Multi-objective optimization design method for integrated turbine box

Abstract

The invention provides an integrated turbine case multi-objective optimization design method which is applied to multi-objective structure optimization design of an integrated turbine case of a high-power-to-weight-ratio engine through combination of a response surface model, a neural network model, heat-fluid-solid coupling simulation and a genetic algorithm. The method mainly comprises the following steps: establishing a geometric structure parameter model of the integrated turbine box, and determining an optimized geometric structure parameter and an optimization target; and determining a structure parameter optimization range, and sampling in a sample space to form a sample library. And performing thermal-fluid-solid coupling simulation calculation on the sample points in the sample library, and establishing a mapping relation between geometric structure parameters and an optimization target by using a response surface approximation function and a BP neural network. And performing multi-target optimization through a genetic algorithm, and screening out a group of optimal structure parameters. The weight of the integrated turbine box is reduced, meanwhile, the structural strength is guaranteed, and the optimization effect is obvious.

Description

Technical Field

[0001] The present invention belongs to the technical field of turbine case structure optimization, and in particular relates to a multi-objective optimization design method for an integrated turbine case. Background Art

[0002] The turbocharger operates as follows: exhaust gas from the engine is directed through the exhaust pipe into the turbine housing, where it rotates and generates power. The turbine then drives the compressor, delivering the supercharged air into the cylinders. With the continuous improvement of engine power-to-weight ratios, lightweight turbo housings are required. The turbocharger turbine end has been upgraded from a traditional flange connection to an integrated exhaust pipe and volute structure. However, under severe thermal loads, the transition from the exhaust pipe to the volute in this integrated turbo housing is prone to stress concentration, which can lead to failure and directly impacts turbocharger reliability. Current approaches to structural optimization of integrated turbo housings mostly rely on simple parametric analysis and comparison results, which fail to simultaneously consider the impact of multiple structural parameters on turbo housing performance. Therefore, considering the requirements for lightweight turbochargers and reliability under high temperatures, and taking into account the unique characteristics of the integrated turbo housing structure, it is necessary to integrate competing or conflicting key objective functions using a multi-objective genetic algorithm, extract key structural design parameters, and utilize scientific methods to develop an accurate and efficient quantitative optimization technique for the structural optimization of the integrated turbo housing.

[0003] Currently, most scholars use numerical simulation methods to optimize the structure of the parts of the turbine box that are prone to failure, and reduce the thermal stress of the turbine box by increasing the fillet, increasing the wall thickness, etc. This method can only consider the impact of a single factor on the structure and performance of the turbine box at a time, and cannot consider the comprehensive impact of multiple structural parameters on the performance of the turbine box. Another method is to prepare turbine boxes with a variety of structural parameters and conduct experimental tests, but the experimental method has a long cycle, poor economy, and low design efficiency. Therefore, in order to ensure the reliability and design accuracy of the integrated turbine box structure, it is very necessary to develop an optimization design method. Only through a reliable multi-objective optimization method can the optimal structural parameters be obtained to maximize the performance of the integrated turbine box. Summary of the Invention

[0004] In view of this, the present invention aims to propose a multi-objective optimization design method for an integrated turbine box to simultaneously meet the requirements of lightweight and reliability of the integrated turbine box, and overcome the problems of long optimization cycle and high cost of the existing integrated turbine box structure.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A multi-objective optimization design method for an integrated turbine case includes the following steps:

[0007] Step 1: Establish a parametric model of the integrated turbine case structure, select the geometric structural parameters to be optimized, including the wall thickness δ and fillet r of the transition from the exhaust pipe to the volute, and determine the optimization targets, including the integrated turbine case weight m and thermal stress σ;

[0008] Step 2: Determine the constraint range of the optimized structural parameters, generate the sample design space, and use the Latin hypercube sampling method to form a sample library;

[0009] Step 3: Using Creo software, the mass of the integrated turbine case under various structural parameters is obtained, and the response surface methodology is used to construct a mapping relationship between the geometric structural parameters and the weight m of the integrated turbine case.

[0010] Step 4: Perform thermal-fluid-solid coupling numerical simulation analysis based on the sample points in the sample library. By calling the prepared macro command files of Creo and ANSYS CFX software, parametric modeling, meshing, pre-processing, simulation calculation and post-processing are completed to obtain the thermal stress σ corresponding to the sample space;

[0011] Step 5: Construct a BP neural network model based on the optimization target value obtained in step 3 to obtain the mapping relationship between geometric structure parameters and thermal stress σ;

[0012] Step 6: Based on the mapping model of steps 4 and 5, the NSGA-Ⅱ genetic algorithm is used to obtain the Pareto frontier solution of the multi-objective optimization of the integrated turbine box, and the optimal solution of the structural parameters is determined according to the strength criterion.

[0013] Furthermore, in step 3, the density of the turbine case material is set in Creo software, and the weight m of the turbine case is calculated by the software. The response surface approximation function is fitted using a second-order polynomial. The input variables are the wall thickness δ and the fillet r, and the output variable is the weight m. The function is expressed as follows:

[0014]

[0015] In the formula, a0 is the constant term to be determined, a j is the coefficient of the second term to be determined, a ij is the undetermined coefficient of the quadratic term, k is the number of design variables x i 、x j is the input variable.

[0016] Furthermore, in step 4, the thermal-fluid-solid coupling numerical simulation steps are as follows: the RNG k-ε model is selected, and the boundary conditions are as follows: the turbine box inlet adopts a pressure inlet, the inlet total pressure is 334 kPa, the inlet temperature is 1023 K, the outlet is a mass flow outlet, which is 0.19 kg / s, the impeller area is set as the rotating domain, the speed is 140000 r / min, and the rest is the stationary domain. The inner wall of the turbine box is set to be impermeable, non-slip, and adiabatic, and the convergence accuracy is set to 10 -5 The number of iteration steps is 800. CFX is used to perform steady-state solution to obtain the temperature and convection heat transfer coefficient of the fluid domain on the inner wall of the turbine box. These are then imported into the structural field and the turbine box thermal stress σ is obtained through simulation calculation.

[0017] Furthermore, in step five, the BP neural network includes an input layer, a hidden layer, and an output layer. The input variables are the wall thickness δ and the fillet r, and the output variable is the thermal stress σ. The BP neural network construction includes the following steps:

[0018] A1. The constructed neural network model has 2 input layer nodes, 8 hidden layer nodes, and 1 output layer node. The input is wall thickness δ and fillet r, and the output is thermal stress σ.

[0019] A2. BP neural network uses Trainlm training method, and the Tan-sigmoid transfer function is used from the input layer to the hidden layer. The transfer function between the hidden layer and the output layer is the Purelin transfer function Ψ(x);

[0020] 75% of the sample space is used as the neural network training set, and 25% is used as the test set. The structural parameters wall thickness δ and fillet r are input vectors, expressed as X = [x1, x2] T , the optimization target thermal stress σ is the output vector, expressed as Y = [y1];

[0021] A3, BP neural network training times is 1000, learning rate is 0.001, BP neural network is trained, and the mean square error loss function is used to represent the neural network prediction value y k p and target value y k The error between the two layers is updated by gradient descent method, and the weights and thresholds of the hidden layer and output layer are updated until the mean square error is less than 0.5, and the training is stopped.

[0022] A4. Use the trained neural network to predict the test set. If the mean square error is greater than 0.5, return to retrain. If it is less than 0.5, output a qualified BP neural network approximate model.

[0023] Furthermore, in step A2, the functional relationship between the input vector and the output vector is:

[0024] Y=y(X)=Ψ[W ki Φ(W ij X+A)+B] (2)

[0025]

[0026] Where W ki is the weight matrix between the hidden layer and the output layer, B is the threshold matrix of the output layer, W ij is the weight matrix between the input layer and the hidden layer, and A is the threshold matrix of the hidden layer.

[0027] Furthermore, in step A3, the neural network predicts the value y k p and target value y k The error between them is expressed as:

[0028] Furthermore, in step six, the mathematical description of the multi-objective optimization of the integrated turbine case is as follows:

[0029]

[0030] The optimization objective functions include: the integrated turbine box has the smallest weight, the highest structural strength, and the optimized design variables must be within a limited range.

[0031] Furthermore, in step 6, the construction of the NSGA-II genetic algorithm includes the following steps:

[0032] B1. Determine the initial model input parameters: wall thickness δ and fillet r, and determine the constraint range of the parameters;

[0033] B2. Determine the objective functions f1 and f2;

[0034] B3. According to the actual parameter variables of the genetic algorithm model, the genetic algorithm initialization parameters are set: the population size is 50, the number of optimization targets is 2, the crossover method is two-by-two crossover according to probability, the number of crossover individuals is 36, the crossover probability is 0.7, the number of mutation individuals is 14, and the mutation probability is 0.2;

[0035] B4. Randomly generate an initialized population, calculate the objective function value corresponding to the initial solution, perform non-dominated sorting on the initial population, obtain a set of non-dominated solutions, and calculate the congestion degree of the initial population;

[0036] B5. Select the initialized population as the parent population, generate the offspring population through genetic operations, merge the parent population and offspring population into a candidate population, perform non-dominated sorting and crowding calculation, and select the top 50 individuals with higher non-dominated levels and higher crowding as the new parent population for genetic operations, and so on;

[0037] B6. Determine whether the convergence conditions are met. If the maximum number of iterations is reached or convergence is to a satisfactory Pareto frontier solution, generate an optimal solution for a set of structural parameters. If not, regenerate a new parent population and iterate again until the maximum number of iterations is reached. Output the optimal solution for the next set of structural parameters after the maximum number of iterations, and terminate the iteration.

[0038] B7. According to the strength criterion σ≤[σ] / S, select the optimal solution from a series of Pareto front solutions.

[0039] Compared with the prior art, the integrated turbine case multi-objective optimization design method described in the present invention has the following advantages:

[0040] (1) The multi-objective optimization design method for an integrated turbine box described in the present invention takes into account both the lightweight and reliability of the integrated turbine box, optimizes the fillet and wall thickness of the transition part, thereby effectively realizing the structural design optimization of the integrated turbine box, reducing the weight of the integrated turbine box while ensuring the structural strength, and maximizing the performance of the material.

[0041] (2) The present invention describes a multi-objective optimization design method for an integrated turbine box. The present invention introduces the response surface method and the BP neural network to construct an approximate model, and introduces the NSGA-Ⅱ genetic algorithm to perform multi-objective optimization on the structural parameters. It can effectively solve the problems of backward optimization means, long optimization cycle, and high cost in traditional simulation methods or experimental methods, greatly improve the calculation efficiency, and can ensure the accuracy of the calculation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0043] Figure 1 This is a schematic diagram of the integrated turbine case structure according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic flow chart of the multi-objective optimization design method for an integrated turbine case according to an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of the spatial distribution of data samples according to an embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram of the process of constructing a BP neural network model according to an embodiment of the present invention;

[0047] Figure 5 Schematic diagram of the multi-objective optimization process of the NSGA-II genetic algorithm according to an embodiment of the present invention;

[0048] Figure 6 Schematic diagram of optimal solution distribution of geometric parameters of the integrated turbine case according to an embodiment of the present invention;

[0049] Figure 7 Schematic diagram of thermal stress distribution of the integrated turbine case before and after optimization according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0051] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0052] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0053] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0054] like Figures 1 to 7As shown, a multi-objective optimization design method for an integrated turbine case is described, comprising the following steps: Step 1: Establishing a parameterized structural model of the integrated turbine case to obtain the optimization objectives and constraints for the optimized structural parameters; Step 2: Determining the constraint range of the optimized structural parameters, generating a sample design space, and sampling within the sample design space to form a sample library; Step 3: Using 3D software to calculate the weight of the integrated turbine case under different structural parameters; Step 4: Using ANSYS CFX software to perform a thermal-fluid-solid coupling simulation of the integrated turbine case to obtain the thermal stress of the integrated turbine case under different structural parameters. The optimized structural parameters in Step 1 include the wall thickness δ and fillet r of the transition from the exhaust pipe to the volute, and the optimization objectives include the weight m and thermal stress σ of the integrated turbine case; the sampling method in Step 2 is Latin hypercube sampling.

[0055] In the above technical solution, in order to ensure that the sample points are evenly distributed in the sample space, the Latin hypercube sampling method is used to generate 40 groups of parameter samples. The distribution of data samples is as follows: Figure 3 shown.

[0056] The response surface method, BP neural network model, and NSGA-Ⅱ genetic algorithm are used to help solve a set of optimal structural parameters to simultaneously meet the requirements of lightweight and reliability of the integrated turbine box. The steps are as follows: Step 1: Construct a response surface approximation function and a neural network model, take the structural parameters in claim 1 as input variables, and the optimization target as the output variable, and obtain the relationship between the geometric structure parameters and the optimization target; Step 2: Based on the approximate model, use the genetic algorithm to obtain the optimal solution of the integrated turbine box geometric structure.

[0057] The integrated turbine case was parametrically modeled using Creo software. By modifying the wall thickness δ and the fillet r, the three-dimensional model of the integrated turbine case with different structural parameters was obtained. The integrated turbine case was made of D5S high-nickel austenitic ductile iron, and the material density was set to 7450 kg / m in Creo software. 3 , and the weight m of the integrated turbine box under each set of structural parameters in the sample library is calculated.

[0058] The response surface methodology was used to construct the functional relationship between the weight m of the integrated turbine case and the structural parameters of wall thickness δ and fillet r.

[0059] The response surface approximation function used is a second-order polynomial, and the function expression is:

[0060]

[0061] Where a0 is the constant term to be determined, a j is the coefficient of the second term to be determined, a ij is the undetermined coefficient of the quadratic term, k is the number of design variables;

[0062] The thermal-fluid-solid simulation method employed was a co-simulation of an integrated turbine case model constructed using Creo and ANSYS CFX. An integrated turbine case model with 40 sets of structural parameters was established in Creo. These 40 sets of integrated turbine case geometric models were pre-processed, meshed, solved, and post-processed in ANSYS to determine the convective heat transfer coefficient on the turbine case's inner wall and the temperature of the fluid domain. The steady-state thermal simulation module was then used to solve the temperature field in the solid domain of the integrated turbine case, thereby calculating the thermal stresses of the integrated turbine case.

[0063] In the above-mentioned integrated turbine box finite element simulation model, due to the complex gas flow and vortex flow in the turbine box, the turbulence model selects the RNG k-ε model. The turbine box uses the high-temperature exhaust gas discharged from the engine cylinder as the working medium, and the fluid model is simplified to a viscous incompressible fluid. The impeller section is the rotating domain, the speed is set to 140000r / min, and the rest is the static domain. The inner wall of the turbine box is set to be impermeable, non-slip, and adiabatic. The turbine box inlet adopts a pressure inlet with a total inlet pressure of 334kPa and an inlet temperature of 1023K. The outlet is a mass flow outlet with a value of 0.19kg / s, and the convergence accuracy is set to 10 -5 The number of iteration steps is 800. CFX is used for steady-state solution to obtain the convective heat transfer coefficient of the inner wall of the turbine box and the temperature of the fluid domain. The results are imported into the steady-state thermal simulation module. The physical properties of the integrated turbine box material are set, including density, specific heat capacity, thermal conductivity, elastic modulus, Poisson's ratio and linear thermal expansion coefficient. The temperature field of the solid domain of the turbine box is obtained by simulation. The inlet flange surface of the turbine box is set as a fixed constraint, and the thermal stress of the turbine box is obtained by simulation.

[0064] The BP algorithm builds an approximate model process as follows Figure 4 As shown. The neural network model used is a BP neural network, and its structure includes an input layer, a hidden layer, and an output layer. The Tan-sigmoid transfer function is used from the input layer to the hidden layer, and the Purelin transfer function is used as the transfer function between the hidden layer and the output layer. The structural parameters wall thickness δ and fillet r are input vectors, and the optimization target thermal stress σ is the output vector. The construction of the neural network model includes the following steps: Step 1: Randomly select 30 of the 40 samples as a training set, and the remaining 10 as a test set; Step 2: The constructed BP neural network model has an input layer node number of 2, a hidden layer node number of 8, an output layer node number of 1, and the weight matrix W between the hidden layer and the output layer of the neural network ki , the threshold matrix B of the output layer, the weight matrix W between the input layer and the hidden layer ij, initialize the threshold matrix A of the hidden layer; Step 3: Train the BP neural network with 1000 training times and a learning rate of 0.001, and update the weights and thresholds of the neural network; Step 4: Calculate the deviation between the predicted value and the target value of the neural network, and stop training when the conditions are met; Step 5: Test the BP neural network with the test set. If the mean square error is greater than 0.5, return to retraining. If it is less than 0.5, output a qualified BP neural network approximate model.

[0065] The mathematical description of the multi-objective optimization of the integrated turbine case is as follows:

[0066]

[0067] The optimization objective functions include: the integrated turbine box has the smallest weight, the highest structural strength, and the optimized design variables must be within a limited range.

[0068] The process of constructing NSGA-Ⅱ genetic algorithm is as follows: Figure 5 As shown, the genetic algorithm used is the NSGA-Ⅱ genetic algorithm, which performs well in dealing with multi-objective optimization problems with its fast non-dominated sorting method, crowding calculation strategy and elite retention mechanism. The construction of the algorithm includes the following steps: Step 1: Determine the initial parameters of the model input: wall thickness δ and fillet r, determine the constraint range of the parameters, and determine the objective function; Step 2: Determine the objective functions f1 and f2; According to the parameter variables of the actual input genetic algorithm model, the genetic algorithm initialization parameters are set: the population size is 50, the number of optimization targets is 2, the crossover method is two individuals crossover by probability, the number of crossover individuals is 36, the crossover probability is 0.7, the number of mutation individuals is 14, and the mutation probability is 0.2; Randomly generate the initialization population and calculate the corresponding initial solution The objective function value is obtained by performing non-dominated sorting on the initial population to obtain a set of non-dominated solutions, and the crowding degree of the initial population is calculated. The initialized population is selected as the parent population, and a series of genetic operations are performed to generate the offspring population. The parent and offspring populations are merged into a candidate population, and non-dominated sorting and crowding degree calculation are performed again. The top 50 individuals with high non-dominated levels and high crowding degrees are selected as the new parent population for genetic operations, and so on. If the maximum number of iterations is reached or convergence reaches a satisfactory Pareto front solution, the final structural parameters are generated. If not, a new parent population is generated and iteration is repeated until the maximum number of iterations is reached. The optimal solution for the structural parameters under the maximum number of iterations is output, and the iteration terminates.

[0069] The criteria for determining the optimal solution of the multi-objective optimization of the integrated turbine case are as follows: a set of Pareto frontier solutions are generated by the NSGA-Ⅱ genetic algorithm, and the reliability of the integrated turbine case is used as a reference for selecting the optimal parameters, that is, σ≤[σ] / S, where [σ] is the fracture strength of the integrated turbine case at the operating temperature, and S is the safety factor. Combined with the integrated turbine case casting process, the optimal structural parameters wall thickness δ and fillet r are determined from a series of non-dominated solutions.

[0070] The following is a specific embodiment of the multi-objective optimization method of the integrated turbine box based on thermal-fluid-solid coupling and genetic algorithm. After 200 generations of NSGA-Ⅱ genetic algorithm optimization, the Pareto frontier solution of the lightweight index and strength of the integrated turbine box is as follows: Figure 6 As shown, it can be obtained from Figure 6 It can be seen from the figures that the optimization design method of the present invention can quantitatively and accurately give the combination of weight and thermal stress of the integrated turbine box under different optimal structures.

[0071] Among the 50 sets of non-dominated optimal solutions, a set of optimal solutions was obtained by screening under the conditions of minimum weight of the integrated turbine box and thermal stress ≤ 336MPa. The integrated turbine box under the optimal geometric structure parameters was modeled and CFD simulation was performed. The simulation results are as follows: Figure 7 The structural strength of the optimized model is compared with that of the initial model, as shown in Table 1. The transition section from the exhaust pipe to the volute of the integrated turbine case experiences the greatest thermal stress. The optimized solution achieves a maximum thermal stress of 328.82 MPa, 1.8% greater than the initial solution, meeting the strength requirements under thermal load. The optimized model weighs 2.95 kg, 8.3% lighter than the initial solution.

[0072] Table 1

[0073] plan Corner fillet r(mm) Wall thickness δ(mm) Weight (kg) Thermal stress (MPa) Initial plan 18 2.7 3.22 321.99 Optimization plan 19.5 2.5 2.95 328.82

[0074] In summary, this application takes the wall thickness δ and the fillet r of the transition part from the exhaust pipe of the integrated turbine box to the volute as optimization design variables, and the weight m and thermal stress σ of the integrated turbine box as optimization targets; adopts the Latin hypercube sampling method to form a sample library, adopts the response surface method to construct the mapping relationship between the geometric structure parameters and the weight m of the integrated turbine box, and adopts the BP neural network model to construct the mapping relationship between the geometric structure parameters and the thermal stress σ; adopts the NSGA-Ⅱ genetic algorithm to obtain the Pareto frontier solution of the multi-objective optimization of the integrated turbine box, and determines the optimal solution of the structural parameters according to the strength criterion, and obtains the optimal design parameters: the fillet r of the transition part is 20.5 mm, and the wall thickness δ is 2.5 mm. The weight of the optimized model is reduced by 8.3% compared with the initial scheme, and meets the strength requirements under thermal load.

[0075] Through the above method, the integrated turbine box parameter design can be converted into the calculation of a multi-objective optimization problem, thereby minimizing the difficulty and workload of the multi-objective optimization design.

[0076] The present invention performs multi-objective optimization of the integrated turbine case based on thermal-fluid-solid coupling simulation and NSGA-Ⅱ genetic algorithm, solves the multi-objective optimization problem of the integrated turbine case, effectively ensures its structural strength while lightweighting the integrated turbine case, and thus ensures the reliability of the supercharger operation.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-objective optimization design method for an integrated turbine case, characterized by: The steps include: Step 1: Establish a parametric model of the integrated turbine case structure, select the geometric structural parameters to be optimized, including the wall thickness δ and fillet r of the transition from the exhaust pipe to the volute, and determine the optimization targets, including the integrated turbine case weight m and thermal stress σ; Step 2: Determine the constraint range of the optimized structural parameters, generate the sample design space, and use the Latin hypercube sampling method to form a sample library; Step 3: Using Creo software, the mass of the integrated turbine case under various structural parameters is obtained, and the response surface methodology is used to construct a mapping relationship between the geometric structural parameters and the weight m of the integrated turbine case. Step 4: Perform thermal-fluid-solid coupling numerical simulation analysis based on the sample points in the sample library. By calling the prepared macro command files of Creo and ANSYS CFX software, parametric modeling, meshing, pre-processing, simulation calculation and post-processing are completed to obtain the thermal stress σ corresponding to the sample space; Step 5: Construct a BP neural network model based on the optimization target value obtained in step 3 to obtain the mapping relationship between geometric structure parameters and thermal stress σ; Step 6: Based on the mapping model of steps 4 and 5, the NSGA-Ⅱ genetic algorithm is used to obtain the Pareto frontier solution of the multi-objective optimization of the integrated turbine box, and the optimal solution of the structural parameters is determined according to the strength criterion.

2. The multi-objective optimization design method for an integrated turbine case according to claim 1, characterized in that: In step 3, the density of the turbine housing material is set in Creo software, and the weight m of the turbine housing is calculated by the software. The response surface approximation function is fitted using a second-order polynomial. The input variables are the wall thickness δ and the fillet r, and the output variable is the weight m. The function is expressed as follows: In the formula, a0 is the constant term to be determined, a j is the coefficient of the second term to be determined, a ij is the undetermined coefficient of the quadratic term, k is the number of design variables x i 、x j is the input variable.

3. The multi-objective optimization design method for an integrated turbine case according to claim 1, characterized in that: In step 4, the thermal-fluid-solid coupling numerical simulation steps are as follows: select the RNG k-ε model, the boundary conditions are: the turbine box inlet adopts a pressure inlet, the inlet total pressure is 334kPa, the inlet temperature is 1023K, the outlet is a mass flow outlet, which is 0.19kg / s, the impeller area is set as the rotating domain, the speed is 140000r / min, the rest is the stationary domain, the inner wall of the turbine box is set to be impermeable, non-slip, and adiabatic, and the convergence accuracy is set to 10 -5 The number of iteration steps is 800. CFX is used to perform steady-state solution to obtain the temperature and convection heat transfer coefficient of the fluid domain on the inner wall of the turbine box. These are then imported into the structural field and the turbine box thermal stress σ is obtained through simulation calculation.

4. The multi-objective optimization design method for an integrated turbine case according to claim 1, characterized in that: In step 5, the BP neural network includes an input layer, a hidden layer, and an output layer. The input variables are the wall thickness δ and the fillet r, and the output variable is the thermal stress σ. The BP neural network construction includes the following steps: A1. The constructed neural network model has 2 input layer nodes, 8 hidden layer nodes, and 1 output layer node. The input is wall thickness δ and fillet r, and the output is thermal stress σ. A2. BP neural network uses Trainlm training method, and the Tan-sigmoid transfer function is used from the input layer to the hidden layer. The transfer function between the hidden layer and the output layer is the Purelin transfer function Ψ(x); 75% of the sample space is used as the neural network training set, and 25% is used as the test set. The structural parameters wall thickness δ and fillet r are input vectors, expressed as X = [x1, x2] T , the optimization target thermal stress σ is the output vector, expressed as Y = [y1]; A3, BP neural network training times is 1000, learning rate is 0.001, BP neural network is trained, and the mean square error loss function is used to represent the neural network prediction value y k p and target value y k The error between the two layers is updated by gradient descent method, and the weights and thresholds of the hidden layer and output layer are updated until the mean square error is less than 0.5, and the training is stopped. A4. Use the trained neural network to predict the test set. If the mean square error is greater than 0.5, return to retrain. If it is less than 0.5, output a qualified BP neural network approximate model.

5. The multi-objective optimization design method for an integrated turbine case according to claim 4, characterized in that: In step A2, the functional relationship between the input vector and the output vector is: Y=y(X)=Ψ[W ki Φ(W ij X+A)+B](2) Where W ki is the weight matrix between the hidden layer and the output layer, B is the threshold matrix of the output layer, W ij is the weight matrix between the input layer and the hidden layer, and A is the threshold matrix of the hidden layer.

6. The multi-objective optimization design method for an integrated turbine case according to claim 4, characterized in that: In step A3, the neural network predicts the value y k p and target value y k The error between them is expressed as:

7. The multi-objective optimization design method for an integrated turbine case according to claim 1, characterized in that: In step 6, the mathematical description of the multi-objective optimization of the integrated turbine case is as follows: The optimization objective functions include: the integrated turbine box has the smallest weight, the highest structural strength, and the optimized design variables must be within a limited range.

8. The multi-objective optimization design method for an integrated turbine case according to claim 7, characterized in that: In step 6, the construction of the NSGA-II genetic algorithm includes the following steps: B1. Determine the initial model input parameters: wall thickness δ and fillet r, and determine the constraint range of the parameters; B2. Determine the objective functions f1 and f2; B3. According to the actual parameter variables of the genetic algorithm model, the genetic algorithm initialization parameters are set: the population size is 50, the number of optimization targets is 2, the crossover method is two-by-two crossover according to probability, the number of crossover individuals is 36, the crossover probability is 0.7, the number of mutation individuals is 14, and the mutation probability is 0.2; B4. Randomly generate an initialized population, calculate the objective function value corresponding to the initial solution, perform non-dominated sorting on the initial population, obtain a set of non-dominated solutions, and calculate the congestion degree of the initial population; B5. Select the initialized population as the parent population, generate the offspring population through genetic operations, merge the parent population and offspring population into a candidate population, perform non-dominated sorting and crowding calculation, and select the top 50 individuals with higher non-dominated levels and higher crowding as the new parent population for genetic operations, and so on; B6. Determine whether the convergence conditions are met, such as reaching the maximum number of iterations or converging to a satisfactory Pareto front solution, and then generate an optimal solution for a set of structural parameters; If it is not satisfied, a new parent population is generated and iterated again until the maximum number of iterations is reached. The optimal solution of the next set of structural parameters with the maximum number of iterations is output and the iteration is terminated. B7. According to the strength criterion σ≤[σ] / S, select the optimal solution from a series of Pareto front solutions.

Images