Optimization method for low-temperature deep-well pump impeller based on PSO algorithm

Abstract

The invention discloses a low-temperature deep-well pump impeller optimization method based on a PSO algorithm. The low-temperature deep-well pump impeller optimization method comprises the following steps that parametric modeling is conducted on an impeller of a low-temperature deep-well pump; combining the impeller subjected to parametric modeling with other flow passage components, and performing numerical simulation calculation to obtain the hydraulic performance of the low-temperature deep-well pump model; the hydraulic performance comprises lift and efficiency; if the low-temperature deep-well pump model needs to be optimized, sensitivity analysis is carried out with geometric design parameters of the impeller as input parameters and hydraulic performance of the low-temperature deep-well pump model as output parameters, a multi-objective optimization MOP agent model is constructed according to a sensitivity analysis result, the multi-objective optimization MOP agent model is optimized based on a particle swarm optimization algorithm, and the low-temperature deep-well pump model is obtained. And obtaining an optimized impeller model. Compared with a traditional optimization method, the method can provide reliable design reference for designers lacking rich design experience, so that the design quality is improved, the calculation precision is ensured, the optimization process is remarkably accelerated, and the required optimization time is shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of low-temperature deep-well pump impellers, and particularly to an optimization method for low-temperature deep-well pump impellers based on the PSO algorithm. Background Art

[0002] Low-temperature deep-well pumps are widely used in LNG ships and other LNG-related fields and have become important pump equipment. With the continuous development of the LNG industry, higher requirements have been put forward for performance indicators such as the efficiency and head of low-temperature deep-well pumps. Since the initial designed low-temperature deep-well pumps have low hydraulic performance, it is necessary to optimize the hydraulic performance of the pumps. Through optimization, not only can the hydraulic performance of the pumps be improved, but also energy consumption can be reduced, thus providing technical support for the sustainable development of LNG-related fields.

[0003] In the field of low-temperature deep-well pump design, designers need to handle higher-dimensional and faster-changing design requirements of low-temperature deep-well pumps, and at the same time, they also need to find excellent design solutions under more computational costs and time constraints. To solve these design problems, compared with traditional methods, intelligent optimization methods have high efficiency and powerful search capabilities. Usually, an approximation model is first constructed, such as the response surface method, Kriging model or artificial neural network. After establishing the approximation model, an intelligent optimization algorithm is selected and implemented to optimize it. The algorithms include particle swarm optimization, genetic algorithm and ant colony algorithm, etc. Using intelligent optimization methods can find design solutions with better performance at a faster speed, effectively improve the performance of the original low-temperature deep-well pumps and shorten the design cycle.

[0004] When designing low-temperature deep-well pumps, designers lacking rich experience often cause the pumps designed for the first time to fail to meet the specified requirements in terms of hydraulic performance, and thus need to be continuously optimized. Traditional optimization methods highly rely on the rich design and optimization experience of designers. Those without rich experience may lead to blindness in the optimization direction, significantly increase the time required for optimization, and the instability of the optimization results is also relatively high. Summary of the Invention

[0005] Aiming at the shortcomings of designers lacking rich experience in designing low-temperature deep-well pumps, the present invention provides an optimization method for low-temperature deep-well pump impellers based on the PSO algorithm. By combining sensitivity analysis and the particle swarm algorithm (PSO), a low-temperature deep-well pump can be designed quickly and effectively, so that the specified hydraulic performance requirements are met during operation.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An optimization method for low-temperature deep-well pump impellers based on the PSO algorithm, comprising the following steps:

[0008] S1. Parametrically model the impeller of the cryogenic deep well pump;

[0009] S2. Combine the parametrically modeled impeller with other flow-through components and perform numerical simulation calculations to obtain the hydraulic performance of the cryogenic deep well pump model; the hydraulic performance includes head and efficiency;

[0010] S3. Compare the hydraulic performance of the cryogenic deep well pump model with the specified hydraulic performance to determine whether the cryogenic deep well pump model needs to be optimized; if optimization is required, then execute step S4;

[0011] S4. Use the geometric design parameters of the impeller as input parameters and the hydraulic performance of the cryogenic deep well pump model as output parameters to perform sensitivity analysis, construct a multi-objective optimization MOP surrogate model based on the sensitivity analysis results, and then optimize the multi-objective optimization MOP surrogate model based on the particle swarm optimization algorithm to obtain an optimized impeller model.

[0012] Preferably, step S4 includes:

[0013] S41. Define the geometric design parameters of the impeller as input parameters, define the hydraulic performance of the cryogenic deep well pump model as output parameters, and use maximizing head and maximizing efficiency as multi-objective functions to perform sensitivity analysis using the global sensitivity analysis method;

[0014] S42. Construct a sample library for the input parameters according to the sensitivity analysis results;

[0015] S43. Construct a multi-objective optimization response surface model based on the sample library;

[0016] S44. According to the multi-objective optimization response surface model, screen out the target input parameters that have the greatest impact on the response variables;

[0017] S45. Based on these target input parameters, construct a multi-objective optimization MOP surrogate model and use the particle swarm optimization algorithm for multi-objective global optimization;

[0018] S46. Find the optimal solution in the Pareto solution set obtained from the multi-objective global optimization to obtain an optimized impeller model;

[0019] S47. Select the remaining input parameters other than the target input parameters, re-parametrically model the optimized impeller model, still using maximizing efficiency and maximizing head as the optimization objectives of the multi-objective function, and repeat the S41 - S46 optimization process to obtain a final optimized impeller model.

[0020] Preferably, step S42 includes:

[0021] S421. According to the sensitivity analysis results, select appropriate ranges of variation for the input parameters;

[0022] S422. Define the optimization objectives of the highest efficiency and the maximum head as multi-objective functions;

[0023] S423. Based on the Monte Carlo sampling method, randomly generate a large number of sample points within the range of input parameter changes, and then construct a basic sample database;

[0024] S424. Through experimental design, specify the efficiency and head of the low-temperature deep-well pump as response variables, and use the advanced Latin hypercube sampling method to sample multiple sample points in the basic sample database, and then construct a sample library.

[0025] Preferably, the geometric design parameters of the impeller include: the impeller inlet diameter, impeller outlet diameter, impeller outlet width, blade inlet setting angle, blade outlet setting angle, blade wrap angle, blade thickness, number of blades, and blade inlet edge position of the low-temperature deep-well pump.

[0026] Preferably, S45 includes:

[0027] S451. Based on these subsets of target input parameters, construct a high-quality response surface fitting model;

[0028] S452. Use the response surface fitting model to construct response surfaces for each response variable;

[0029] S453. Integrate all response surfaces into a multi-objective optimization MOP surrogate model;

[0030] S454. Adopt a pre-trained particle swarm optimization algorithm to perform multi-objective global optimization on the multi-objective optimization MOP surrogate model.

[0031] Preferably, step S46 includes:

[0032] S461. Find the optimal solution that meets the optimization objectives in the Pareto front solution set distribution obtained by multi-objective global optimization to obtain the optimized impeller model.

[0033] The beneficial effects of the present invention are:

[0034] Combining the sensitivity analysis method with the particle swarm optimization algorithm can quickly obtain a hydraulic model that meets the hydraulic performance requirements without rich design experience. Compared with traditional optimization methods, the optimization method that combines the sensitivity analysis method, the multi-objective optimization MOP surrogate model, and the particle swarm optimization algorithm for the low-temperature deep-well pump can significantly improve the hydraulic performance of the initial designed low-temperature deep-well pump, shorten the optimization time, and promote efficient optimization. The optimal prediction meta-model for target optimization constructed based on the sensitivity analysis results can quickly predict the hydraulic performance under different combinations of input parameters. The particle swarm optimization algorithm is used to globally optimize multiple targets (head and efficiency) of the optimal prediction meta-model for target optimization, effectively avoiding local optimal solutions, finding the global optimal combination of impeller geometric parameters, and balancing the relationship between the performance indicators of different targets to obtain an impeller model with the optimal comprehensive performance. Description of the Drawings

[0035] Figure 1 It is the flow chart of the optimization method for the impeller of the low-temperature deep-well pump based on the PSO algorithm;

[0036] Figure 2 It is the distribution diagram of the CoP value between the input parameters and the response variable (efficiency);

[0037] Figure 3 It is the distribution diagram of the CoP value between the input parameters and the response variable (head);

[0038] Figure 4 It is the flow chart of the particle swarm optimization algorithm;

[0039] Figure 5 It is the distribution diagram of the Pareto solution set and the Pareto front solution set. Detailed Implementation Manner

[0040] Referring to Figure 1 This application embodiment provides an optimization method for the impeller of a low-temperature deep-well pump based on the PSO algorithm, including the following steps:

[0041] S1. Parametrically model the impeller of the low-temperature deep-well pump;

[0042] S2. Combine the parametrically modeled impeller with other flow components and perform numerical simulation calculations to obtain the hydraulic performance of the low-temperature deep-well pump model; the hydraulic performance includes head and efficiency;

[0043] S3. Compare the hydraulic performance of the low-temperature deep-well pump model with the specified hydraulic performance to determine whether the low-temperature deep-well pump model needs to be optimized; if optimization is required, execute step S4;

[0044] S4. Take the geometric design parameters of the impeller as input parameters and the hydraulic performance of the low-temperature deep-well pump model as output parameters to conduct a sensitivity analysis. Then, construct a multi-objective optimization MOP surrogate model based on the sensitivity analysis results. Further, optimize the multi-objective optimization MOP surrogate model based on the particle swarm optimization algorithm to obtain an optimized impeller model.

[0045] Among them, step S1 includes: In modeling software (such as SpaceClaim or ANSYS), first define the basic geometric parameters of the impeller, such as the impeller inlet diameter, impeller outlet diameter, impeller outlet width, blade inlet setting angle, blade outlet setting angle, blade wrap angle, blade thickness, number of blades, and blade inlet edge position. According to the defined parameters, construct the basic geometric model of the impeller. For example, create basic structures such as the hub and blades. Through parametric tools, associate the geometric parameters with the model so that the model can be automatically updated when the parameters are modified. For example, when changing the number of blades or the blade angle, the model will automatically adjust the distribution and shape of the blades. Use CFD (Computational Fluid Dynamics) technology to simulate the performance of the impeller under different parameter combinations, and analyze performance indicators such as efficiency and head. Determine the optimal parameter combination through methods such as orthogonal experiments.

[0046] Through parametric modeling, parameters such as the impeller inlet diameter, outlet diameter, blade angle, and number of blades can be flexibly adjusted. After combining the parameterized impeller with other flow-through components, a complete pump internal flow passage model is formed, providing a basis for subsequent numerical simulations. Numerical simulation is an important means to evaluate and optimize the hydraulic performance of pumps, and usually uses Computational Fluid Dynamics (CFD) technology.

[0047] For example, in the embodiment of the present application, transfer the parameterized impeller model to TurboGrid software for high-quality structured mesh generation. The other flow-through components of the low-temperature deep-well pump are meshed through ICEM CFD software. Finally, combine the mesh model of the impeller and the mesh models of each flow-through component for numerical simulation calculations, and obtain the hydraulic performance of the low-temperature deep-well pump through numerical simulation calculations.

[0048] Among them, the specified hydraulic performance in step S3 refers to the target performance indicators set by the user according to the usage scenario, operating conditions requirements, and design standards of the low-temperature deep-well pump. The specified hydraulic performance includes: efficiency (η) and design head (H). Efficiency refers to the energy conversion efficiency of the pump, usually requiring a certain minimum standard; the design head refers to the height that the pump needs to lift the liquid.

[0049] For example, by analyzing the efficiency and head of the hydraulic performance, its efficiency is 48% and the head is 71m, which does not meet the specified hydraulic performance requirements of an efficiency of 50% and a head of 74m. Then, through subsequent optimization, the specified requirements are met.

[0050] Step S4 includes:

[0051] S41. Define the geometric design parameters of the impeller as input parameters, define the hydraulic performance of the low-temperature deep-well pump model as output parameters, take the maximum head and the highest efficiency as multi-objective functions, and perform sensitivity analysis using the global sensitivity analysis method;

[0052] S42. Construct a sample library for the input parameters according to the sensitivity analysis results;

[0053] S43. Construct a multi-objective optimization response surface model based on the sample library;

[0054] S44. According to the multi-objective optimization response surface model, screen out the target input parameters that have the greatest influence on the response variables;

[0055] S45. Based on these target input parameters, construct a multi-objective optimization MOP surrogate model and perform multi-objective global optimization using the particle swarm optimization algorithm;

[0056] S46. Find the optimal solution in the Pareto solution set obtained by multi-objective global optimization to obtain the optimized impeller model;

[0057] S47. Select the remaining input parameters other than the target input parameters, re-parametrize and model the optimized impeller model, still taking the highest efficiency and the maximum head as the optimization objectives of the multi-objective function, and repeat the optimization process of S41 - S46 to obtain the finally optimized impeller model.

[0058] The global sensitivity analysis method can evaluate the influence degree of each input parameter on the output parameter (i.e., hydraulic performance). Through global sensitivity analysis, the geometric design parameters that have the greatest influence on the head and efficiency can be identified, and these parameters are the key variables in the optimization process.

[0059] In the sensitivity analysis, a suitable change interval of the input parameters should be selected for the input parameters. Determine the value range of the target input parameters according to the pump design experience integrated in the OptiSLang optimization module and relevant research results. In this embodiment, the target input parameters to be optimized and the parameter change ranges are shown in Table 1.

[0060] Table 1

[0061]

[0062] Among them, step S42 includes:

[0063] S421. According to the sensitivity analysis results, select a suitable change range of the input parameters for the input parameters;

[0064] S422. Define the optimization objectives of taking the highest efficiency and the maximum head as multi-objective functions;

[0065] S423. Randomly generate a large number of sample points within the range of input parameter changes based on the Monte Carlo sampling method, and then construct a basic sample database;

[0066] S424. Through experimental design, specify the efficiency and head of the cryogenic deep well pump as response variables, and use the advanced Latin hypercube sampling method to sample multiple sample points in the basic sample database, and then construct a sample library.

[0067] The mathematical expression of multi-objective optimization can be defined as:

[0068] min x F(x) = [f 1 (x), f 2 (x), …, f K (x)]

[0069] g i (x) ≥ 0, i ∈ [1, M]

[0070] h j (x) = 0, j ∈ [1, L]

[0071] Let D be the feasible region of the above multi-objective optimization problem, that is:

[0072] D = {x|g i (x) ≥ 0, i ∈ [1, M], h j (x) = 0, j ∈ [1, L]}

[0073] In S423, the constructed basic sample database includes multiple sample values of the target input parameters and the output parameter values corresponding to the multiple sample values. The output parameter values are also numerical simulation values obtained in a similar way using the above logic.

[0074] In S424, the constructed sample library also includes multiple sample values of the target input parameters and the output parameter values corresponding to the multiple sample values. In this embodiment, the advanced Latin hypercube sampling method is used to generate 50 sample points for the target input parameters, as shown in Table 2.

[0075] Table 2

[0076]

[0077]

[0078]

[0079] In this embodiment, step S45 includes:

[0080] S451. Construct a high-quality response surface fitting model based on these subsets of target input parameters;

[0081] S452. Use the response surface fitting model to construct a response surface for each response variable;

[0082] S453. Integrate all the response surfaces into a multi-objective optimization MOP surrogate model;

[0083] S454. Employ a pre-trained particle swarm optimization algorithm to perform multi-objective global optimization on the multi-objective optimization MOP surrogate model.

[0084] By constructing a multi-objective optimization response surface model, an approximate value of the output parameters can be found, enabling designers to directly obtain the numerical values of the output parameters in the design space of a large number of sample points.

[0085] The CoP (Correlation of Parameters) value represents the strength of the correlation between the input parameters and the output parameters.

[0086] Through the MOP response surface model, calculate the CoP value through sensitivity analysis, and select the input parameter with the largest CoP value as the target input parameter. Then, construct a response surface fitting model based on these target input parameters. Use the response surface fitting model to construct high-quality response surfaces for each response variable, and a multi-objective optimization MOP surrogate model can be obtained. During the optimization process, define this multi-objective optimization MOP surrogate model as the calculation and solution tool, and by analyzing the magnitude of the CoP values of each input parameter and the response variable, the degree of influence of the input parameter on the response variable can be obtained. As Figure 2 and Figure 3 shown is the relationship between the CoP values of the input parameters and the response variable.

[0087] For example, based on the multi-objective optimization response surface model, it is found that the impeller inlet diameter, impeller outlet diameter, impeller outlet width, blade inlet setting angle, blade outlet setting angle, and blade wrap angle have a great influence on the response variable, and a particle swarm optimization algorithm is used to perform global optimization. First, set the velocity and position parameters of the particles for the particle swarm optimization algorithm, as shown in Table 3.

[0088] Table 3

[0089]

[0090]

[0091] Determine the best initial fitness value of the particles, modify the parameters of the particle velocity and position using the prepared formula, and check and update the fitness value of each particle. For the improved particles, update their fitness value to obtain the optimal solution, as Figure 3The flowchart of the particle swarm optimization algorithm is shown. Finally, the optimal fitness value is found among all the particles, and this optimal fitness value is the optimal solution. The particles in the particle swarm optimization algorithm have mathematical expressions for regular movement and adjustment of their positions, such as the formula:

[0092] v i = ω × v i + c 1 × rand() × (pbest i - x i ) + c 2 × rand()

[0093] × (gbest - x i )

[0094] x i = x i + v i

[0095] In the formula, i = 1, 2, …, N, where N is the total number of particles; ω is the inertia factor with a value of 0.9; v i is the particle velocity; rand() is a random number between 0 and 1; x i is the current particle position; c 1 , c 2 are the learning factors with a value of 2.

[0096] After training the particles in the particle swarm algorithm, the multi - objective optimization MOP surrogate model can be multi - object optimized by the trained particle swarm optimization algorithm. The first optimization objective is to maximize the efficiency of the low - temperature deep - well pump, and the second optimization objective is to maximize the head of the low - temperature deep - well pump. In this embodiment, step S46 includes:

[0097] S461. Find the optimal solution that meets the optimization objectives in the distribution of the Pareto - front solution set obtained from the multi - objective global optimization to obtain the optimized impeller model.

[0098] The Pareto solution set is obtained through multi - objective optimization, and the distribution of the Pareto - front solution set is statistically analyzed. As Figure 4 shown in the diagram of the Pareto solution set and the Pareto - front solution set distribution, find the solution set that meets the specified hydraulic performance requirements on the front - solution set distribution. The parameters of the solution set are: the impeller outlet diameter is optimized to 195 mm, the impeller inlet diameter is optimized to 80 mm, the impeller outlet width is optimized to 80 mm, the blade inlet setting angle is optimized to 11°, the blade outlet setting angle is optimized to 20°, and the blade wrap angle is optimized to 190°. The geometric parameters of this solution set are the optimized result scheme, and the impeller optimization model is determined.

[0099] By re-parametrically modeling the optimized impeller model, the input parameters are selected as blade thickness, number of blades, and blade inlet edge position. The multi-objective function defined by the output parameters of the low-temperature deep-well pump is the highest efficiency and the largest head. Using the global sensitivity analysis method and the particle swarm optimization algorithm, the same settings and processes as above are carried out for optimization. The final optimized impeller model is obtained from the Pareto front solution set distribution obtained. The final geometric design parameters of the impeller are shown in Table 4. By comparing the hydraulic performance of the low-temperature deep-well pump before and after optimization, the efficiency after optimization is 57.8%, an increase of 9.8%; the head is 76 m, an increase of 5 m.

[0100] Table 4

[0101]

[0102] Obviously, the above embodiments are the preferred embodiments of the present invention to illustrate the technical solutions of the present invention, rather than limiting it. Of course, it does not limit the patent scope of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cryogenic deep well pump impeller optimization method based on PSO algorithm, characterized in that: The following steps are involved: S1. Parametric modeling of the impeller of the cryogenic deep well pump; S2. Combine the impeller and other flow-through components after parameterized modeling, and perform numerical simulation calculations to obtain the hydraulic performance of the low-temperature deep well pump model; the hydraulic performance includes head and efficiency; S3, comparing the hydraulic performance of the low-temperature deep well pump model with the specified hydraulic performance to determine whether the low-temperature deep well pump model needs to be optimized; If optimization is required, execute step S4; S4. Taking the geometric design parameters of the impeller as input parameters and the hydraulic performance of the low-temperature deep well pump model as output parameters, a sensitivity analysis is performed, and a multi-objective optimization MOP proxy model is constructed based on the sensitivity analysis results. The multi-objective optimization MOP proxy model is then optimized based on the particle swarm optimization algorithm to obtain the optimized impeller model.

2. The cryogenic deep well pump impeller optimization method based on particle swarm optimization algorithm according to claim 1, characterized in that: Step S4 includes: S41. Define the geometric design parameters of the impeller as input parameters, define the hydraulic performance of the low-temperature deep well pump model as output parameters, take maximum head and highest efficiency as multi-objective functions, and use the global sensitivity analysis method to perform sensitivity analysis; S42. Construct a sample library for input parameters based on the sensitivity analysis results; S43, constructing a multi-objective optimization response surface model based on the sample library; S44, screening out the target input parameters that have the greatest impact on the response variables according to the multi-objective optimization response surface model; S45, based on these target input parameters, construct a multi-objective optimization MOP agent model, and use a particle swarm optimization algorithm to perform multi-objective global optimization; S46, finding the optimal solution in the Pareto solution set obtained by multi-objective global optimization, and obtaining an optimized impeller model; S47, selecting the remaining input parameters except the target input parameters, re-parameterizing the optimized impeller model, still taking the highest efficiency and the maximum head as the optimization goals of the multi-objective function, repeating the S41-S46 optimization process to obtain the final optimized impeller model.

3. The low-temperature deep well pump impeller optimization method based on particle swarm optimization algorithm according to claim 2 is characterized in that: Step S42 includes: S421. Selecting a suitable input parameter variation range for the input parameter according to the sensitivity analysis result; S422, defining maximum efficiency and maximum lift as optimization goals of the multi-objective function; S423, based on the Monte Carlo sampling method, randomly generate a large number of sample points within the range of input parameter variation, and then build a basic sample database; S424. Through experimental design, the efficiency and head of the cryogenic deep well pump are specified as response variables, and the advanced Latin hypercube sampling method is used to sample multiple sample points in the basic sample database to construct a sample library.

4. The low-temperature deep well pump impeller optimization method based on the PSO algorithm according to claim 3 is characterized in that: The geometric design parameters of the impeller include: the impeller inlet diameter, impeller outlet diameter, impeller outlet width, blade inlet placement angle, blade outlet placement angle, blade wrap angle, blade thickness, blade number and blade inlet edge position of the cryogenic deep well pump.

5. The low-temperature deep well pump impeller optimization method based on PSO algorithm according to claim 2 is characterized in that: S45 includes: S451, constructing a high-quality response surface fitting model based on these target input parameter subsets; S452, construct a response surface for each response variable using the response surface fitting model; S453, integrate all response surfaces into a multi-objective optimization MOP proxy model; S454. Use the pre-trained particle swarm optimization algorithm to perform multi-objective global optimization on the multi-objective optimization MOP agent model.

6. The cryogenic deep well pump impeller optimization method based on PSO algorithm according to claim 2, characterized in that: Step S46 includes: S461. Find the optimal solution that meets the optimization objectives in the Pareto front solution set distribution obtained from the multi-objective global optimization, and obtain the optimized impeller model.

Images