CN102819651A - Simulation-based parameter optimizing method for precise casting process of single crystal turbine blade - Google Patents

Abstract

The invention discloses a simulation-based single crystal turbine blade precision casting process parameter optimization method, which is used to solve the technical problem that the existing single crystal turbine blade precision casting process parameter optimization method has poor optimization effect. The technical solution is to design the gating system model of the single crystal turbine blade, use the finite element analysis method to divide the gating system model into units, extract and analyze the precision casting process parameters of the blade, and obtain the surface size of the precision casting blade through the single factor process test. Precision casting process parameters and reasonable parameter variation ranges that have a greater impact on precision; numerical simulation of the pouring process based on the established test table; through data processing and analysis, evaluate the deformation of the casting relative to the design model, and establish a BP neural network model, The search range of precision casting process parameter optimization is gradually narrowed by adopting the small step size search method. The optimization effect of the single crystal turbine blade precision casting process parameter optimization method is improved.

Description

Optimization method of precision casting process parameters for single crystal turbine blades based on simulation

technical field

The present invention relates to a single crystal turbine blade precision casting process parameter optimization method, in particular to a single crystal turbine blade precision casting process parameter optimization method based on simulation.

Background technique

The complex hollow turbine blade is the core technology of the high thrust-to-weight ratio engine, the core component of the engine, and the easily broken and failed parts. Its performance level, especially its ability to withstand high temperatures, is an important symbol of the advanced level of a type of engine, and in a certain sense, it is also a significant symbol of the level of a country's aviation industry. Due to its complex internal cooling structure, high requirements for aerodynamic shape and dimensional accuracy, and long-term service under conditions of strong thermal shock and complex cyclic thermal stress, the structural design and manufacturing technology of hollow turbine blades has become the core of high thrust-to-weight ratio aeroengines technology.

Hollow turbine blades are generally precision cast with directional crystallization or single crystal without margin. Since the turbine blades are thin-walled structures (with a wall thickness of 0.5mm-2mm) composed of a large number of free-form surfaces and complex inner cavities, the surface precision of precision casting blades is low and the wall Large thickness drift, unstable quality, and high scrap rate have always been the bottleneck restricting the development of new aero-engines in my country.

Precision casting of blades is a complex dynamic process in which multiple deformation factors and multiple stress and resistance sources are coupled with each other, and what affects blade size is also the coupling effect of multiple factors. The domestic casting process generally adopts the "experience + test" method, that is, relying on production experience process parameters, through repeated field tests to study the influence of precision casting process parameters on the forming quality of single crystal turbine blades. This method is low in intelligence and automation, high in cost, and has a long research and development cycle. There is no scientific theoretical basis, and it is difficult to effectively optimize the parameters of the precision casting process. It can no longer meet the needs of the rapid development and fierce competition of the modern market.

Contents of the invention

In order to overcome the disadvantage of poor optimization effect of the existing single crystal turbine blade investment casting process parameter optimization method, the present invention provides a simulation-based single crystal turbine blade precision casting process parameter optimization method. In this method, the gating system of the single crystal turbine blade is designed, the gating system model is divided into units by the finite element analysis method, the parameters of the precision casting process of the blade are extracted and analyzed, and the influence on the accuracy of the precision casting blade surface size is obtained through the single factor process test. Large investment casting process parameters and reasonable parameter variation range; design interactive orthogonal experiments, and perform numerical simulation of the pouring process according to the established test table to obtain the deformation of the turbine blade castings during the pouring process; through data processing and analysis, evaluate For the deformation of the casting relative to the design model, by establishing the BP neural network model and using the small step size search method, the search range for the optimization of the precision casting process parameters can be gradually narrowed, and the optimization effect of the single crystal turbine blade precision casting process parameter optimization method can be improved. .

The technical solution adopted by the present invention to solve the technical problem is: a simulation-based optimization method for single crystal turbine blade precision casting process parameters, which is characterized in that it includes the following steps:

Step 1: Establish a single crystal turbine blade gating system model, and use the finite element analysis method to divide the gating system model into units.

Step 2: Extract and analyze the precision casting process parameters of the single crystal turbine blade, and obtain the precision casting process parameters and reasonable parameter variation ranges that have a great influence on the dimension accuracy of the single crystal turbine blade surface through a single factor process test.

Step 3: Design an interactive orthogonal test form according to the investment casting process parameters in Step 2.

Step 4: Carry out numerical simulation of the pouring process according to the established interactive orthogonal test table, so as to obtain the deformation of the turbine blade casting during the pouring process. Firstly, the numerical simulation boundary conditions are applied, including the thermophysical parameters of the alloy material and the mold shell material, the alloy temperature of the initial pouring, the alloy temperature at which the numerical calculation is terminated, the interface heat transfer coefficient between the alloy material and the precision casting mold shell, and the constraint conditions of the model displacement . Through the solution of the stress field in the precision casting process, the stress distribution of each node of the turbine blade mesh model in the precision casting process is obtained, and then the displacement of each node is derived, and the displacement field model is established.

Step 5: Perform data processing and analysis according to the simulation results obtained in Step 3. Use the equal parameter method to intercept multiple two-dimensional sections of the blade at different heights, use the method of corresponding points to calculate the two-dimensional displacement distribution of the casting section, and then express the three-dimensional displacement field of the casting from the set of two-dimensional displacement distributions of multiple sections, and evaluate The deformation of the casting relative to the design model. Combined with the interactive orthogonal test table in step 3, the neural network training samples and the optimal investment casting process parameters are obtained. Specific steps are as follows:

[1] Export the blade simulation model through the ViewCAST module of ProCAST, and the data format is "*.sm";

[2] Convert "*.sm" format to "*.STL" format;

[3] Import the casting system model of the single crystal turbine blade and the CAD model of the single crystal turbine blade wax model in step 1 for three-dimensional registration;

[4] After 3D registration, 5 to 8 section lines are cut along the height direction of the model to obtain the two-dimensional cross section of the single crystal turbine blade casting system model and the single crystal turbine blade wax model CAD model at the same height, and the cross section is derived at the same time Wire;

[5] Discretize the intercepted section lines with equal parameters, sort the discrete points, and use the UG secondary development module to read the discrete points into the calculation of the displacement between the corresponding discrete points;

[6] Combined with the interactive orthogonal test table in step 3, normalize the data and analyze the range, and obtain the neural network training samples and the optimal investment casting process parameter combination.

Step 6: Establish a BP neural network model, and use the neural network training samples obtained in step 5 to train the neural network.

Step 7: Combining with the BP neural network model established in step 6, adopt the small step size search method to gradually narrow the search range of investment casting process parameters optimization, and finally make the surface displacement corresponding to the optimized parameters smaller than the required value. The process of small step size search optimization method is as follows:

[1] Increase and decrease the micro-step size δ _i (i=1, 2, 3, 4) for each variable near the obtained maximum point, and combine these parameters into a combination of multiple sets of precision casting process parameters, That is, a new orthogonal table is generated;

[2] Through the calculation of BP neural network model, a series of blade surface displacements △Z are obtained;

[3] Find the investment casting process parameter combination corresponding to the smaller blade surface displacement;

Return to step [1] until the value of displacement ΔZ of the blade profile meets the requirement.

The beneficial effects of the present invention are: due to the design of the single crystal turbine blade gating system, the finite element analysis method is used to divide the gating system model into units, the blade precision casting process parameters are extracted and analyzed, and the accuracy of the precision casting is obtained through the single factor process test. The precision casting process parameters and reasonable parameter variation ranges that greatly affect the blade profile size accuracy; design interactive orthogonal experiments, and conduct numerical simulations of the pouring process according to the established test table to obtain the deformation of the turbine blade castings during the pouring process; through Data processing and analysis, evaluating the deformation of the casting relative to the design model, through the establishment of a BP neural network model, and the use of a small step size search method, gradually narrowed the search range for the optimization of investment casting process parameters. The optimization effect of the single crystal turbine blade precision casting process parameter optimization method is improved.

The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

Description of drawings

Fig. 1 is a flow chart of the simulation-based method for optimizing the technical parameters of single crystal turbine blade precision casting in the present invention.

Fig. 2 is a model diagram of a certain type of turbine power blade used in the method of the present invention.

Figure 3 is a CAD model of the gating system used in the method of the present invention.

Fig. 4 is the variation curve of profile displacement with drawing speed in the method of the present invention.

Fig. 5 is a schematic diagram of the temperature field during the solidification process of a certain type of turbine power blade used in the method of the present invention.

Fig. 6 is a flowchart of BP neural network training of the method of the present invention.

Fig. 7 is a search optimization flow chart of the method of the present invention with a small step size.

Detailed ways

The following embodiments refer to FIGS. 1-7.

Step 1: Use a certain type of turbine power blade, the main parameters of which are blade body length 106.10mm, maximum chord length 56.21mm, maximum inscribed circle radius 5.7mm, leading edge radius 4.12mm, and trailing edge radius 1.25mm. The blade is made of the second generation single crystal superalloy DD6, and the mold shell is made of silica sand. According to casting feeding theory and actual production experience, the blade casting process and gating system are designed for power blades, using top injection, 2 pieces in a set.

Step 2: Use commercial finite element pre-processing software Hypermesh (product of Altair, USA) to divide the model into units based on non-uniform mesh division technology. First, import the gating system model into Hypermesh and discretize it into tetrahedral units. The quality meets the quality requirements of general enterprise finite element analysis. In this embodiment, more than 95% of the unit quality is required to meet: unit warpage less than 5.0, unit long-short side ratio less than 5.0, skew less than 60.0, and unit Jacobian ratio greater than 0.7. The total number of tetrahedral units is 737,000.

Step: 3: Extract the investment casting process parameters that have a greater impact on the accuracy and size of the precision casting surface of the turbine blade, such as drawing speed, pouring temperature, mold shell preheating temperature, cold copper temperature, mold shell thickness, heat preservation temperature, etc. , the present invention only studies the first four parameters, and the remaining parameters are kept at fixed values in the test.

A single factor process test was used to study the influence of investment casting process parameters on the precision dimension of the turbine blade profile. The test form and test results are shown in Table 1.

Table 1 Single factor process test table and results

In order to be able to more intuitively see the influence and regularity of each process parameter on the size of the investment casting surface, a drawing tool is used to plot the precision casting process parameters as the abscissa and the resulting displacement of the profile as the ordinate.

According to the drawn curve, it can be found that with the change of investment casting process parameters, the size of the blade surface shows an obvious change rule, so as to determine the reasonable selection of parameters and the factor level table of the orthogonal test, as shown in Table 2 Show.

Table 2 Factor level table

Step 4: From step 2, the investment casting process parameters that need to be optimized include drawing speed, pouring temperature, mold shell preheating temperature, and cold copper temperature. Three levels are selected for each factor, and the interaction between the three factors of drawing speed, pouring temperature, and mold shell preheating temperature is considered at the same time, and the L23 (313) interaction table design is adopted. The specific test table is as follows:

Table 3 Interactive test table

A Pulling speed B Pouring temperature C Mold shell preheating temperature D Cold copper temperature

Step 5: According to the orthogonal test table in step 3, use ProCAST to carry out the numerical simulation of the precision casting process of the turbine blade. The alloy is DD6 high-temperature nickel-based alloy, whose liquidus temperature is 1380°C and solidus temperature is 1310°C. Its thermal conductivity is 33.2W/m·K, its density is 8780kg/m ³ , and its specific heat is 99.0KJ/Kg/K. The mold shell is made of silica sand with a thermal conductivity of 0.59W/m·K, a density of 1520kg/m ³ and a specific heat of 1.20KJ/Kg/K. The alloy temperature at which the numerical simulation was terminated was 600 °C. The displacement constraints are fixed at the bottom of the sprue and at the bottom of the blade seeding section and at the bottom of the cold copper.

Step 6: Process and analyze the numerical simulation results of the directional solidification process. The blade simulation model and the blade CAD model that have completed the three-dimensional registration are cut along the Z-axis to 5 section lines with heights of 320mm, 330mm, 340mm, 350mm, 360mm, 370mm, 380mm, and 390mm. The processing and analysis results are shown in the following table:

Table 4

27 groups of experiments were used as neural network training samples. The optimal investment casting process parameters were drawing speed 5mm/min, pouring temperature 1550°C, mold shell preheating temperature 1470°C, and cold copper temperature 24°C.

Step 7: Using a three-layer artificial neural network, train and establish a neural network optimization model. The neural network adopts the BP network based on the Liebenberg-Marquardt optimization algorithm, the hidden layer is a Sigmoid activation function, and the output layer is a Purelin activation function. The neural network structure of investment casting process parameters and blade surface displacement is: 4 nodes in the input layer, whose parameters are the extracted investment casting process parameters, including drawing speed, pouring temperature, mold shell preheating temperature, and cold copper temperature. The output layer is a node whose parameter is the output of the numerical simulation results, that is, the displacement of the blade profile. Use the training samples to train the established neural network and verify the neural network. If the error is within the allowable range, the threshold and weight matrix of the neural network will be obtained as the neural network matrix. In this way, a mapping relationship model is established, which can map the relationship between the precision casting process parameters and the displacement of the blade profile. Figure 6 is a BP neural network structure diagram, where A is the drawing speed, B is the pouring temperature, C is the preheating temperature of the mold shell, and D is the cold copper temperature.

Step 8: Combining with neural network, adopt small step size search method to optimize investment casting process parameters. The initial micro-step size δ _i (i=1, 2, 3, 4) is δ ₁ =0.2mm/min, δ ₂ =20°C, δ ₃ =20°C, δ ₄ =1°C, and the precision of the blade surface size ΔZ=0.185mm. After the first calculation, △Zmin=0.1820mm, already meets the precision requirement. The finally obtained optimized investment casting process parameter group is as follows:

table 5

Pulling speed pouring temperature Shell preheating temperature cold copper temperature displacement final optimization plan 4.8 1550 1480 twenty four 0.1820

Claims (1)

1. single crystal turbine blade precision casting technology parameter optimization method based on emulation is characterized in that may further comprise the steps:

Step 1: set up single crystal turbine blade running gate system model, adopt finite element method that the running gate system model is carried out dividing elements;

Step 2:, and obtain precision casting technology parameter and the Reasonable Parameters variation range bigger to single crystal turbine blade molding surface size precision influence through single factor engineer testing to single crystal turbine blade precision casting technology parameter extraction and discrimination;

Step 3: according to the mutual orthogonal test form of the precision casting technology parameter designing in the step 2;

Step 4: carry out the numerical simulation of casting process according to the mutual orthogonal test form of setting up, to obtain the turbo blade casting deformation situation in the casting process; At first apply the numerical simulation boundary condition, comprise the thermal physical property parameter of alloy material and formwork material, initial cast alloy temperature, end the interface heat exchange coefficient between alloy temperature, alloy material and the finish cast die shell of numerical evaluation, the constraint condition of model displacement; Through finding the solution of essence casting process stress field, draw the stress distribution of smart each node of casting process turbo blade grid model, and then derive the displacement of each node, set up the displacement field model;

Step 5: the simulation result that obtains according to step 3 carries out data processing and analysis; A plurality of two-dimensional sections on the parametric method intercepting blade differing heights such as use; Adopt the two-dimension displacement in the method calculating foundry goods cross section of corresponding point to distribute; Expressed the three-D displacement field of foundry goods again by the two-dimension displacement distributed collection in a plurality of cross sections, the assessment foundry goods is with respect to the distortion situation that designs a model; Mutual orthogonal test form in the integrating step three obtains train samples and best precision casting technology parameter; Concrete steps are following:

[1] the ViewCAST module through ProCAST derives the blade realistic model, and data layout is " * .sm ";

[2] be " * .STL " form with " * .sm " format conversion;

[3] step 1 is set up single crystal turbine blade running gate system model and single crystal turbine blade wax-pattern cad model importing carrying out three-dimensional registration;

[4] after three-dimensional registration, along 5～8 section lines of model short transverse intercepting, obtain single crystal turbine blade running gate system model and single crystal turbine blade wax-pattern cad model at the two-dimensional section of sustained height, derive section line simultaneously;

[5] section line to intercepting waits parameter discrete, and with the discrete point ordering, uses the UG Secondary Development Module and discrete point is read in the displacement that calculates between the corresponding discrete point;

[6] the mutual orthogonal test form in the integrating step three is done normalization processing and range analysis to data, obtains train samples and best precision casting technology parameter combinations;

Step 6: set up the BP neural network model, with the train samples neural network training that obtains in the step 5;

Step 7: the BP neural network model of setting up in the integrating step six, adopt little step length searching way, dwindle the hunting zone of precision casting technology parameter optimization gradually, finally make the corresponding profile displacement of parameters optimization less than requiring numerical value; Little step length searching optimization method process is following:

[1] near the value point that has obtained, each variable is increased and reduces small step-length δ _i(i=1,2,3,4) mix into many group precision casting technology combinations of parameters to these parameters, promptly generate new orthogonal table;

[2] calculate a series of blade profile displacement △ Z through the BP neural network model;

[3] search and obtain the corresponding precision casting technology parameter combinations of vanelets profile displacement more;

Return step [1] till blade profile displacement △ Z numerical value reaches requirement.

Images