Manufacturing method of equal-wall-thickness metal stator cavity

Abstract

The invention discloses a manufacturing method of an equal-wall-thickness metal stator cavity, and belongs to the technical field of metal precision machining. Comprising the following steps: constructing a global coordinate system and establishing a conversion relation with a matrix coordinate system; the power and speed of laser cladding are adaptively controlled based on the cavity curvature, and a heat input gradient field is formed; constructing and fusing a five-dimensional multi-physical field data field; predicting wall thickness deviation by using a physical constraint neural network, and generating a tool path through stochastic differential programming; carrying out light pressure synergistic micro-machining; and enabling the wall thickness precision to be converged to a target range by adopting a learning type iterative controller. According to the method, multiple physical field factors are integrated, accurate control over the machining process is achieved through an intelligent algorithm, the problems of low manufacturing precision and low efficiency in the prior art are solved, and the manufacturing precision and efficiency of the equal-wall-thickness metal stator cavity can be improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of metal precision machining, in particular to a manufacturing method of an equal-wall-thickness metal stator cavity. BACKGROUND

[0002] In the fields of aerospace, high-end equipment manufacturing, etc., the equal-wall-thickness metal stator cavity is a key part, and the manufacturing precision directly affects the performance and reliability of the equipment. When machining the equal-wall-thickness metal stator cavity with a complex shape, the traditional manufacturing methods such as numerical control machining and casting are difficult to meet the requirements of high precision, high efficiency and low cost at the same time. With the continuous improvement of product performance requirements, the requirements for the wall thickness precision and surface quality of the metal stator cavity are becoming more and more strict, and the existing technology has the following problems: During the machining process, factors such as thermal deformation and stress deformation make it difficult to accurately control the wall thickness, and it is difficult to meet the high-precision requirements; The effective analysis and regulation of the multi-physical field coupling are lacked, and it is difficult to realize the optimization of the machining process; the traditional control strategy is difficult to adapt to the complex and changeable machining conditions, and the machining stability is poor and the efficiency is low. Therefore, a new manufacturing method is needed, which can comprehensively consider the multi-physical field factors, realize the accurate control of the machining process through intelligent algorithm, and improve the manufacturing precision and efficiency of the equal-wall-thickness metal stator cavity. SUMMARY

[0003] The purpose of the application is to provide a manufacturing method of an equal-wall-thickness metal stator cavity, which solves the problems of low manufacturing precision and low efficiency in the prior art by collecting and fusing multi-physical field data, predicting and deciding through intelligent algorithm, and cooperatively regulating the machining process, and realizes high-precision and high-efficiency manufacturing of the equal-wall-thickness metal stator cavity.

[0004] To achieve the above-mentioned purpose, the application provides a manufacturing method of an equal-wall-thickness metal stator cavity, comprising the following steps: S1, arranging a plurality of laser interferometers on a machining platform to construct a global coordinate system, marking a reference point on the surface of a metal base, and establishing a conversion relationship between the base coordinate system and the global coordinate system; S2, adaptively controlling the laser power and scanning speed in the laser cladding process based on the curvature of the cavity to form a heat input gradient field matched with the curvature distribution; S3, constructing a five-dimensional multi-physical field data field containing temperature field, strain rate field, sound pressure field, curvature and material hardness distribution, and performing data fusion; S4, predicting the wall thickness deviation by using a physically constrained neural network, solving the optimal compensation decision based on stochastic differential programming, and converting it into a tool path; S5, softening the material by laser preheating, cooperatively regulating the laser softening and cutting force, and performing light pressure cooperative micro-machining; S6. Establish convergence criteria and adopt a learning-based iterative controller. Based on a closed-loop iterative mechanism combining PID control and reinforcement learning, the wall thickness accuracy gradually converges to the target range.

[0005] Preferably, in step S1, constructing the coordinate system and establishing the transformation relationship specifically involves: Multiple sets of laser interferometers are arranged on the processing platform to construct a global coordinate system. Reference points are marked on the surface of the metal substrate. By measuring the coordinates of the reference points in the two coordinate systems, the rotation matrix and translation vector are calculated, and the transformation relationship between the substrate coordinate system and the global coordinate system is established.

[0006] Preferably, in step S2, the specific method for dynamically adjusting the laser power and scanning speed is as follows: Laser power according to adjust; Scanning speed according to adjust; Where sigmoid is the activation function, and its expression is: ; To scan the arc length of the path, For path curvature, For real-time temperature, and For reference power and speed, and This represents the material property coefficient.

[0007] Preferably, in step S3, temperature field, strain rate field, and sound pressure field data are collected in real time, and combined with curvature and hardness distribution data obtained during processing, a five-dimensional multiphysics data field is constructed, which is represented as follows: ; in, Indicates the temperature field distribution. Represents the strain rate field. The Laplace operator represents the sound pressure field. Indicates the curvature of the current processing area. Indicates the material hardness distribution; The data field is then fused using an improved Tucker decomposition, expressed by the following formula: ; in, The core tensor includes the main feature information of the data field; , , , These are the factor matrices for each mode, corresponding to feature extraction matrices of different dimensions in the data field.

[0008] Preferably, in step S4, the multiphysics data and spatial coordinates are input into the physical constraint neural network, the architecture of which is as follows: ; in, This is a physical model based on thermo-coupling equations, where x represents spatial coordinates. These are the parameters of the physical model; This is a neural network model, with spatial coordinates x and multiphysics tensors as inputs. , For mesh parameters; This is the physical constraint function.

[0009] Preferably, in step S4, the optimization model of the stochastic differential programming is: ; The Ito process constraints are satisfied as follows: ; in, To compensate for the displacement vector, The compensation amount for discrete points; For the cost function; To compensate for the second derivative norm of the path; For smoothing weighting factors; and These are the drift coefficient and diffusion coefficient of the compensation displacement, respectively; For Wiener process increments.

[0010] Preferably, in step S5, the coordinated control of laser softening and cutting force is based on the following model: ; ; in, Temperature-dependent yield strength; The yield strength at room temperature; The softening coefficient; This refers to the laser heating temperature. The glass transition temperature of the material; This refers to the softening temperature range. For cutting force; The material cutting coefficient, For the cutting area, For cutting speed, It is a speed-sensitive factor.

[0011] Preferably, in step S6, the iterative equation of the learning controller is: ; in, For the first The control parameter vector for each iteration (including parameters such as laser power and cutting speed); For the first The control parameter vector for the next iteration; , and These are PID parameters, used to adjust the control parameters' response to the current error, eliminate the system's static error, and suppress the system's overshoot, respectively. For the first Iterative wall thickness error; To reinforce the output of the learning module, based on the current state vector Provide optimized control strategies; For the first The state vector of the next iteration includes information such as current error, temperature field, stress field, and control parameters.

[0012] Preferably, in step S6, the convergence criterion is: ; in, For cavity volume, This is the actual wall thickness. For target wall thickness, This is the volume average error threshold, used to determine whether the volume average wall thickness error meets the requirements; This represents the maximum value of the wall thickness gradient, reflecting the degree of drastic change in wall thickness in space. This is the critical gradient threshold, used to determine whether the wall thickness gradient is within the allowable range; When both of the above conditions are met, the wall thickness accuracy is considered to have converged to the target range, and the processing process ends.

[0013] Preferably, the laser cladding uses a high-power fiber laser with a wavelength of 1064nm and a power adjustment range of 50-1000W; the cutting process uses a micro milling cutter with a diameter range of 0.1-2mm.

[0014] Therefore, the present invention provides a method for manufacturing a metal stator cavity with equal wall thickness using the above-described structure, which has the following beneficial effects: (1) By constructing and fusing a multi-physics data field, this invention comprehensively considers the influence of multiple physical field factors such as temperature field, strain rate field, and sound pressure field on the processing process, which can more accurately reflect the processing status and provide a basis for precise control. (2) The present invention uses physical constraint neural network and stochastic differential programming to predict and compensate for wall thickness deviation. By combining physical model and data-driven method, the prediction accuracy and the optimality of decision are improved, and the wall thickness accuracy is effectively controlled. (3) Based on the adaptive control of cavity curvature to control the heat input of laser cladding, and the coordinated regulation of laser softening and cutting force, this invention can adapt to the processing requirements of complex shapes, reduce the influence of thermal deformation and cutting force, and improve the quality and efficiency of the processed surface. (4) The present invention uses a learning-type iterative controller that combines PID control and reinforcement learning to realize closed-loop iterative optimization of the processing process, which can adapt to different processing conditions and improve processing stability and manufacturing accuracy.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of a method for manufacturing a metal stator cavity with uniform wall thickness according to the present invention. Figure 2 This is a schematic diagram of the structure of a method for manufacturing a metal stator cavity with equal wall thickness according to the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0019] Example like Figures 1-2 As shown, the present invention provides a method for manufacturing a metal stator cavity of uniform wall thickness, and the method is... Figure 2 The system shown is implemented by including the following steps: S1: Coordinate System Construction and Transformation: Construct a global coordinate system, mark reference points on the metal substrate, and establish the transformation relationship between the substrate coordinate system and the global coordinate system. Multiple sets of laser interferometers are deployed on the machining platform to construct the global coordinate system. Reference points are marked on the metal substrate surface. By measuring the coordinates of the reference points in both coordinate systems, the rotation matrix and translation vector are calculated, establishing the transformation relationship between the substrate coordinate system and the global coordinate system, providing a foundation for subsequent machining accuracy control. S2: Thermal Input Gradient Field Formation: Based on the cavity curvature, the laser power and scanning speed of the laser cladding process are adaptively controlled to form a thermal input gradient field. Curvature data for various parts of the cavity are acquired in real time. Based on the curvature changes, the laser power and scanning speed during the laser cladding process are calculated and adjusted according to the following formula: ; in, To scan the arc length of the path, For path curvature, For real-time temperature, and For reference power and speed, and These are material property coefficients; sigmoid is the activation function, and its expression is: This is used to smooth the power adjustment process. This control strategy matches the heat input of the laser cladding process with the cavity curvature, reducing thermal deformation and improving processing accuracy.

[0020] S3: Construction and Fusion of Multiphysics Data Field: Constructing and fusing a five-dimensional multiphysics data field including temperature field, strain rate field, sound pressure field, curvature and hardness distribution. Temperature field, strain rate field, and sound pressure field data are collected in real time using devices such as temperature sensors, strain measurement devices, and sound pressure sensors. Combined with curvature and hardness distribution data obtained during processing, a five-dimensional multiphysics data field is constructed, represented as follows: in, Indicates the temperature field distribution. Represents the strain rate field. The Laplace operator represents the sound pressure field. Indicates the curvature of the current processing area. Indicates the material hardness distribution; The data field is then fused using an improved Tucker decomposition, expressed by the following formula: ; in, The core tensor includes the main feature information of the data field; , , , Each mode has a factor matrix, corresponding to feature extraction matrices of different dimensions of the data field. Through data fusion, comprehensive information about the processing is obtained, providing a basis for subsequent decision-making.

[0021] S4: Wall Thickness Deviation Prediction and Toolpath Generation: A physical constraint neural network is used to predict wall thickness deviation. The optimal compensation decision is solved using stochastic differential programming to generate the toolpath. Multiphysics data and spatial coordinates are input into the physical constraint neural network, and its architecture is as follows: ; in, This is a physical model based on thermo-coupling equations, where x represents spatial coordinates. These are the parameters of the physical model; This is a neural network model, with spatial coordinates x and multiphysics tensors as inputs. , For mesh parameters; This is a physical constraint function. It is used to constrain the correction of neural networks to avoid violating physical laws.

[0022] Based on the predicted wall thickness deviation, the optimal compensation decision is solved using stochastic differential programming, and the optimization model is as follows: ; The Ito process constraints are satisfied as follows: ; in, To compensate for the displacement vector, The compensation amount for discrete points; For the cost function; To compensate for the second derivative norm of the path; For smoothing weighting factors; and These are the drift coefficient and diffusion coefficient of the compensation displacement, respectively; This is the Wiener process increment. Toolpaths are generated based on optimal compensation decisions to achieve precise control of the machining process.

[0023] S5: Photopressure Co-processing for Micromachining: This method utilizes laser preheating to soften the material, and coordinates the laser softening and cutting force for photopressure co-processing. Before cutting, the material is preheated using a laser, and the laser softening and cutting force are controlled according to the following formula: ; ; in, Temperature-dependent yield strength; The yield strength at room temperature; The softening coefficient; This refers to the laser heating temperature. The glass transition temperature of the material; This refers to the softening temperature range. For cutting force; The material cutting coefficient, For the cutting area, For cutting speed, It is a speed-sensitive factor. Through synergistic regulation, cutting forces are reduced, and the surface quality and machining efficiency are improved.

[0024] S6: Closed-loop iterative control: Establish a convergence criterion and utilize a learning-based iterative controller combining PID control and reinforcement learning to converge the wall thickness accuracy to the target range. The convergence criterion is: ; in, For cavity volume, This is the actual wall thickness. For target wall thickness, This is the volume average error threshold, used to determine whether the volume average wall thickness error meets the requirements; This represents the maximum value of the wall thickness gradient, reflecting the degree of drastic change in wall thickness in space. This is the critical gradient threshold, used to determine whether the wall thickness gradient is within the allowable range; When both of the above conditions are met, the wall thickness accuracy is considered to have converged to the target range, and the processing process ends.

[0025] The learning-based iterative controller updates the control parameters according to the following iterative equation: ; in, For the first The control parameter vector for each iteration (including parameters such as laser power and cutting speed); For the first The control parameter vector for the next iteration; , and These are PID parameters, used to adjust the control parameters' response to the current error, eliminate the system's static error, and suppress the system's overshoot, respectively. For the first Iterative wall thickness error; To reinforce the output of the learning module, based on the current state vector Provide optimized control strategies; For the first The state vector of the next iteration includes information such as current error, temperature field, stress field, and control parameters.

[0026] Specifically, taking the manufacturing of a thick-walled metal stator cavity for a certain type of aero-engine as an example, the manufacturing method of the present invention will be described in detail: S1. Coordinate System Construction and Transformation: Three sets of orthogonal laser interferometers are arranged on the machining platform to construct a global coordinate system, achieving a measurement accuracy of ±0.5μm / m. A reference hole is machined on the metal substrate surface. A ruby ​​probe is used to probe the reference hole, and the coordinates of the reference point in both the global and substrate coordinate systems are measured using the wavelength counting function of the laser interferometers. The rotation matrix and translation vector are calculated using the least squares method to establish the transformation relationship between the two coordinate systems, achieving a coordinate transformation accuracy of ±0.5μm.

[0027] S2. Thermal Input Gradient Field Formation: Acquire a 3D model of the cavity and calculate the curvature data of each part. During laser cladding, monitor the arc length of the scanning path in real time. Temperature measurement value Calculate and adjust the laser power according to the formula. and scanning speed Among them, the reference power =500w, base speed =10mm / s, material property coefficient =0.5, =0.3, material critical value =800℃, ambient temperature =25℃.

[0028] S3. Multiphysics Data Field Construction and Fusion: Temperature sensors, strain gauges, and acoustic pressure sensors are deployed in the processing area to collect temperature, strain rate, and acoustic pressure field data in real time. Curvature and hardness distribution data are acquired through online measurement equipment to construct a five-dimensional multiphysics data field. An improved Tucker decomposition is used for data fusion, and a core tensor is set. The dimensions and sizes of each factor matrix are used to extract data features.

[0029] S4. Wall Thickness Deviation Prediction and Toolpath Generation: Input multiphysics data and spatial coordinates into a physical constraint neural network to train network parameters. and Determine the physical constraint functions Based on the predicted wall thickness deviation, the optimal compensation decision is solved using stochastic differential programming, and a compensation cost function is set. Smoothing weight factor Parameters such as these are used to generate toolpaths based on optimal compensation decisions.

[0030] S5. Photo-pressure co-processing micromachining: Before cutting, the material is preheated using a laser, and the laser softening and cutting force are controlled according to a formula. The room temperature yield strength is among the parameters. =500MPa, softening coefficient =0.2, laser heating temperature The glass transition temperature of the material is set according to processing requirements. =600℃, softening temperature range =100℃, material cutting coefficient =2000, the cutting area is determined based on the tool and cutting parameters, and the cutting speed... =20mm / s, velocity sensitivity factor =0.1.

[0031] S6. Closed-loop iterative control: Set convergence criterion parameters and volume average error threshold. =0.01mm, critical gradient threshold =0.05mm / mm. Initialize the PID parameters of the learning iterative controller. =0.5, =0.1, =0.2. During the processing, the control parameters are updated according to the real-time measured wall thickness error and the iterative equation is used to continuously optimize the processing until the wall thickness accuracy meets the convergence criterion requirements. Through the above implementation process, the manufacturing method of the present invention has achieved high-precision manufacturing of the metal stator cavity with thick wall for this type of aero-engine, with wall thickness accuracy reaching ±5μm and surface roughness Ra≤0.8μm. The processing efficiency has been improved by more than 30% compared with the traditional method.

[0032] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for manufacturing a metal stator cavity of uniform wall thickness, characterized in that, Includes the following steps: S1. Arrange multiple sets of laser interferometers on the processing platform to construct a global coordinate system, mark reference points on the surface of the metal substrate, and establish the transformation relationship between the substrate coordinate system and the global coordinate system; S2. Based on the cavity curvature, the laser power and scanning speed during the laser cladding process are adaptively controlled to form a thermal input gradient field that matches the curvature distribution; S3. Construct a five-dimensional multiphysics data field that includes temperature field, strain rate field, sound pressure field, curvature and material hardness distribution, and perform data fusion. S4. Use a physical constraint neural network to predict wall thickness deviation, solve the optimal compensation decision based on stochastic differential programming and transform it into tool path; S5. By preheating and softening the material with laser, and synergistically controlling the laser softening and cutting force, optical pressure-coordinated micro-machining is performed. S6. Establish convergence criteria and adopt a learning-based iterative controller. Based on a closed-loop iterative mechanism combining PID control and reinforcement learning, the wall thickness accuracy gradually converges to the target range.

2. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S1, the construction of the coordinate system and the establishment of transformation relationships are specifically as follows: Multiple sets of laser interferometers are arranged on the processing platform to construct a global coordinate system. Reference points are marked on the surface of the metal substrate. By measuring the coordinates of the reference points in the two coordinate systems, the rotation matrix and translation vector are calculated, and the transformation relationship between the substrate coordinate system and the global coordinate system is established.

3. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S2, the specific method for dynamically adjusting the laser power and scanning speed is as follows: Laser power according to adjust; Scanning speed according to adjust; Where sigmoid is the activation function, and its expression is: ; To scan the arc length of the path, For path curvature, For real-time temperature, and For reference power and speed, and This represents the material property coefficient.

4. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S3, temperature field, strain rate field, and sound pressure field data are collected in real time. Combined with curvature and hardness distribution data obtained during processing, a five-dimensional multiphysics data field is constructed, which is represented as follows: ； in, Indicates the temperature field distribution. Represents the strain rate field. The Laplace operator represents the sound pressure field. Indicates the curvature of the current processing area. Indicates the material hardness distribution; The data field is then fused using an improved Tucker decomposition, expressed by the following formula: ； in, The core tensor includes the main feature information of the data field; , , , These are the factor matrices for each mode, corresponding to feature extraction matrices for different dimensions of the data field.

5. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S4, multiphysics data and spatial coordinates are input into a physical constraint neural network, the architecture of which is as follows: ； in, This is a physical model based on thermo-coupling equations, where x represents spatial coordinates. These are the parameters of the physical model; This is a neural network model, with spatial coordinates x and multiphysics tensors as inputs. , For mesh parameters; This is the physical constraint function.

6. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S4, the optimization model of the stochastic differential programming is: ； The Ito process constraints are satisfied as follows: ； in, To compensate for the displacement vector, The compensation amount for discrete points; For the cost function; To compensate for the second derivative norm of the path; For smoothing weighting factors; and These are the drift coefficient and diffusion coefficient of the compensation displacement, respectively; For Wiener process increments.

7. A method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S5, the coordinated control of laser softening and cutting force is based on the following model: ； ； in, Temperature-dependent yield strength; The yield strength at room temperature; The softening coefficient; This refers to the laser heating temperature. The glass transition temperature of the material; This refers to the softening temperature range. For cutting force; The material cutting coefficient, For cutting area, For cutting speed, It is a speed-sensitive factor.

8. The method for manufacturing a metal stator cavity with uniform wall thickness according to claim 1, characterized in that: In step S6, the convergence criterion is: ； in, For cavity volume, This is the actual wall thickness. For target wall thickness, This is the volume average error threshold, used to determine whether the volume average wall thickness error meets the requirements; This represents the maximum value of the wall thickness gradient, reflecting the degree of drastic change in wall thickness in space. This is the critical gradient threshold, used to determine whether the wall thickness gradient is within the allowable range; When both of the above conditions are met, the wall thickness accuracy is considered to have converged to the target range, and the processing process ends.

9. A method for manufacturing a metal stator cavity with uniform wall thickness according to claim 8, characterized in that: In step S6, the iterative equation of the learning controller is: ； in, For the first The control parameter vector for the next iteration; For the first The control parameter vector for the next iteration; , and These are PID parameters, used to adjust the control parameters' response to the current error, eliminate the system's static error, and suppress the system's overshoot, respectively. For the first Iterative wall thickness error; To reinforce the output of the learning module, based on the current state vector Provide optimized control strategies; For the first The state vector of the next iteration.

10. A method for manufacturing a metal stator cavity with uniform wall thickness according to any one of claims 1-9, characterized in that: The laser cladding uses a high-power fiber laser with a wavelength of 1064nm and a power adjustment range of 50-1000W; the cutting process uses a micro milling cutter with a diameter range of 0.1-2mm.

Images