchanger and intelligent design according to the present invention includes the following steps:

[0007] (1)Collect the actual operation data of plate heat exchangers in different industries and save them in the form of data isolation; use the data analysis center of the cloud platform to save the existing calculation empirical formulas of Colburn j factor and Fanning friction factor f, the corresponding applicable ranges, and the experimental / simulation environments, experimental bench data and simulation data, and preprocess the multi-source data;

[0008] (2)Construct a thermodynamic calculation model of the plate heat exchanger with adaptive j and f calculations, and use the cross-condition weighted transfer learning method to achieve adaptive calculation of j and f;

[0009] (3)Collaborative multi-objective optimization of the structural parameters on the cold and hot sides of the heat exchanger based on the improved multi-objective frost ice optimization algorithm;

[0010] (4)Designers or users select the design scheme that meets the actual application requirements according to the actual application scenario using cosine similarity - TOPSIS.

[0011] Furthermore, the preprocessing in step (1) is specifically as follows: cleaning, assimilation, and normalization processing.

[0012] Furthermore, in step (2), the thermodynamic modeling of the heat exchanger includes the following steps:

[0013] (21)Use the ε-NTU method for heat transfer calculation, and the formula is as follows:

[0014]

[0015] Among them, C r = C min / C max , C min and C max are the minimum and maximum heat capacities of the fluid respectively; C = mC p , C p is the specific heat at constant pressure, and m is the mass flow rate; the expression of NTU is:

[0016]

[0017] Among them, U is the total heat transfer coefficient; A is the total heat transfer area; h is the heat transfer coefficient; the heat transfer coefficient h is calculated using the heat transfer factor j:

[0018] h = j·G·C p ·Pr -(2 / 3)

[0019] Among them, j is solved using the cross-condition weighted transfer learning method;

[0020] The mass velocity G = m / A ff , A ffis the free flow area, expressed as;

[0021] A ffa =(H a -t a )(1 - n a t a )L b N a

[0022] A ffb =(H b -t b )(1 - n b t b )L a N b

[0023] where the subscript a represents the hot side and b represents the cold side;

[0024] In the heat exchanger, the number of fins N on the hot side a is less than the number of fins on the cold side, which can be expressed as: N a =N b -1. The heat transfer areas of the hot side and the cold side are expressed as:

[0025] A a =L a L b N a [1 + 2n a (H a -t a )]

[0026] A b =L a L b N b [1 + 2n b (H b -t b )]

[0027] The total heat transfer area A of the heat exchanger HT is expressed as

[0028] A HT =A a +A b

[0029] The heat transfer rate Q can be expressed as

[0030] Q = εC min (T a,1 -T b,1 )

[0031] (22) Calculate the pressure drop: The calculation method of the Reynolds number is:

[0032]

[0033] Among them, D h is the hydraulic diameter, expressed as:

[0034]

[0035] The friction factor f is solved by using the cross - operating - condition weighted transfer learning method;

[0036] The pressure drops on the hot side and the cold side are expressed as:

[0037]

[0038] Furthermore, in step (2), the cross - operating - condition weighted transfer learning method includes the following steps:

[0039] S21 Use the data generated in step (1) and adopt CNN for in - depth feature mining;

[0040] S22 Divide the source operating conditions and the target operating conditions according to the data volume of different operating conditions;

[0041] S23 Construct a j, f intelligent source learning model. For heat exchangers of different brands, adopt federated feature cross - vertical transfer learning to transfer the source learning model parameters to the target operating conditions;

[0042] S24 Use Q - Learning to automatically adjust the model parameters according to the influence of the operating conditions and actions on the heat transfer and flow of the heat exchanger, and generate j, f calculation expressions; where the actions include changing the flow rate and adjusting the heat exchanger parameters; the j, f calculation expressions include explicit expressions such as traditional empirical formulas and implicit relational expressions expressed by deep - learning models.

[0043] 5. A method for rapid performance calculation and intelligent design of a plate heat exchanger according to claim 1, characterized in that, in step (3), the multi - objective optimization includes: economic index TAC, physical properties, structural materials, and thermodynamic properties, i.e., entropy increase; where the physical properties include: heat transfer efficiency, pressure drop, flow - heat transfer coefficient, and the structural materials include: weight, total heat transfer area.

[0044] Furthermore, the specific process of the collaborative multi - objective optimization of the hot and cold side structural parameters of the heat exchanger based on the improved multi - objective frost - ice optimization algorithm is as follows:

[0045] (31) Conduct a sensitivity analysis of the main hot and cold side design parameters of the plate heat exchanger, and the number of them is denoted as dim.

[0046] (32) Determine the upper limit Ub and the lower limit Lb of the design parameters, set the operating parameters of the heat exchanger according to the actual operating conditions, and input the economic parameters of the heat exchanger;

[0047] (33) Set the maximum number of iterations, population size N, and external archive size Nr of the multi-objective frost ice optimization algorithm, and generate the initial population according to the cubic chaotic map combined with the dynamic reverse learning strategy; the formula is as follows:

[0048] P ij = repmat(Lb, N, 1)+C 1 (N, dim).*repmat((Ub - Lb), N, 1)

[0049]

[0050] where C 1～3 is a random number generated by the cubic chaotic map between 0 and 1, and repmat(Lb, N, 1) means replicating Lb into an N*1-dimensional matrix;

[0051] (34) Encode the hot and cold side design parameters as soft frost particles, and utilize the strong randomness and coverage of the frost particles to simulate the growth characteristics of soft frost for the optimization of design parameters;

[0052]

[0053] θ = πt / (10T max );

[0054] ξ = 1 - [(wt) / T max / w;

[0055]

[0056] where is a new set of hot and cold side structure design parameters; P best,j is the parameter that can obtain the best performance in the current set of all structure design parameters; C 4 is a random number in the range of (-1, 1), which controls the optimization direction of the structure design parameters and changes with θ as the number of iterations increases; ξ is the environmental factor, which simulates the influence of the external environment according to the iteration to ensure the convergence of the algorithm; [·] represents rounding; the default value of w is 5, which controls the number of segments of the step function to avoid premature convergence of the algorithm; ψ is the adhesion degree, which is a random number in the range of (0, 1) and is used to control the distance between the centers of two sets of design parameters; t is the current number of iterations; T max is the maximum number of iterations; E is the adhesion coefficient, which affects the mutation probability of the alternative design parameters and increases with the increase of the number of iterations; C 5 is a random number in the range of (0, 1), and it and the adhesion coefficient control whether the design parameters need to mutate;

[0057] (35) According to the hard frost growth mechanism, multiple sets of design parameter sets are cross - fused to improve the ability to jump out of the local optimum in the direction of finding the optimal structural design parameter set within the feasible region of the research problem. The formula is as follows:

[0058] C 6 <F nor (P j )

[0059] Among them, C 6 is a random number between (-1, 1);

[0060] (36) By comparing the objective function values of the updated design parameters with those of the pre - updated design parameters, ensure that the population evolves in a better direction in each iteration.

[0061] Objective function calculation:

[0062]

[0063] Among them, the index with subscript ref represents the performance index corresponding to the reference design scheme;

[0064] (37) According to non - dominated sorting and crowding distance, select the better population and update the population using the non - linear Gaussian - Cauchy mutation perturbation strategy:

[0065]

[0066] In the formula represents the design parameters after non - linear Gaussian - Cauchy mutation perturbation, R 1 is the non - linear inertia factor, Cauchy is a random variable that satisfies the Cauchy distribution, and Gaussian is a random variable that satisfies the Gaussian distribution;

[0067] (38) Repeat steps 32 - 38 until the maximum number of iterations is reached or other stopping conditions are met;

[0068] (39) Output the final Pareto front, that is, the set of non - dominated design schemes and the corresponding heat exchanger performance indicators.

[0069] Furthermore, in step (4), the specific process of improving C - TOPSIS is as follows:

[0070] (41) Construct a three - dimensional matrix by corresponding the heat exchanger design scheme quantity, design variables of the design scheme, and performance indicators;

[0071] (42) The user inputs the degree of importance for different performances of each heat exchanger, and the system automatically calculates the user's subjective weight w sub ; Calculate the objective weight w according to the information entropyobj , based on this, the comprehensive weight w is obtained overall , and the calculation formula of the comprehensive weight is as follows:

[0072] w overall =(1 - a 1 )w obj + a 1 w sub

[0073] In the formula, a 1 is an empirical correction factor, which is used to judge whether the importance attached by the user to different performances of the heat exchanger is worthy of reference according to whether the user has engineering experience, make full use of the actual engineering experience of senior experts, and avoid the uncertainty introduced by excessive intervention of inexperienced users;

[0074] (43) Construct a weighted normalized decision matrix, and determine the ideal optimal solution and the ideal worst solution, where each performance index of the ideal optimal solution is the best, and each performance index of the ideal worst solution is the worst;

[0075] (44) Calculate the comprehensive goodness: Calculate the cosine similarity between each design solution and the ideal optimal solution D idealbest and the ideal worst solution D idealworst ; Calculate the comprehensive goodness of each solution according to the cosine similarity; The formula is as follows:

[0076]

[0077] In the formula, represents the weight column vector, D i =(index 1 , index 2 , …, index m ) represents the performance index vector of the i-th design solution; The subscript w represents after weighting.

[0078] A plate heat exchanger performance rapid calculation and intelligent design system described in the present invention includes:

[0079] A preprocessing module: used to collect the actual operation data of plate heat exchangers in different industries and store them in the form of data isolation; Utilize the data analysis center of the cloud platform to store the existing Colburn j factor and Fanning friction factor f calculation empirical formulas, corresponding applicable ranges, and experimental / simulation environments, experimental bench data and simulation data, and preprocess the multi-source data;

[0080] A thermodynamics module: used to construct a thermodynamics calculation model of a plate heat exchanger with adaptive j and f calculations, and use the cross-condition weighted transfer learning method to achieve adaptive calculation of j and f;

[0081] Multi-objective optimization module: used for collaborative multi-objective optimization of the structural parameters on the cold and hot sides of the heat exchanger based on the improved multi-objective frost ice optimization algorithm;

[0082] Cosine similarity module: used for designers or users to select design solutions that meet the actual application requirements according to the actual application scenario by using cosine similarity - TOPSIS.

[0083] An electronic device according to the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements the method for rapid performance calculation and intelligent design of a plate heat exchanger described in any one of the above.

[0084] A storage medium according to the present invention stores a computer program, and when the computer program is executed by a processor, it implements the method for rapid performance calculation and intelligent design of a plate heat exchanger described in any one of the above.

[0085] Advantageous effects: Compared with the prior art, the present invention has the following remarkable advantages:

[0086] 1. The present invention provides a method and system for intelligent optimization decision-making of the performance of a plate heat exchanger, providing ideas for domestic self-developed heat exchanger design software with independent intellectual property rights. By self-programming the thermodynamic calculation model and all calculation programs for intelligent optimization, it solves the copyright restrictions of relying on commercial software and also avoids the limitation of the time-consuming and laborious manual setting and adjustment of parameters step by step, greatly accelerating the design time of the heat exchanger.

[0087] 2. The system is provided with an experience correction factor in the decision-making process. On the one hand, it can learn and accept the engineering experience of senior experts, and at the same time avoid the uncertainty introduced by the ineffective intervention of inexperienced users, improving the design accuracy of the heat exchanger;

[0088] 3. When designing the system, the non-uniformity of the cold and hot sides is considered, as well as various factors such as cost, structure, and performance, which is more in line with the complex and changeable actual working conditions, improving the design accuracy and efficiency of the heat exchanger and reducing the manufacturing cost of the heat exchanger.

[0089] 4. Adaptive adjustment of the heat transfer factor j and the friction factor f improves the inclusiveness and expansibility of the system.

[0090] 5. The design system of the present invention can isolate and save the operation data of plate heat exchangers in different industries in the data acquisition module, and at the same time store general physical property data, simulation data, and empirical formulas and other general data in the cloud platform data analysis center for convenient sharing and updating, which not only ensures the mutual learning of experience between industries but also avoids the problem of data leakage. Description of the Drawings

[0091] Figure 1Data flow direction of the cloud platform of the present invention;

[0092] Figure 2 Flow chart of the thermodynamic calculation model of the plate heat exchanger with adaptive j and f calculations of the present invention;

[0093] Figure 3 System diagram of the present invention. Detailed implementation manners

[0094] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0095] As Figure 1 and Figure 3 shown, an embodiment of the present invention provides a method for rapid calculation and intelligent design of the performance of a plate heat exchanger, including the following steps:

[0096] S1: The cloud platform data analysis center stores the existing Colburn j factor and Fanning friction factor f calculation empirical formulas, corresponding applicable ranges, and experimental / simulation environments, experimental bench data and simulation data, and cleans, assimilates, and normalizes the multi-source data to facilitate the adaptive calculation of j and f factors for later modeling.

[0097] S2: Construct a thermodynamic calculation model of the plate heat exchanger with adaptive j and f calculations, and use the cross-condition weighted transfer learning method to achieve the adaptive calculation of j and f; as Figure 1 shown, it is the data flow direction of steps S1 and S2. Figure 2 It is the calculation flow chart of the thermodynamic calculation model of the plate heat exchanger with adaptive j and f calculations. The thermodynamic modeling of the heat exchanger includes the following steps:

[0098] Use the ε-NTU method for heat transfer calculation:

[0099] The ε-NTU method is described as follows:

[0100]

[0101] In the formula:

[0102] C r = C min / C max C min and C max are the minimum and maximum heat capacities of the fluid, respectively.

[0103] C = mC p C p is the specific heat at constant pressure, and m is the mass flow rate.

[0104] The expression of NTU is:

[0105]

[0106] In the formula:

[0107] U is the overall heat transfer coefficient;

[0108] A is the total heat transfer area;

[0109] h is the heat transfer coefficient;

[0110] The heat transfer coefficient h can be calculated using the heat transfer factor j:

[0111] h = j·G·C p ·Pr -(2 / 3)

[0112] In the formula, j is solved using the cross-condition weighted transfer learning method.

[0113] The mass flow rate G = m / A ff ,A ff is the free flow area and can be expressed as;

[0114] A ffa =(H a -t a )(1 - n a t a )L b N a

[0115] A ffb =(H b -t b )(1 - n b t b )L a N b

[0116] In the formula, the subscript a represents the hot side and b represents the cold side;

[0117] In the heat exchanger, the number of fins N on the hot side a is less than the number of fins on the cold side and can be expressed as: N a =N b -1. The heat transfer areas of the hot side and the cold side can be expressed as:

[0118] A a =L a L b N a [1 + 2n a (H a -t a )]

[0119] A b =L a L b N b [1 + 2nb (H b -t b )]

[0120] Therefore, the total heat transfer area A of the heat exchanger HT can be expressed as

[0121] A HT = A a + A b

[0122] The heat transfer rate Q can be expressed as

[0123] Q = εC min (T a,1 - T b,1 )

[0124] Pressure drop calculation:

[0125] The calculation method of the Reynolds number is:

[0126]

[0127] where D h is the hydraulic diameter and is expressed as:

[0128]

[0129] The friction factor f is solved by using the cross-condition weighted transfer learning method.

[0130] The pressure drops on the hot side and the cold side can be expressed as:

[0131]

[0132] The cross-condition weighted transfer learning method includes the following steps:

[0133] a Use the data generated in step S1 and adopt CNN for in-depth feature mining;

[0134] b Divide the source condition and the target condition according to the data volume under different conditions;

[0135] c Construct the j, f intelligent source learning model, and for heat exchangers of different brands, use the federated feature horizontal and vertical transfer learning to transfer the source learning model parameters to the target condition;

[0136] d Use Q-Learning to automatically adjust the model parameters according to the influence of the condition and the action on the heat transfer and flow of the heat exchanger, and generate the j, f calculation expressions.

[0137] The actions include but are not limited to changing the flow rate and adjusting the heat exchanger parameters.

[0138] The calculation expressions of j and f are not limited to the explicit expressions of traditional empirical formulas, but also include implicit relational expressions expressed by deep learning models.

[0139] S3: Coordinated multi-objective optimization of the structural parameters on the cold and hot sides of the heat exchanger based on the improved multi-objective frost ice optimization algorithm, considering economic indicators (TAC), physical properties (heat transfer efficiency, pressure drop, flow heat transfer coefficient), structural materials (weight, total heat transfer area), and thermodynamic properties (entropy increase). The specific process of the coordinated multi-objective optimization of the structural parameters on the cold and hot sides of the heat exchanger based on the improved multi-objective frost ice optimization algorithm is as follows:

[0140] S31 Conduct sensitivity analysis on the main design parameters on the cold and hot sides of the plate heat exchanger, and the number of them is denoted as dim.

[0141] S32 Determine the upper limit Ub and the lower limit Lb of the design parameters, set the operating parameters of the heat exchanger according to the actual operating conditions, and input the economic parameters of the heat exchanger;

[0142] S33 Set the maximum number of iterations, population size (N), and external archive size (Nr) of the multi-objective frost ice optimization algorithm, and generate the initial population according to the cubic chaotic mapping combined with the dynamic reverse learning strategy.

[0143] P ij = repmat(Lb, N, 1) + C 1 (N, dim).*repmat((Ub - Lb), N, 1)

[0144]

[0145] In the formula, C 1～3 is a random number generated by the cubic chaotic mapping between 0 and 1, and repmat(Lb, N, 1) means replicating Lb into an N*1-dimensional matrix;

[0146] S34 Encode the design parameters on the cold and hot sides as soft frost particles, and utilize the strong randomness and coverage of the frost particles to simulate the growth characteristics of soft frost for the optimization of design parameters;

[0147]

[0148] θ = πt / (10T max );

[0149] ξ = 1 - [(wt) / T max / w;

[0150]

[0151] Among them, is a new set of structural design parameters on the cold and hot sides; P best,jis the parameter that can obtain the best performance among all current structural design parameter sets; C 4 is a random number within the range of (-1, 1), which controls the optimization direction of structural design parameters and changes with the number of iterations along with θ; ξ is the environmental factor, which simulates the influence of the external environment according to the iteration to ensure the convergence of the algorithm; [·] represents rounding; the default value of w is 5, which controls the number of segments of the step function to avoid premature convergence of the algorithm; ψ is the adhesion degree, which is a random number within the range of (0, 1) and is used to control the distance between the centers of two groups of design parameters; t is the current number of iterations; T max is the maximum number of iterations; E is the adhesion coefficient, which affects the mutation probability of alternative design parameters and increases with the number of iterations; C 5 is a random number within the range of (0, 1), and it and the adhesion coefficient control whether the design parameters need to mutate.

[0152] According to the hard frost growth mechanism, S35 cross - fuses multiple groups of design parameter sets to enhance the ability to jump out of local optima in the direction of finding the optimal structural design parameter set within the feasible region of the research problem.

[0153] C 6 <F nor (P j )

[0154] In the formula, C 6 is a random number between (-1, 1);

[0155] S36 ensures the evolution of the population in a better direction in each iteration by comparing the objective function values of the updated design parameters with those of the pre - updated design parameters.

[0156] Objective function calculation:

[0157]

[0158] In the formula, the index with the subscript ref represents the performance index corresponding to the reference design scheme.

[0159] S37 selects the better population according to non - dominated sorting and crowding distance, and updates the population using the non - linear Gaussian - Cauchy mutation perturbation strategy:

[0160]

[0161] In the formula represents the design parameters after non - linear Gaussian - Cauchy mutation perturbation, R 1 is the non - linear inertia factor, Cauchy is a random variable that satisfies the Cauchy distribution, and Gaussian is a random variable that satisfies the Gaussian distribution.

[0162] Repeat steps S32 - S38 until the maximum number of iterations is reached or other stopping conditions are met.

[0163] S39 Output result: Output the final Pareto front, that is, the set of non - dominated design solutions and the corresponding heat exchanger performance indicators.

[0164] S4: The designer or user selects the design solution that meets the actual application requirements using cosine similarity - TOPSIS (C - TOPSIS) according to the actual application scenario. The specific process of improving C - TOPSIS is as follows:

[0165] S41: Construct a three - dimensional matrix corresponding to the heat exchanger design solutions, design variables of the design solutions, and performance indicators.

[0166] S42: The user inputs the degree of importance attached to different performances of each heat exchanger, and the system automatically calculates the user's subjective weight w sub , and in addition, calculate the objective weight w obj based on information entropy, and obtain the comprehensive weight w overall . The comprehensive weight calculation formula is:

[0167] w overall =(1 - a 1 )w obj +a 1 w sub

[0168] In the formula, a 1 is the empirical correction factor, used to judge whether the degree of importance attached by the user to different performances of the heat exchanger is worthy of reference according to whether the user has engineering experience, making full use of the actual engineering experience of senior experts and avoiding the uncertainty introduced by excessive intervention of inexperienced users.

[0169] S43: Construct a weighted normalized decision matrix, and determine the ideal optimal solution and the ideal worst solution, where each performance indicator of the ideal optimal solution is the best, and each performance indicator of the ideal worst solution is the worst.

[0170] S44: Calculate the comprehensive goodness. Calculate the cosine similarity between each design solution and the ideal optimal solution D idealbest and the ideal worst solution D idealworst ; calculate the comprehensive goodness of each solution according to the cosine similarity.

[0171]

[0172] In the formula, represents the weight column vector, D i =(index 1 ,index 2 ,…,indexm ) represents the performance index vector of the i-th design scheme; the subscript w indicates after weighting.

[0173] An embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements any one of the methods for rapid performance calculation and intelligent design of a plate heat exchanger.

[0174] An embodiment of the present invention further provides a storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, it implements any one of the methods for rapid performance calculation and intelligent design of a plate heat exchanger.

Claims

1. A plate heat exchanger performance rapid calculation and intelligent design method, characterized in that: The following steps are involved: (1) Collect actual operating data of plate heat exchangers in different industries and save them in the form of data isolation; use the cloud platform data analysis center to save the existing Colburn j factor and Fanning friction factor f calculation empirical formula, the corresponding applicable scope and experimental / simulation environment, bench data and simulation data, and pre-process multi-source data; (2) Construct a thermodynamic calculation model of plate heat exchanger with adaptive j and f calculation, and use cross-operating condition weighted transfer learning method to realize adaptive calculation of j and f; (3) Collaborative multi-objective optimization of heat exchanger cold and hot side structural parameters based on an improved multi-objective frost optimization algorithm; (4) Designers or users use cosine similarity-TOPSIS to select design solutions that meet actual application needs based on actual application scenarios.

2. A plate heat exchanger performance rapid calculation and intelligent design method according to claim 1, characterized in that: The preprocessing in step (1) is specifically as follows: cleaning, assimilation, and normalization.

3. A plate heat exchanger performance rapid calculation and intelligent design method according to claim 1, characterized in that: In step (2), the heat exchanger thermodynamic modeling includes the following steps: (21) The heat transfer calculation is performed using the ε-NTU method, and the formula is as follows: Among them, C r =C min / C max , C min and C max are the minimum and maximum heat capacities of the fluid respectively; C = mC p , C p is the specific heat at constant pressure, m is the mass flow rate; the expression of NTU is: Among them, U is the total heat transfer coefficient; A is the total heat transfer area; h is the heat transfer coefficient; the heat transfer coefficient h is calculated using the heat transfer factor j: h=j G C p ·Pr -(2 / 3) Among them, j is solved using the cross-operating condition weighted transfer learning method; Mass flow rate G = m / A ff ,A ff is the free flow area, expressed as; A ffa =(H a -t a )(1-n a t a )L b N a A ffb =(H b -t b )(1-n b t b )L a N b Wherein, subscript a represents the hot side and b represents the cold side; In the heat exchanger, the number of hot side fin layers N a Less than the number of fins on the cold side, which can be expressed as: N a =N b -1. The heat transfer area on the hot side and the cold side is expressed as: A a =L a L b N a [1+2n a (H a -t a )] A b =L a L b N b [1+2n b (H b -t b )] Total heat exchange area of ​​heat exchanger A HT Expressed as A HT =A a +A b The heat transfer rate Q can be expressed as Q=εC min (T a,1 -T b,1 ) (22) Calculation of pressure drop: The Reynolds number calculation method is: Among them, D h is the hydraulic diameter, expressed as: The friction factor f is solved using the cross-operating condition weighted transfer learning method; The pressure drop on the hot and cold sides is expressed as:

4. A plate heat exchanger performance rapid calculation and intelligent design method and system according to claim 1, characterized in that: In step (2), the cross-operating condition weighted transfer learning method includes the following steps: S21 uses the data generated in step (1) to perform deep feature mining using CNN; S22 divides the source working condition and the target working condition according to the amount of data of different working conditions; S23 builds j and f intelligent source learning models. For heat exchangers of different brands, federated feature horizontal and vertical transfer learning is used to migrate the source learning model parameters to the target working conditions. S24 uses Q-Learning to automatically adjust model parameters and generate j and f calculation expressions based on the impact of operating conditions and actions on the heat transfer and flow of the heat exchanger. Actions include changing the flow rate and adjusting heat exchanger parameters. The j and f calculation expressions include explicit expressions such as traditional empirical formulas and implicit relationships expressed by deep learning models.

5. A plate heat exchanger performance rapid calculation and intelligent design method according to claim 1, characterized in that: In step (3), the multi-objective optimization includes: economic indicator TAC, physical properties, structural materials, and thermodynamic properties, namely entropy increase; wherein the physical properties include: heat transfer efficiency, pressure drop, flow heat transfer coefficient, and the structural materials include: weight and total heat transfer area.

6. A plate heat exchanger performance rapid calculation and intelligent design method according to claim 5, characterized in that: The specific process of collaborative multi-objective optimization of the cold and hot side structural parameters of the heat exchanger based on the improved multi-objective frost and ice optimization algorithm is as follows: (31) Sensitivity analysis of the main design parameters of the hot and cold sides of the plate heat exchanger, the number of which is denoted as dim. (32) Determine the upper limit Ub and lower limit Lb of the design parameters, set the heat exchanger operating parameters according to the actual operating conditions, and enter the heat exchanger economic parameters; (33) Set the maximum number of iterations of the multi-objective frost optimization algorithm, the population size N, the external archive size Nr, and generate the initial population based on the cubic chaos map combined with the dynamic reverse learning strategy; the formula is as follows: <h2 style=";text-align:left;direction:ltr">P<h2 style=";text-align:left;direction:ltr"> ij <h2 style=";text-align:left;direction:ltr"> =repmat(Lb,N,1)+C1(N,dim).*repmat((Ub-Lb),N,1) Among them, C 1～3 Generates random numbers between 0 and 1 for the cubic chaotic map, repmat(Lb,N,1) means to copy Lb into an N*1 dimensional matrix; (34) The design parameters of the hot and cold sides are encoded into soft frost particles, and the strong randomness and coverage of frost particles are used to simulate the growth characteristics of soft frost to optimize the design parameters; θ=πt / (10T max ); ξ=1-[(wt) / T max ] / w; in, is a new set of cold and hot side structural design parameters; P best,j is the parameter that can obtain the best performance among all the current structural design parameter sets; C4 is a random number in the range of (-1, 1), which controls the optimization direction of the structural design parameters and changes with θ as the number of iterations increases; ξ is an environmental factor, which simulates the influence of the external environment according to the iteration to ensure the convergence of the algorithm; [·] indicates rounding; the default value of w is 5, which controls the number of segments of the step function to avoid premature convergence of the algorithm; ψ is the degree of adhesion, which is a random number in the range of (0, 1) and is used to control the distance between the centers of the two sets of design parameters; t is the current number of iterations; T max is the maximum number of iterations; E is the adhesion coefficient, which affects the probability of variation of the alternative design parameters and increases with the number of iterations; C5 is a random number in the range of (0, 1), which controls whether the design parameters need to vary with the adhesion coefficient; (35) According to the hard frost growth mechanism, multiple sets of design parameter sets are cross-fused to improve the ability to find the direction of the optimal structural design parameter set within the feasible domain of the research problem and escape from the local optimum. The formula is as follows: Among them, C6 is a random number between (-1,1); (36) By comparing the objective function values ​​of the updated design parameters with those before the update, it is ensured that the population evolves in a better direction in each iteration. Objective function calculation: Among them, the index with the subscript ref represents the performance index corresponding to the reference design solution; (37) According to the non-dominated sorting and crowding distance, a better population is selected, and the population is updated using the nonlinear Gauss-Cauchy mutation perturbation strategy: In the formula represents the design parameters after nonlinear Gauss-Cauchy variation disturbance, R1 is the nonlinear inertia factor, Cauchy is a random variable that satisfies the Cauchy distribution, and Gaussian is a random variable that satisfies the Gaussian distribution; (38) Repeat steps 32-38 until the maximum number of iterations is reached or other stopping conditions are met; (39) Output the final Pareto frontier, that is, the set of non-inferior design solutions and the corresponding heat exchanger performance indicators.

7. A plate heat exchanger performance rapid calculation and intelligent design method according to claim 1, characterized in that: In step (4), the specific process of improving C-TOPSIS is: (41) Construct a three-dimensional matrix corresponding to the heat exchanger design scheme quantity, design scheme design variables and performance indicators; (42) The user inputs the importance they attach to the different performance of each heat exchanger, and the system automatically calculates the user's subjective weight w sub ; Calculate the objective weight w based on information entropy obj Based on this, we get the comprehensive weight w overall , the comprehensive weight calculation formula is: In overall =(1-a1)in obj +a1w sub Where a1 is the experience correction factor, which is used to judge whether the user has engineering experience to determine whether the importance he attaches to different performance of the heat exchanger is worth referring to, making full use of the actual engineering experience of senior experts and avoiding the uncertainty introduced by excessive intervention of inexperienced users; (43) constructing a weighted normalized decision matrix, and determining an ideal optimal solution and an ideal worst solution, wherein each performance indicator of the ideal optimal solution is optimal, and each performance indicator of the ideal worst solution is the worst; (44) Calculate the comprehensive merit: Calculate the difference between each design scheme and the ideal optimal scheme D idealbest and the ideal worst solution D idealworst The cosine similarity of each solution is calculated according to the cosine similarity. The formula is as follows: In the formula, represents the weight column vector, D i =(index1,index2,…,index m ) represents the performance index vector of the i-th design scheme; the subscript w represents the weighted one.

8. A plate heat exchanger performance rapid calculation and intelligent design system, characterized in that: include: Preprocessing module: used to collect actual operation data of plate heat exchangers in different industries and save them in the form of data isolation; Using the cloud platform data analysis center, save the existing Colburn j factor and Fanning friction factor f calculation empirical formulas, the corresponding applicable scope and experimental / simulation environment, test bench data and simulation data, and pre-process the multi-source data; Thermodynamics module: used to build a thermodynamic calculation model for plate heat exchangers with adaptive j and f calculations, and to achieve adaptive calculations of j and f using cross-operating condition weighted transfer learning methods; Multi-objective optimization module: used for collaborative multi-objective optimization of the cold and hot side structural parameters of the heat exchanger based on the improved multi-objective frost and ice optimization algorithm; Cosine similarity module: It is used by designers or users to select design solutions that meet actual application needs based on actual application scenarios using cosine similarity-TOPSIS.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, a plate heat exchanger performance rapid calculation and intelligent design method according to any one of claims 1 to 7 is implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for rapid performance calculation and intelligent design of a plate heat exchanger according to any one of claims 1 to 7 is implemented.

Images