Method for improving cavitation performance of centrifugal pump

Abstract

The invention discloses a method for improving cavitation performance of a centrifugal pump, which comprises the following steps of: carrying out parametric design on an impeller structure of the centrifugal pump, screening a plurality of parameters which obviously influence the net positive suction head from numerous impeller structure parameters, designing a plurality of test samples based on the parameters, carrying out calculation, training the sample parameters by using an RBF (Radial Basis Function) neural network, and calculating the cavitation performance of the centrifugal pump. And then optimizing the trained RBF neural network model by using a genetic algorithm so as to obtain an optimal structure parameter combination of the cavitation performance of the centrifugal pump. And designing and manufacturing the impeller according to the obtained optimal structural parameters, and mounting the impeller on the centrifugal pump, so that the cavitation performance of the centrifugal pump is improved. By optimizing and improving the related structure variables and parameters influencing the cavitation performance, the cavitation resistance of the centrifugal pump can be greatly improved.

Description

Technical Field

[0001] The invention relates to the improvement of a centrifugal pump, in particular to a method for improving the cavitation performance of a centrifugal pump, and belongs to the technical field of centrifugal pumps. Background Art

[0002] As one of the most commonly used pumps for conveying liquids, centrifugal pumps are suitable for various occasions and have a wide range of applications. They also play an irreplaceable role in some special fields. With the rapid development of centrifugal pump technology at home and abroad, users' performance requirements for centrifugal pumps are also constantly increasing. In addition to requirements for head, efficiency, etc. that directly affect the performance of centrifugal pumps, the requirements for the anti-cavitation performance of centrifugal pumps are also gradually increasing. In the design of centrifugal pumps, the shape and size design of the impeller has a direct impact on the performance of the centrifugal pump. An unreasonable design will make the centrifugal pump more prone to cavitation, which will directly affect its service life.

[0003] Therefore, it is very important to ensure the working conditions of centrifugal pumps, improve the working efficiency of centrifugal pumps and avoid the occurrence of cavitation in the design and use of centrifugal pumps. In the design process, some designers often pursue higher theoretical efficiency and ignore the impact of cavitation on efficiency, which makes the designed products unable to meet the use requirements, causing waste of resources for manufacturers and losing product industry competitiveness. Therefore, in the process of centrifugal pump design, it is particularly important to improve the cavitation performance. Under the premise of ensuring that the head and efficiency meet the requirements, the anti-cavitation performance of the centrifugal pump should be improved as much as possible. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a method for improving the cavitation performance of a centrifugal pump. The present invention can greatly improve the anti-cavitation performance of a centrifugal pump by optimizing and improving the relevant structural variables and parameters that affect the cavitation performance.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions: A method for improving the cavitation performance of a centrifugal pump, the steps are as follows: 1) Take the centrifugal pump impeller as the starting point for improving the cavitation performance, and select several major influencing variables that significantly affect the NPSH from the impeller structure variables; 2) Based on the main influencing variables, multiple training samples are obtained. Each training sample consists of the specific parameters of each main influencing variable and the corresponding NPSH. The RBF neural network is trained through the training samples. During training, the specific parameters of the main influencing variables are used as model input, and the NPSH is used as model output. The mapping relationship between the main influencing variables and the NPSH is established to obtain the RBF neural network model. 3) Use genetic algorithm to optimize the trained RBF neural network model to obtain the main influencing variable parameters with the best cavitation performance of the centrifugal pump, which are the main influencing variable optimization parameters; 4) Based on the optimized parameters of the main influencing variables obtained in step 3), the impeller is designed and manufactured and installed on the centrifugal pump, that is, the cavitation performance of the centrifugal pump is improved.

[0006] Furthermore, the impeller structure variables include: blade inlet placement angle β 1 , Blade outlet placement angle β 2 , blade wrap angle φ , Impeller outlet width b 2 , Impeller inlet diameter D 1 , Impeller outlet diameter D 2 and number of blades Z .

[0007] Furthermore, there are four main influencing variables, namely, blade inlet placement angle β 1 , blade wrap angle φ , Impeller outlet width b 2 and impeller inlet diameter D 1 .

[0008] Furthermore, in step 1), a significance analysis test is used to screen several main influencing variables that significantly affect the NPSH from the impeller structural variables.

[0009] Furthermore, in step 2) when obtaining training samples, the specific parameters of each main influencing variable are taken between the upper limit and the lower limit of the main influencing variable.

[0010] Furthermore, in step 2), when the RBF neural network is trained, the mean square error is used as the evaluation of the training fitting effect. If the mean square error is less than 0.5, it is considered that the trained RBF neural network model meets the use requirements. The training is completed and the RBF neural network model is output.

[0011] Furthermore, the genetic algorithm described in step 3) is a NGSA-III genetic algorithm.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention utilizes PB significance analysis test design to screen the impeller parameters that significantly affect the cavitation margin, which can more clearly identify the parameters that are sensitive to cavitation performance, thereby performing design.

[0013] 2. The present invention uses RBF neural network to train sample data, and can establish the connection between parameters and objective function with fewer sample training results, obtain a fitting curve with better fitting accuracy, and thus establish a corresponding neural network model.

[0014] 3. The present invention uses NSGA-Ⅲ genetic algorithm to optimize the neural network model within a specified range. Combined with the actual requirements of the target parameters, the optimal parameter combination can be obtained. The impeller designed based on the optimal parameters can significantly improve the anti-cavitation performance of the centrifugal pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which: Figure 1 -Flowchart of the present invention for improving the cavitation performance of a centrifugal pump.

[0016] Figure 2 - Flowchart of neural network training of the present invention.

[0017] Figure 3 -Genetic algorithm optimization flow chart of the present invention.

[0018] Figure 4 -Comparison diagram of the centrifugal pump impeller of the present invention before and after optimization. DETAILED DESCRIPTION

[0019] In order to reduce the NPSH of a centrifugal pump, improve the anti-cavitation performance of the centrifugal pump, and thus improve the efficiency and service life of the centrifugal pump, the present invention performs parameterized design on the impeller structure of the centrifugal pump, selects several parameters that significantly affect the NPSH from numerous impeller structural parameters, designs multiple test samples based on the parameters, trains the sample parameters using an RBF neural network, and then uses the NGSA-Ⅲ genetic algorithm to optimize the trained RBF neural network model, thereby obtaining the optimal structural parameter combination for the cavitation performance of the centrifugal pump.

[0020] The embodiments of the present invention are described in further detail below.

[0021] The present invention provides a method for improving the cavitation performance of a centrifugal pump. The specific steps are as follows. Figure 1 , 1) Through the study of the cavitation principle of centrifugal pumps, it is found that the impeller structure has a significant impact on cavitation. Therefore, the centrifugal pump impeller is used as a starting point for improving the cavitation performance, and several major influencing variables that significantly affect the cavitation margin are selected from the impeller structure variables; 2) Based on the main influencing variables, multiple training samples are obtained. Each training sample consists of the specific parameters of each main influencing variable and the corresponding NPSH. The RBF neural network is trained through the training samples. During training, the specific parameters of the main influencing variables are used as model input, and the NPSH is used as model output. The mapping relationship between the main influencing variables and the NPSH is established to obtain the RBF neural network model. 3) Use genetic algorithm to optimize the trained RBF neural network model to obtain the main influencing variable parameters with the best cavitation performance of the centrifugal pump, which are the main influencing variable optimization parameters; 4) Based on the optimized parameters of the main influencing variables obtained in step 3), the impeller is designed and manufactured and installed on the centrifugal pump, that is, the cavitation performance of the centrifugal pump is improved.

[0022] For the impeller, the structural variables include the following seven: blade inlet placement angle β 1 , Blade outlet placement angle β 2 , blade wrap angle φ , Impeller outlet width b 2 , Impeller inlet diameter D 1 , Impeller outlet diameter D 2 and number of blades Z The present invention utilizes a significance analysis test to screen out several main influencing variables that significantly affect the NPSH from the seven impeller structural variables.

[0023] In the embodiment, the PB significance experiment of the present invention is carried out according to the following design scheme, and each impeller structure variable takes two values, namely, an upper limit value and a lower limit value. Then, the upper and lower limit values ​​of the seven variables are respectively input into the PB significance experiment software, and 12 of the experimental arrays obtained are selected as the specific data sets of the significance test, as shown in Table 1.

[0024] Table 1 PB significance experimental design The T value and P value of each variable are obtained through software calculation, among which the blade inlet placement angle β 1 , Blade outlet placement angle β 2 , blade wrap angle φ , Impeller outlet width b 2 , Impeller inlet diameter D 1 , Impeller outlet diameter D 2 and number of bladesZ The P values ​​are 0.015, 0.081, 0.027, 0.034, 0.068, 0.473 and 0.124 respectively. Therefore, the importance of the seven variables is ranked as follows: blade inlet angle β 1 , blade wrap angle φ , Impeller outlet width b 2 , Impeller inlet diameter D 1 , Blade outlet placement angle β 2 Number of blades Z and impeller outlet diameter D 2 The specific PB experimental calculation results are shown in Table 2.

[0025] Table 2 PB experimental calculation results In the specific implementation, four main influencing variables are selected according to the obtained importance, namely, blade inlet placement angle β 1 , blade wrap angle φ , Impeller outlet width b 2 and impeller inlet diameter D 1 Of course, the number of main influencing variables is not fixed and can be adjusted according to actual conditions.

[0026] Step 2) When obtaining training samples, the specific parameters of each main influencing variable are taken between the upper limit and the lower limit of the main influencing variable.

[0027] Step 2) When training the RBF neural network, the mean square error is used as the evaluation of the training fitting effect. If the mean square error is less than 0.5, it is considered that the trained RBF neural network model meets the use requirements. The training is completed and the RBF neural network model is output. Figure 2 It is a flowchart of the neural network training of the present invention. As can be seen from the figure, the embodiment of the present invention divides the data set into a training set and a test set, the training set is 50 groups of sample data, and the test set is 10 groups of sample data.

[0028] Since the neural network model of the present invention has only one output parameter, that is, only one parameter, NPSH, needs to be optimized, there is no special requirement for genetic algorithm optimization, and most algorithms can meet the use scenario of the present invention. In the embodiment, the genetic algorithm used in step 3) is NGSA-III genetic algorithm. Figure 3 This is the optimization flow chart of the genetic algorithm of the present invention.

[0029] Figure 4This is a comparison diagram of the centrifugal pump impeller of the present invention before and after optimization according to the above embodiment. Figure 4 It can be seen that the cavitation margin of the centrifugal pump is lower after optimization than before optimization. At the same cavitation margin value, the optimized centrifugal pump has fewer bubbles and does not spread over a large range, indicating that the cavitation phenomenon of the centrifugal pump is delayed after optimization, and its anti-cavitation performance is stronger.

[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described with reference to the preferred embodiments of the present invention, it should be understood by those skilled in the art that various changes may be made in form and details without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for improving the cavitation performance of a centrifugal pump, characterized in that: The steps are as follows, 1) Take the centrifugal pump impeller as the starting point for improving the cavitation performance, and select several major influencing variables that significantly affect the NPSH from the impeller structure variables; 2) Based on the main influencing variables, multiple training samples are obtained. Each training sample consists of the specific parameters of each main influencing variable and the corresponding NPSH. The RBF neural network is trained through the training samples. During training, the specific parameters of the main influencing variables are used as model input, and the NPSH is used as model output. The mapping relationship between the main influencing variables and the NPSH is established to obtain the RBF neural network model. 3) Use genetic algorithm to optimize the trained RBF neural network model to obtain the main influencing variable parameters with the best cavitation performance of the centrifugal pump, which are the main influencing variable optimization parameters; 4) Based on the optimized parameters of the main influencing variables obtained in step 3), the impeller is designed and manufactured and installed on the centrifugal pump, that is, the cavitation performance of the centrifugal pump is improved.

2. A method for improving the cavitation performance of a centrifugal pump according to claim 1, characterized in that: The impeller structure variables include: blade inlet placement angle β 1. Blade outlet placement angle β 2. Blade wrap angle φ , Impeller outlet width b 2. Impeller inlet diameter D 1. Impeller outlet diameter D 2 and number of blades Z .

3. A method for improving the cavitation performance of a centrifugal pump according to claim 2, characterized in that: There are four main influencing variables, namely, blade inlet placement angle β 1. Blade wrap angle φ , Impeller outlet width b 2 and impeller inlet diameter D 1.

4. A method for improving the cavitation performance of a centrifugal pump according to claim 1, characterized in that: In step 1), a significance analysis test is used to screen several main influencing variables that significantly affect the NPSH from the impeller structural variables.

5. The method for improving the cavitation performance of a centrifugal pump according to claim 1, characterized in that: Step 2) When obtaining training samples, the specific parameters of each main influencing variable are taken between the upper limit and the lower limit of the main influencing variable.

6. A method for improving the cavitation performance of a centrifugal pump according to claim 1, characterized in that: Step 2) When training the RBF neural network, the mean square error is used as the evaluation of the training fitting effect. If the mean square error is less than 0.5, it is considered that the trained RBF neural network model meets the use requirements. The training is completed and the RBF neural network model is output.

7. A method for improving the cavitation performance of a centrifugal pump according to claim 1, characterized in that: The genetic algorithm described in step 3) is the NGSA-III genetic algorithm.

Images