A method for inverse identification and prediction of initial residual stress

Abstract

This invention belongs to the technical field of thin-walled part processing and discloses a method for inverse identification and prediction of initial residual stress. The prediction method includes: S1 For the same batch of parts, eliminating the initial residual stress of the test parts, milling them, and measuring their surface residual stress; S2 Setting the initial residual stress to zero, adjusting simulation parameters so that the difference between the simulated residual stress and the surface residual stress obtained in step S1 is within an acceptable threshold range, saving the current simulation parameters, applying different initial residual stresses, simulating to obtain different surface residual stresses, and constructing a database with a one-to-one correspondence between initial residual stress and surface residual stress; S3 Using the database to construct a prediction model, inputting the surface residual stress of the part to be tested into the prediction model to obtain the initial residual stress of the part to be tested, thus achieving inverse identification of the initial residual stress. This invention solves the problems of high destructiveness and cumbersome measurement of residual stress in existing technologies.

Description

Technical Field

[0001] This invention belongs to the technical field of thin-walled part processing, and more specifically, relates to an inverse identification and prediction method for initial residual stress. Background Technology

[0002] Aero engines comprise key thin-walled components such as casings, impellers, and shafts. The engine's service performance largely depends on the surface properties of these components and the assembly of the parts. This patent focuses primarily on the machinability of these components. Controlling deformation during the design and manufacturing process is a crucial step in the form-property co-manufacturing of thin-walled parts. These components are characterized by their thin walls, and the deformation generated during machining is affected not only by the forces and heat generated during machining but also by the initial residual stress within the component. Detecting this initial residual stress has always been a challenge in the manufacturing process. Research on methods for detecting this initial residual stress is beneficial for revealing the deformation mechanism within the component under the influence of initial residual stress.

[0003] Traditional methods for measuring initial residual stress at great depths include neutron beam methods, delamination drilling methods, delamination X-ray diffraction methods, and crack compliance methods. Among these, the neutron beam method requires scarce and expensive equipment, while the other methods require destructive delamination of the part, thus disrupting the original stress balance within the material, and the measured residual stress is not the original initial residual stress within the material. These methods suffer from drawbacks such as significant damage to the part, cumbersome measurement methods, and the results being severely affected by the delamination method. Therefore, a non-destructive and convenient method for obtaining the initial residual stress of a material is needed. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement needs of the prior art, the present invention provides an inverse identification and prediction method for initial residual stress, which solves the problems of the large damage to parts caused by measuring residual gravitational force and the cumbersome measurement in the prior art.

[0005] To achieve the above objectives, according to the present invention, an inverse identification and prediction method for initial residual stress is provided, the prediction method comprising the following steps:

[0006] S1 For the same batch of parts, the batch of parts is divided into two parts, one part is used as the parts to be tested and the other part is used as the test parts. The test parts are subjected to stress relief treatment to eliminate their initial residual stress. The test parts with the initial residual stress eliminated are milled and then the surface residual stress of the test parts is measured.

[0007] S2 uses finite element simulation to simulate the initial residual stress and surface residual stress of the part. The initial residual stress is set to zero, and the simulated surface residual stress is obtained. The simulated residual stress is compared with the surface residual stress of the test part obtained in step S1. The simulation parameters are adjusted until the difference between the simulated residual stress and the surface residual stress of the test part is within an acceptable threshold range. The current simulation parameters are saved, different initial residual stresses are applied, and different surface residual stresses are obtained through simulation. A database with a one-to-one correspondence between initial residual stress and surface residual stress is constructed.

[0008] S3 uses the database to construct a prediction model, tests the surface residual stress of the part to be tested, and inputs the obtained surface residual stress into the prediction model to obtain the initial residual stress of the part to be tested, thereby realizing the inverse identification of the initial residual stress.

[0009] More preferably, in step S1, the stress relief treatment employs annealing to eliminate initial residual stress.

[0010] More preferably, in step S1, the residual stress on the surface is measured using X-ray measurement.

[0011] More preferably, in step S2, the simulation parameters are the angle between the rake face and the base plane, and the clearance angle is the angle between the clearance face and the cutting plane, i.e., the rake angle and the clearance angle.

[0012] More preferably, the front angle is simulated using the effective average front angle instead of the actual front angle.

[0013] More preferably, the effective average rake angle is calculated according to the following expression:

[0014]

[0015] Among them, h a For the cutting thickness, r e Let α0 be the radius of the cutting edge arc of the tool, and α0 be the nominal rake angle of the tool. θ is an empirical constant, typically taking the value 2. f This is the separation angle between the tool and the chip, typically taken as 37.6°.

[0016] More preferably, in step S2, the different residual stresses are applied from the depth direction of the part.

[0017] More preferably, in step S3, the prediction model adopts a neural network prediction model.

[0018] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0019] 1. This invention utilizes machine learning inverse identification methods to identify the intrinsic relationship between initial residual stress and surface residual stress. By performing simple milling operations on the blank and measuring the surface residual stress after machining, the internal residual stress of the blank material after forging, heat treatment, and aging treatment is obtained, referred to as initial residual stress. Unlike surface residual stress after machining, which is mostly distributed within 1 mm of the surface layer, initial residual stress is distributed throughout the entire material and is released after the material is removed during machining. It has a crucial impact on the deformation of thin-walled aerospace parts.

[0020] 2. The initial residual stress prediction method provided by this invention characterizes the residual stress inside the material. Compared with existing measurement methods, it is a reverse identification method. It identifies the processing result in the forward direction through processing parameters and processing conditions, thereby optimizing the processing parameters and conditions. Based on machine learning neural networks, it takes a processing database such as initial residual stress and processing conditions as input. By learning the relationship between initial residual stress and residual stress on the processed surface, it performs reverse identification of initial residual stress, providing a measurement basis for the deformation control of part processing.

[0021] 3. In this invention, the initial residual stress of the test part is zeroed out. For the test part, the initial residual stress is unknown because it is located inside. The surface residual stress after milling will be affected by the initial residual stress and this relationship is unknown. When the test part is stress-relieved, its initial residual stress is approximately 0. After milling, the surface residual stress value is measured to know the magnitude of the surface residual stress value after milling when the initial residual stress is 0. At this time, applying 0 initial residual stress can obtain the simulated surface residual stress. When the simulated surface residual stress value and the known surface processing value of the workpiece with 0 initial residual stress are within an acceptable threshold range, it is considered that the influence mechanism caused by processing and simulation is the same. The purpose of zeroing out is to prove that the value obtained by simulation can approximately represent the actual value.

[0022] 4. This invention calibrates a zero-stress block, using the approximate zero initial residual stress after annealing as a benchmark. It employs finite element method to apply the initial residual stress, simulates the residual stress after processing, and performs database construction and machine learning inverse identification fitting with the actual processed surface to obtain the original internal residual stress distribution of the part, providing deformation control guidance for the production of aerospace thin-walled parts. Attached Figure Description

[0023] Figure 1 This is a flowchart of the inverse identification and prediction method for initial residual stress constructed according to a preferred embodiment of the present invention;

[0024] Figure 2This is a schematic diagram of a heat treatment process based on Ti6Al4V material constructed according to a preferred embodiment of the present invention;

[0025] Figure 3 This is a finite element simulation result diagram of the initial residual stress loading of 0 MPa in the x-direction constructed according to the preferred embodiment of the present invention;

[0026] Figure 4 This is a finite element simulation result diagram of the initial residual stress loading of 100 MPa in the x-direction according to the preferred embodiment of the present invention;

[0027] Figure 5 This is a structural diagram of initial residual stress inverse identification performed according to the preferred machine learning neural network of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0029] like Figure 1 As shown, an inverse identification and prediction method for initial residual stress is proposed, which includes the following steps:

[0030] Step 1: Take one workpiece from the same batch of parts (assuming they have undergone the same process and have the same initial residual stress) and perform a heat treatment process (after heat treatment, it becomes a Class B workpiece). The purpose of heat treatment is to make the initial residual stress value distribution of the workpiece approximately zero. Here, the stress test block after annealing should be ±20 MPa after measurement. At this time, it is used as a calibration block and is approximately a zero stress block.

[0031] like Figure 2 As shown, the heat treatment process described above is carried out using a vacuum annealing furnace. The specific annealing process and holding time are usually... Figure 2 The process begins with heating to 600℃ at a rate of 100℃ / h, followed by holding at that temperature for 3 hours in a vacuum annealing furnace to prevent oxidation of the metal workpiece during annealing. Afterward, the workpiece is removed and air-cooled to complete the stress-relief annealing. It is assumed that the annealing stress of the part is zero at this point. This process uses Ti6Al4V material as an example; different materials require adjustments to the annealing process.

[0032] Step Two: The initial residual stress of the same batch of specimens is calibrated to zero. The calibration method involves milling the heat-treated zero-stress specimens. In this patent, Ti6Al4V is used as the sample, with a tool parameter of φ17R0.4, a specially made tool from Seco, model CT-JHP780170R040.0Z4. Appropriate machining parameters are selected. The tool parameters chosen in this invention are: cutting width 0.3mm, cutting depth 20mm, spindle speed 1200r / min, cutting speed 64m / min, and feed per tooth 0.1mm. These machining parameters will yield different results with different tools and materials, requiring testing for selection. The selection of machining parameters is mainly based on the following two points: a small radial depth of cut during milling to avoid excessive influence on the initial residual stress distribution of the workpiece; and machining parameters should result in a high surface roughness (approximately 20μm) to meet the detection standards of the X-ray residual stress analyzer.

[0033] The residual stress on the machined surface of a part is simulated using the finite element method. For zero-stress specimens, no initial residual stress is applied in the finite element simulation. During the finite element simulation, the tooling and machining conditions should be set as closely as possible to reality, such as cutting fluid conditions and tool coating information. The finite element simulation is adjusted so that the error between the simulation results with zero initial residual stress and the measured values ​​of the actual machining results is within 10%. At this point, the simulated environment is considered to approximate the actual environment, meaning that the effects caused by the simulation are approximately the same as the effects caused by machining.

[0034] In the finite element simulation process, orthogonal turning is used instead of milling for simulation. Based on the conversion between milling and turning, some optimization and settings of the simulation parameters are required. The effective average rake angle is used instead of the rake angle for simulation. The calculation process of the effective rake angle is shown below:

[0035]

[0036] In the formula, ha is the cutting thickness, and r e Let α0 be the radius of the cutting edge radius of the tool, and α0 be the nominal rake angle of the tool. θ is an empirical constant, typically taken as 2. f This is the separation angle between the tool and the chip, typically taken as 37.6°.

[0037] Step 3: Calibrate other initial residual stresses. The surface residual stress values ​​obtained under the same processing conditions will differ due to the influence of the initial residual stress. Different initial residual stresses are applied in the finite element method. Here, initial residual stresses are applied at intervals of 10 MPa. Because the range of initial residual stress in the layer distribution is from -100 to 100 MPa, the initial residual stress is applied from -100 MPa to 100 MPa, resulting in 20 sets of finite element simulation experiments.

[0038] like Figure 3 As shown, this is the zero-value calibration curve for the initial residual stress without loading. The residual stress value within a 10 μm range is taken as the residual stress value on its surface (the X-ray residual stress meter measures values ​​within approximately 10 μm). Figure 3 The initial residual stress value is approximately 130 MPa tensile stress, and the trough value of its spoon-shaped curve is -270 MPa compressive stress.

[0039] The analysis can be performed by applying different initial residual stresses along the depth direction. The residual stress mainly has three values ​​σ. xx σ yy σ xy (In 2D simulation).

[0040] like Figure 4 The figure shows the simulation curve of the initial residual stress value under a load of 100 MPa. The residual stress value within a 10 μm range is taken as the residual stress value on its surface. Figure 4 The initial residual stress value is approximately 230 MPa tensile stress, and the trough value of its spoon-shaped curve is -120 MPa compressive stress. Here, the surface stress value is used as the relevant standard for simulation. This simulation is a demonstration case, and the values ​​produced by different workpiece materials and processing parameters may be different.

[0041] Step 4: After applying the initial residual stress, perform a simulation to obtain the residual stress value of the simulated machined surface. Figure 4 The corresponding value is 230 MPa. Since it has been calibrated with a 10% error to zero, it is assumed that the finite element simulation with initial residual stress is the same as the actual machining condition of the unannealed workpiece affected by initial residual stress. Figure 3 The corresponding stress is 100 MPa. A database of initial residual stress and surface simulation residual stress is created using finite element simulation values ​​at 10 MPa intervals. The same processing parameters as the zero-stress test block are used to process the untreated workpiece (Class A workpiece). The surface processing stress is measured using an X-ray residual stress meter. The measured surface processing stress under the influence of the initial residual stress is then compared with the surface residual stress value obtained from the finite element simulation with the initial residual stress applied for inverse identification.

[0042] like Figure 5 As shown, a machine learning neural network is used, taking the initial residual stress loading value and processing conditions as inputs. The machine learning obtains the inverse identification mapping relationship between the initial residual stress loading value and the simulated surface residual stress after processing. Then, the initial residual stress along the thickness direction of the workpiece is obtained by inverse mapping from the residual stress value of the actual processed surface. This is used to predict the initial residual stress of parts in the same batch.

[0043] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for inverse identification and prediction of initial residual stress, characterized in that, The prediction method includes the following steps: S1 For the same batch of parts, the batch of parts is divided into two parts, one part is used as the parts to be tested and the other part is used as the test parts. The test parts are subjected to stress relief treatment to eliminate their initial residual stress. The test parts with the initial residual stress eliminated are milled and then the surface residual stress of the test parts is measured. S2 uses finite element simulation to simulate the initial residual stress and surface residual stress of the part. The initial residual stress is set to zero, and the simulated surface residual stress is obtained. The simulated surface residual stress is compared with the surface residual stress of the test part obtained in step S1. The simulation parameters are adjusted until the difference between the simulated surface residual stress and the surface residual stress of the test part is within an acceptable threshold range. The current simulation parameters are saved, different initial residual stresses are applied, and different surface residual stresses are obtained through simulation. A database with a one-to-one correspondence between initial residual stress and surface residual stress is constructed. S3. A prediction model is constructed using the database, and the surface residual stress of the part to be tested is tested. The surface residual stress obtained from the test is input into the prediction model to obtain the initial residual stress of the part to be tested, thereby realizing the inverse identification of the initial residual stress. The inverse identification process is as follows: A machine learning neural network is used to take the initial residual stress loading value and processing conditions as input. The machine learning is used to obtain the inverse identification mapping relationship between the initial residual stress loading value and the surface residual stress after simulation processing. Then, the initial residual stress along the thickness direction of the workpiece is obtained by inverse mapping from the residual stress value of the actual surface after processing. This is used to predict the initial residual stress of parts in the same batch.

2. The inverse identification and prediction method for initial residual stress as described in claim 1, characterized in that, In step S1, the stress relief treatment employs annealing to eliminate initial residual stress.

3. The inverse identification and prediction method for initial residual stress as described in claim 1 or 2, characterized in that, In step S1, the residual stress on the surface is measured using X-ray measurement.

4. The inverse identification and prediction method for initial residual stress as described in claim 1 or 2, characterized in that, In step S2, the simulation parameters are the angle between the rake face and the base plane, and the clearance angle is the angle between the clearance face and the cutting plane, i.e., the rake angle and the clearance angle.

5. The inverse identification and prediction method for initial residual stress as described in claim 4, characterized in that, The effective average front angle is used instead of the actual front angle in the simulation.

6. The inverse identification and prediction method for initial residual stress as described in claim 5, characterized in that, The effective average front angle is calculated according to the following expression: in, h a For cutting thickness, r e The radius of the cutting edge arc of the tool. α 0 is the nominal rake angle of the cutting tool. This is an empirical constant with a value of 2. θ f The separation angle between the tool and the chip is 37.6°.

7. The inverse identification and prediction method for initial residual stress as described in claim 1 or 2, characterized in that, In step S2, the different residual stresses are applied from the depth direction of the part.

8. The inverse identification and prediction method for initial residual stress as described in claim 1, characterized in that, In step S3, the prediction model adopts a neural network prediction model.

Images