Assembly deviation prediction method considering surface macro and micro characteristics and assembly deformation

Abstract

The invention provides an assembly deviation prediction method considering surface macro and micro characteristics and assembly deformation. Comprising the steps of part surface macro-micro feature representation based on primary function modal superposition, non-ideal surface model construction considering an actual machining surface, and non-ideal surface contact state calculation based on a difference plane and point cloud registration method. And solving non-ideal surface contact deformation based on a conjugate gradient-fast Fourier algorithm and establishing an assembly deviation transfer model considering surface macro and micro characteristics and assembly deformation. According to the method, surface form and position errors generated by part machining errors and assembly deformation errors caused by loads in the assembly process are fully considered, the deviation domain model and the Minkowski summation method are combined to simulate transmission of the assembly deviation, the prediction precision of the product assembly deviation is effectively improved, and meanwhile the operation speed is remarkably increased; and real-time monitoring and timely deviation adjustment in the assembling process are achieved.

Description

Technical Field

[0001] The present invention belongs to the research field of precision assembly technology, and specifically relates to an assembly deviation prediction method that takes into account surface macro and micro features and assembly deformation. Background Art

[0002] Assembly, the final step in complex product manufacturing, is a crucial indicator of product quality and performance. With the increasing demand for high-quality, high-quality complex equipment in high-end manufacturing fields like aviation and aerospace, the accuracy of product assembly precision analysis is becoming increasingly demanding. However, the complex assembly processes, numerous sensitive factors, and poor traceability of complex products lead to the continuous propagation and accumulation of assembly deviations, making precise quantification and control difficult. This leads to poor assembly accuracy and low first-pass yields, severely impacting product performance. Therefore, accurately predicting assembly deviations is crucial.

[0003] Assembly deviation prediction primarily involves the construction of deviation models and the study of assembly deviation propagation. Currently, assembly deviation calculations are often performed under ideal contact surface conditions using extreme value or statistical methods. However, component manufacturing errors, assembly loads, and contact deformation caused by positioning and tightening all have a certain impact on assembly deviation analysis results. Therefore, modeling the geometric and positional errors of the assembly interface and accurately calculating contact deformation are key to accurately constructing deviation prediction models.

[0004] Finite element analysis (FEA) is the most commonly used method for calculating load-induced assembly contact deformation. However, its high computational cost and complex pre-processing make it difficult to meet the demand for real-time prediction of assembly process deviations, which can extend product production cycles and cause economic losses. To address these issues, this paper uses a basis function modal superposition method to simulate the macro- and microscopic surface features of components containing actual processing information. A boundary element algorithm is proposed to accurately and rapidly calculate assembly contact deformation. Based on a deviation domain model, this method achieves precise prediction of assembly deviations that considers both surface macro- and microscopic features and assembly deformation. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems of complex products caused by complex assembly processes, multiple sensitive factors and poor traceability, such as the continuous propagation and accumulation of assembly deviations, the difficulty in precise quantification of control, and the resulting poor assembly accuracy and low first-time assembly pass rate. To this end, the present invention proposes an assembly deviation prediction method that considers surface macro and micro features and assembly deformation, including the following steps:

[0006] S1: The basis function modal superposition method is used to characterize the macroscopic and microscopic features of the surfaces of two typical parts: plane and cylindrical surfaces;

[0007] S2: Correlate the basis function coefficients with the actual machining surface roughness parameters to generate a non-ideal surface containing actual machining information;

[0008] S3: Calculate the initial contact points between the constructed planes and cylindrical surfaces containing macro and micro features by using the difference plane method and point cloud registration method respectively;

[0009] S4: Based on the conjugate gradient-fast Fourier transform algorithm, the contact deformation of the corresponding discrete points between two non-ideal contact surfaces under load is solved, the contact deformation surface is fitted, and the small displacement spin is calculated;

[0010] S5: Establish an assembly deviation transmission path that considers the serial-parallel connection of assembly dimensions, gradually calculate the small displacement spins of the geometric functional characteristics in the path and establish a deviation domain model based on this, simulate the transmission of assembly deviations based on the Minkowski summation method, and construct an assembly deviation prediction model.

[0011] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S1 specifically includes:

[0012] S11: The kernel function of discrete cosine transform is used to characterize the macro and micro features of the plane. The expression of the kernel function of discrete cosine transform is:

[0013]

[0014] Where: M, N represent the number of discrete points in the length and width directions of the rectangular plane respectively, m, n represent the sampling point numbers in the length and width directions of the plane respectively, u, v represent the sampling frequencies in the length and width directions respectively, and m, u = 0, 1, ..., M-1, n, v = 0, 1, ..., N-1;

[0015] S12: The Legendre-Fourier function is used to characterize the macroscopic and microscopic features of the cylindrical surface. The expression of the Legendre-Fourier function is:

[0016]

[0017] Where z∈[-1,1], θ∈[-π,π], i is the Fourier series, j is the polynomial coefficient, P j (z) represents the order of Legendre polynomial.

[0018] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S2 specifically includes:

[0019] S21: Obtain the roughness parameters of the generated surface under different combinations of basis function coefficients through numerical simulation, and establish a database of the relationship between the skewness Ssk and kurtosis Sku of the generated surface under different combinations of basis function coefficients;

[0020] S22: Divide the training set, test set and validation set, and apply the random forest algorithm to establish the relationship model between basis function coefficients and roughness parameters;

[0021] S23: Select a corresponding basis function coefficient combination according to the surface roughness coefficient of the part to generate a non-ideal surface containing actual processing information.

[0022] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S3 specifically includes:

[0023] S31: The contact between two non-ideal surfaces is equivalent to the contact between an ideal surface and a non-ideal difference plane;

[0024] S32: Traverse all the minimum points on the equivalent difference plane and calculate the three initial contact points of the two equivalent planes;

[0025] S33: Calculate the small displacement rotation of the plane feature where the initial contact point is located;

[0026] S34: searching for the point with the minimum distance between the target cylindrical surface point cloud and the source cylindrical surface point cloud based on an iterative closest point algorithm, and determining the corresponding relationship between the two groups of point clouds;

[0027] S35: Calculate the optimal transformation matrix between two sets of point clouds using a sequential quadratic programming algorithm;

[0028] S36: Calculate the small displacement spin of two non-ideal cylindrical features after point cloud registration.

[0029] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S4 specifically includes:

[0030] S41: Measure the pressure distribution between two planes and calculate the deformation of corresponding discrete points on the two equivalent planes using fast Fourier transform;

[0031] S42: Using the conjugate gradient algorithm to iteratively obtain the contact pressure and gap of the corresponding discrete points on the two equivalent planes;

[0032] S43: Apply the least squares method to fit the discrete points of the equivalent non-ideal difference plane after the assembly deformation calculation, and obtain the small displacement spinor considering the macro and micro features of the two planes and the assembly deformation error;

[0033] S44: The contact of the non-ideal cylindrical surface features obtained by simulation is converted into the contact of two arc surfaces after being unfolded along the axis, and S41 to S43 are repeated to obtain a small displacement rotation considering the macro and micro features of the two cylindrical surfaces and the assembly deformation error.

[0034] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S41 specifically includes:

[0035] S411: Use pressure film to measure the pressure distribution between contact surfaces, and obtain a pressure matrix with twice the number of discrete points by zero padding;

[0036] S412: Obtaining the influence coefficient matrix K in the doubled calculation area;

[0037]

[0038] Where x m =x i +Δx / 2,x p =x i -Δx / 2,y m =y j +Δy / 2,y p =y j -Δy / 2. Δx and Δy are the grid sizes along the x and y directions respectively;

[0039] S413: Calculate the Fourier transform matrices of the pressure matrix and the influence coefficient matrix respectively, multiply the corresponding items of the two and perform inverse Fourier transform to obtain the deformation of the contact of the corresponding discrete points on the two equivalent planes.

[0040] According to the above aspects and any possible implementation, a further implementation is provided, wherein step S5 specifically includes:

[0041] S51: Analyze the geometric functional characteristics related to assembly, determine the assembly functional requirements that need to be guaranteed, and establish the assembly deviation transmission path considering the assembly dimension chain-parallel characteristics;

[0042] S52: Calculate the range of small displacement screw values ​​of different types of geometric feature pairs in the assembly deviation transmission path, and establish the deviation domain model of the corresponding geometric features;

[0043] S53: For the series and parallel situations in the assembly transfer path, the Minkowski summation and intersection methods are respectively applied to combine the deviation domains to calculate the cumulative deviation value of the corresponding assembly deviation transfer path.

[0044] The present invention provides an assembly deviation prediction method that considers surface macro and micro features and assembly deformation. Compared with the prior art, it has at least the following beneficial effects:

[0045] The embodiment of the present invention is aimed at the prediction of assembly deviation of complex products. It uses the basis function modal superposition method to characterize the macro-micro features of two typical parts, planes and cylinders, to generate non-ideal surfaces containing actual processing information. The initial contact points between the planes and cylinders containing macro-micro features are calculated by the difference plane method and the point cloud registration method respectively. The contact deformation of the corresponding discrete points between the two non-ideal contact surfaces under the action of the load is solved based on the conjugate gradient-fast Fourier algorithm. The contact state between the two features is accurately solved while improving the efficiency of assembly deformation calculation. The assembly deviation prediction model is constructed based on the deviation domain model and the Minkowski summation method. The influence of the macro-micro features of the part surface and the assembly deformation is comprehensively considered, and the accumulation of assembly deviation is solved by geometric calculation. The technical solution of the present invention associates the basis function coefficients with the actual processing surface roughness parameters, effectively improving the prediction accuracy of product assembly deviation while significantly improving the calculation speed. In the assembly process, the assembly deviation is adjusted in time according to the real-time prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the assembly deviation prediction method considering surface macro and micro features and assembly deformation in step S1 of the present invention;

[0047] Figure 2 It is a partial plane macro-micro feature map represented by the kernel function of the discrete cosine transform in step S2 of the present invention;

[0048] Figure 3 It is a macro-micro feature map of a portion of the cylindrical surface characterized by the Legendre-Fourier function in step S2 of the present invention;

[0049] Figure 4 is a corresponding relationship diagram between the basis function coefficient combination and the roughness coefficient Sku in step S3 of the present invention;

[0050] Figure 5 The exploded view of the assembly and the assembly constraint feature diagram in step S5 of the present invention;

[0051] Figure 6 is the assembly deviation transfer path diagram of the assembly in step S5 of the present invention;

[0052] Figure 7 is the value range of the corresponding small displacement rotation of each assembly functional element of the assembly in step S5 of the present invention;

[0053] Figure 8 is the assembly deviation domain cumulative calculation diagram in step S5 of the present invention;

[0054] Figure 9 It is the assembly accuracy required by the assembly function in step S5 of the present invention. DETAILED DESCRIPTION

[0055] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] The flowchart of the assembly deviation prediction method considering surface macro-micro features and assembly deformation provided in this embodiment is as follows: Figure 1 As shown, the following steps are included:

[0057] S1: The basis function modal superposition method is used to characterize the macroscopic and microscopic features of the surfaces of two typical parts: plane and cylindrical surfaces;

[0058] S2: Correlate the basis function coefficients with the actual machining surface roughness parameters to generate a non-ideal surface containing actual machining information;

[0059] S3: Calculate the initial contact points between the constructed planes and cylindrical surfaces containing macro and micro features by using the difference plane method and point cloud registration method respectively;

[0060] S4: Based on the conjugate gradient-fast Fourier transform algorithm, the contact deformation of the corresponding discrete points between two non-ideal contact surfaces under load is solved, the contact deformation surface is fitted, and the small displacement spin is calculated;

[0061] S5: Establish an assembly deviation transmission path that considers the serial-parallel connection of assembly dimensions, gradually calculate the small displacement spins of the geometric functional characteristics in the path and establish a deviation domain model based on this, simulate the transmission of assembly deviations based on the Minkowski summation method, and construct an assembly deviation prediction model.

[0062] In a specific embodiment, the S1 specifically includes:

[0063] The kernel function of discrete cosine transform is used to characterize the macro and micro features of the plane. The expression of the kernel function of discrete cosine transform is:

[0064]

[0065] Where: M, N represent the number of discrete points in the length and width directions of the rectangular plane respectively, m, n represent the sampling point numbers in the length and width directions of the plane respectively, u, v represent the sampling frequencies in the length and width directions respectively, and m, u = 0, 1, ..., M-1, n, v = 0, 1, ..., N-1.

[0066] Some of the generated plane macro and micro features are as follows Figure 2 As shown in the figure, the first 16 groups of planar macro-micro features obtained by simulation when different combinations of coefficients u and v are given.

[0067] Furthermore, the Legendre-Fourier function is used to characterize the macroscopic and microscopic features of the cylindrical surface. The expression of the Legendre-Fourier function is:

[0068]

[0069] Where z∈[-1,1], θ∈[-π,π], i is the Fourier series, j is the polynomial coefficient, P j (z) represents the order of the Legendre polynomial, which is expressed in the form of Rodrigues' formula as:

[0070]

[0071] The Legendre polynomials in this form {P j (z)}(j=0,1,2,…) is an orthogonal function group. By changing j, different axial profiles of cylindrical surface features are described. By changing j, different radial profiles of cylindrical surface features are described. ij With B ij is the influence coefficient, expressed as follows:

[0072]

[0073] Where k is the number of axial sampling points, l is the number of radial sampling points, r kl is the distance from each sampling point on the cylinder to the axis, z k is the coordinate of the axial sampling point, θ l is the radial distance between adjacent sampling points.

[0074] The generated macro and micro features of some cylindrical surfaces are as follows Figure 3 As shown in the figure, the first 16 groups of macro- and micro-features of cylindrical surfaces obtained by simulation when different combinations of coefficients i and j are given.

[0075] In a specific embodiment, the S2 specifically includes:

[0076] Numerical simulations were used to obtain the skewness Ssk and kurtosis Sku of the generated surfaces for different combinations of basis function coefficients. When Ssk is 0 and Sku is 3, the surface features exhibit a Gaussian distribution. The surfaces generated for different values ​​of the coefficients u and v∈[0,9] were then superimposed on each other to establish a dataset correlating the basis function coefficients with the generated surfaces Ssk and Sku.

[0077] Furthermore, 70% of the data was used as the training set of the model, and the remaining 30% of the data was divided into a test set and a validation set in a 1:1 ratio. The random forest algorithm was applied to establish a relationship model between the coefficients u1, v1, u2, v2 and the Ssk and Sku values ​​of the superimposed surface, and the basis function coefficient combination corresponding to the non-ideal surface within the specific Ssk and Sku value range was determined, and the "Gini index" was used as the partitioning attribute of the decision tree.

[0078] Typically, non-ideal surfaces produced by turning and shaping processes have a positive distribution, while grinding, honing, and milling processes produce non-ideal surfaces with high kurtosis and negative distribution. Therefore, the Ssk and Sku values ​​are divided into four intervals with cutoff points of -0.1, 0.1, 1, and 2.5, 3, and 3.5, respectively. Based on the Ssk and Sku value intervals that produce non-ideal surfaces in different machining methods, the corresponding relationship between the basis function coefficients and the corresponding machined surface topography is revealed. The established random forest prediction model is then used to select reasonable combinations of basis function coefficients, thereby more realistically simulating the macro- and microscopic features of different machined surfaces.

[0079] Furthermore, the corresponding basis function coefficient combination is selected according to the surface roughness coefficient of the part to generate a non-ideal surface containing actual processing information. Figure 4 It is a binary tree visualization expression of the Sku classification model, and the corresponding prediction paths are given when u1=1, v1=4, u2=6, and v2=7. At this time, the generated surface height distribution is relatively symmetrical and the surface features are relatively flat. In addition to the path given in the figure, the other prediction paths marked as "1" can also generate non-Gaussian surfaces with this surface feature, thereby characterizing non-ideal surfaces containing actual processing information such as shape and position errors and roughness.

[0080] In a specific embodiment, the S3 specifically includes:

[0081] The contact between two non-ideal surfaces containing actual machining features generated by the superposition of basis function modes is equivalent to the contact between an ideal surface and a non-ideal difference plane.

[0082] Next, traverse all the minimum points on the equal-price-difference plane and determine whether any point on the plane formed by three minimum points is less than the height value of the corresponding discrete point on the equal-price-difference plane, and whether the intersection of the assembly force line and the plane is within the triangle formed by the three minimum points. If these conditions are met, these three points are used as the initial contact points between the equal-price-difference plane and the ideal plane.

[0083] Furthermore, the small displacement spinor of the equivalent non-ideal difference plane feature at the initial contact point is calculated, and the expression is:

[0084] SDT=[u,v,w,α,β,γ] T

[0085] Where u, v, and w represent the small displacements along the x, y, and z directions, respectively; α, β, and γ represent the small rotations along the x, y, and z directions, respectively.

[0086] Furthermore, based on the iterative closest point algorithm, the point with the minimum distance between the target cylindrical point cloud and the source cylindrical point cloud is searched to determine the point-to-point correspondence between the two sets of point clouds. Under the premise of ensuring that the two sets of matched point clouds do not interfere with each other, the sequential quadratic programming algorithm is used to iteratively optimize the following formula:

[0087]

[0088] |R·p i +Tq i |·n i ≥0

[0089] Where p i is the source point cloud, q i is the target point cloud set, and n is the unit vector in the direction of the assembly force line. After obtaining the optimal rigid transformation matrix, the small displacement spinor considering the characteristics of the two non-ideal cylindrical surfaces is calculated.

[0090] In view of the different characteristics of planar features and cylindrical features, this method uses the difference plane method and point cloud registration to obtain the initial contact state of part features, and accurately obtains the initial contact points containing actual processing information before calculating the assembly deformation.

[0091] In a specific embodiment, the S4 specifically includes:

[0092] The pressure distribution between the contact surfaces is measured using a pressure film, and the deformation of the corresponding discrete points on the two equivalent planes is calculated using fast Fourier transform. The contact deformation expression is:

[0093]

[0094] Where p k,l is the pressure at the node (k, l), K i-k,j-l is the pressure-displacement influence coefficient.

[0095] Specifically, the pressure matrix p' with twice the number of discrete points is obtained by zero padding, and the influence coefficient matrix K is obtained in the twice calculation area, which is expressed as:

[0096]

[0097] Where: x m =x i +Δx / 2,x p =x i -Δx / 2,y m =y j +Δy / 2,y p =y j -Δy / 2. Δx and Δy are the grid sizes along the x and y directions respectively, E * is the composite elastic modulus.

[0098] The Fourier transform matrices of the pressure matrix and the influence coefficient matrix are calculated respectively, the corresponding items of the two are multiplied and the inverse Fourier transform is performed to obtain the deformation matrix in the time domain and the contact deformation of the corresponding discrete points on the two equivalent planes.

[0099] Furthermore, the gap after assembly deformation calculation is obtained by adding the contact deformation of the corresponding discrete points of the two equivalent planes to the initial gap. Based on the variational method and the principle of minimum potential energy, the conjugate gradient algorithm is used to iteratively calculate the minimum value of the expression W(p), which is:

[0100]

[0101] Where c is the contact gap matrix between the two surfaces, and p is the contact pressure matrix between the two surfaces.

[0102] Furthermore, the least squares method was applied to fit the discrete points of the equivalent non-ideal difference plane after the assembly deformation calculation, resulting in a small displacement spin that takes into account the macro- and microscopic features of the two planes and the assembly deformation errors. Simultaneously, the contact of the simulated non-ideal cylindrical surface features was transformed into the contact of two arc surfaces after unfolding along the axis. The calculation method for the contact deformation of the corresponding discrete points of the two plane features was repeated to obtain a small displacement spin that takes into account the macro- and microscopic features of the two cylindrical surfaces and the assembly deformation errors.

[0103] This method discretizes the problem of solving assembly contact deformation and accelerates the calculation efficiency through the boundary element algorithm of conjugate gradient-fast Fourier algorithm, providing the premise for timely feedback of assembly process information.

[0104] In a specific embodiment, the S5 specifically includes:

[0105] like Figure 5 As shown in the figure, analysis of the geometric functional features related to assembly shows that the three parts contain a total of 11 geometric functional features.

[0106] Furthermore, an assembly deviation transmission path considering the assembly dimension chain-parallel characteristics is established, such as Figure 6 As shown in the figure, there are 2 groups of internal functional element pairs IFEP, 5 groups of series functional element pairs SFEP, 1 group of parallel functional element pairs PFEP and the assembly functional requirements FR that need to be guaranteed. Four assembly deviation transmission paths can be established, namely: 1.0→1.1→2.1→2.5→3.1, 1.0→1.2→2.2→2.5→3.1, 1.0→1.3→2.3→2.5→3.1, and 1.0→2.4→2.5→3.1.

[0107] Furthermore, using the small displacement spin calculation method given in S1 to S4 considering the macro and micro features of the two contact surfaces and the assembly deformation error, 1000 Monte Carlo simulations are performed on each set of geometric feature pairs in the assembly deviation transmission path, and the value range of the small displacement spin of different types of assembly functional element pairs in the assembly deviation transmission path is calculated, as shown in Figure 2. Figure 7 As shown in the figure, taking IFEP1 as an example, it represents the numerical simulation results of the small displacement in the z direction and the small rotation in the x and y directions corresponding to the assembly feature GFE1.0 / 1.1. It can be seen that the constraint ranges of z, α, and β are (-0.0114, 0.0176), (-3.7025e-3, 3.9364e-3), (-3.6737e-3, 9.0287e-4), and the quartiles are (-6.579e-3, 1.042e-2), (-1.791e-3, 2.046e-3), and (-6.403e-3, 1.376e-3), respectively.

[0108] Furthermore, the obtained small displacement spinor is represented by the vertex values ​​in the convex hull model to form a deviation space to represent all possible changes of the target feature within the tolerance range, and the deviation domain model of the corresponding geometric feature is constructed. For the series and parallel situations in the assembly transfer path, the Minkowski summation and intersection methods are applied to combine the deviation domains and calculate the cumulative deviation value of the corresponding assembly deviation transfer path, as shown in the figure. Figure 8 As shown, the expression is:

[0109]

[0110] Where, represents the Minkowski summation operation, ∩ represents the Minkowski intersection operation, and D represents the deviation domain of the corresponding functional element.

[0111] Finally, the assembly accuracy in the z direction is (-0.0760, 0.0843), as shown in Figure 9 shown.

[0112] This method comprehensively considers the influence of macro and micro features of part surface and assembly deformation, solves the accumulation of assembly deviation through geometric calculation, effectively improves the prediction accuracy of product assembly deviation while significantly improving the calculation speed, and realizes timely adjustment of assembly deviation according to real-time prediction results during the assembly process.

[0113] The above drawings and explanations are only a specific embodiment of the present invention, but the specific protection scope of the present invention is not limited to the above explanations. Any technical ideas disclosed in the present invention and any simple replacement or change of the technical solution of the present invention should be within the scope of protection of the present invention.

Claims

1. A method for predicting assembly deviation considering macro and micro features of joint surfaces and assembly deformation, characterized in that: The following steps are involved: S1: The basis function modal superposition method is used to characterize the macroscopic and microscopic features of the surfaces of two typical parts: plane and cylindrical surfaces; S2: Correlate the basis function coefficients with the actual machining surface roughness parameters to generate a non-ideal surface containing actual machining information; S3: Calculate the initial contact points between the constructed planes and cylindrical surfaces containing macro and micro features by using the difference plane method and point cloud registration method; S4: Based on the conjugate gradient-fast Fourier algorithm, the contact deformation of the corresponding discrete points between two non-ideal contact surfaces under load is solved, the contact deformation surface is fitted and the small displacement spinor is calculated; S5: Establish an assembly deviation transmission path considering the serial-parallel connection of assembly dimensions, gradually calculate the small displacement spins of the geometric functional characteristics in the path and establish a deviation domain model based on this, simulate the transmission of assembly deviations based on the Minkowski summation method, and construct an assembly deviation prediction model.

2. The assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 1 is characterized in that: The step S1 specifically includes: S11: Using the kernel function of discrete cosine transform to characterize the macro and micro features of the plane; S12: Use Legendre-Fourier function to characterize the macro and micro features of cylindrical surfaces.

3. The assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 1 is characterized in that: The step S2 specifically includes: S21: obtaining roughness parameters of the generated surface under different basis function coefficient combinations through numerical simulation and establishing a data set; S22: Divide the training set, test set and validation set, and apply the random forest algorithm to establish the relationship model between basis function coefficients and roughness parameters; S23: Select a corresponding basis function coefficient combination according to the surface roughness coefficient of the part to generate a non-ideal surface containing actual processing information.

4. The assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 1 is characterized in that: The step S3 specifically includes: S31: Equivalent the contact between two non-ideal surfaces to the contact between an ideal surface and a non-ideal difference plane; S32: traverse all the minimum points on the equivalent difference plane and calculate the three initial contact points of the two equivalent planes; S33: Calculate the small displacement spin of the plane feature where the initial contact point is located; S34: searching for the point with the minimum distance between the target cylindrical surface point cloud and the source cylindrical surface point cloud based on an iterative closest point algorithm, and determining the corresponding relationship between the two groups of point clouds; S35: using a sequential quadratic programming algorithm to calculate the optimal transformation matrix between two sets of point clouds; S36: Calculate the small displacement spin of the two non-ideal cylindrical features after point cloud registration.

5. The assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 1 is characterized in that: The step S4 specifically includes: S41: Measure the pressure distribution between two planes and calculate the deformation of corresponding discrete points on two equivalent planes using fast Fourier transform; S42: using the conjugate gradient algorithm to iteratively obtain the contact pressure and gap of the corresponding discrete points on two equivalent planes; S43: Apply the least squares method to fit the discrete points of the equivalent non-ideal difference plane after the assembly deformation calculation, and obtain the small displacement spinor considering the macro and micro features of the two planes and the assembly deformation error; S44: The contact of the non-ideal cylindrical surface features obtained by simulation is converted into the contact of two arc surfaces after being unfolded along the axis, and S41 to S43 are repeated to obtain a small displacement rotation factor that takes into account the macro and micro features of the two cylindrical surfaces and the assembly deformation error.

6. According to the assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 5, the step S41 specifically comprises: S411: Use the pressure film to measure the pressure distribution between the contact surfaces, and obtain a pressure matrix with twice the number of discrete points by zero filling; S412: Obtain the influence coefficient matrix K in the double calculation area: In the formula, x m =x i +Δx / 2,x p =x i -Δx / 2,y m =y j +Δy / 2,y p =y j -Δy / 2. Δx and Δy are the grid sizes along the x and y directions respectively; S413: Calculate the Fourier transform matrices of the pressure matrix and the influence coefficient matrix respectively, multiply the corresponding items of the two and perform inverse Fourier transform to obtain the deformation of the contact of the corresponding discrete points of the two equivalent planes.

7. The assembly deviation prediction method considering surface macro and micro features and assembly deformation according to claim 1 is characterized in that: The S5 specifically includes: S51: Analyze the geometric functional elements related to assembly, determine the assembly functional requirements that need to be guaranteed, and establish the assembly deviation transmission path considering the chain-parallel characteristics of assembly dimensions; S52: Calculate the value range of small displacement screw of different types of geometric feature pairs in the assembly deviation transmission path, and establish the deviation domain model of the corresponding geometric feature; S53: For the series-parallel situation in the assembly transfer path, the Minkowski summation method and the intersection method are respectively applied to combine the deviation domains to calculate the cumulative deviation value of the corresponding assembly deviation transfer path.

Images