Blade measurement method and system based on three-axis measurement and MBD model fusion

Abstract

The invention discloses a blade measurement method and system based on three-axis measurement and MBD model fusion, and the method comprises the steps: carrying out the preliminary measurement of a blade through a three-axis measurement instrument, fitting the blade into a first NURBS curved surface, obtaining the actual normal vector of the blade based on the first NURBS curved surface, and adjusting the posture of the three-axis measurement instrument through the actual normal vector; constructing a convolutional neural network, training the convolutional neural network based on the MBD model to obtain a blade curved surface measurement point prediction model, and planning measurement points by using the blade curved surface measurement point prediction model; calculating a deviation comprehensive index, adjusting the measurement speed and the distance between the measurement points of the three-axis measurement instrument based on the deviation comprehensive index, calculating the similarity between the measurement points, and feeding back the technological parameters based on the similarity. According to the invention, the problems of low measurement precision, low efficiency, poor fusion effect of measurement data and the MBD model and difficulty in batch measurement in the existing blade measurement technology are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of blade measurement, and particularly relates to a blade measurement method and system based on the fusion of three-axis measurement and MBD model. Background Art

[0002] In the field of aero-engine blade measurement, the traditional three-axis measurement technology has the following problems: (1) Insufficient intelligence in measurement path: When measuring complex curved surfaces, the distribution of measurement points is unreasonable and the accuracy is low; (2) Poor data fusion effect: The measurement data is independent of the MBD model, and only simple coordinate matching is performed, without exploring deep information such as surface normal vectors and feature correlations; (3) Low efficiency in batch measurement: Lack of a multi-blade synchronous measurement mechanism, and sequential measurement leads to excessive time consumption; (4) Weak quality control ability: Dependence on manual setting of tolerances, lack of intelligent analysis, and difficulty in predicting quality trends.

[0003] Although existing technologies have attempted to combine MBD models with multi-device measurements, there are still problems such as complex device integration, simple algorithms, and insufficient efficiency, making it difficult to meet the requirements of modern industries for high precision, high efficiency, and intelligence. For example, in "He Xiaofeng. Research and Development of a Blade CMM Detection Path Planning System Based on MBD [D]. Jiangsu University [2025-02-19]": This literature constructs a blade detection system based on three-axis measurement and MBD. In terms of hardware composition, it mainly relies on a three-axis measuring instrument. By precisely controlling the movement of the measuring head in the X, Y, and Z directions, the blade surface is measured point by point. During the measurement process, according to the pre-set measurement path, the measuring head moves on the blade surface to collect the three-dimensional coordinate data of each point on the blade surface. At the software level, first, the MBD model is parsed to extract key information such as the theoretical dimensions, shape, and tolerances of the blade, and a theoretical data model is constructed. Then, the actual measurement data collected by the three-axis measuring instrument is compared with the theoretical data in the MBD model. In terms of the data comparison algorithm, a method based on coordinate matching is adopted. By calculating the coordinate deviation between the measurement point and the corresponding position point in the theoretical model, the compliance degree of the actual manufacturing of the blade with the design requirements is judged. For example, when measuring the surface profile of the blade, a series of measured surface point coordinates are compared one by one with the theoretical surface coordinates in the MBD model, and the coordinate difference is calculated to evaluate the machining accuracy of the blade surface. Although this technology compares the measurement data with the MBD model, it only stays at the simple coordinate matching level and does not fully explore the fusion value of other rich information in the measurement data and the MBD model (such as surface normal vectors, feature correlations, etc.). This makes the analysis of blade deviations relatively one-sided, unable to comprehensively and accurately evaluate the manufacturing quality of the blade, and difficult to effectively guide the improvement of production processes. When measuring multiple blades in batches, due to the lack of an effective synchronous measurement mechanism and an efficient data management strategy, each blade needs to be measured and data processed sequentially, and the measurement process takes a long time, unable to meet the requirements of rapid blade detection in large-scale production, severely restricting the improvement of production efficiency. Summary of the Invention

[0004] The present invention aims to solve the deficiencies of the existing technology and provides the following solutions:

[0005] A blade measurement method based on the fusion of three-axis measurement and MBD model, comprising the following steps:

[0006] The blade is preliminarily measured by a three-axis measuring instrument and fitted into a first NURBS surface. Based on the first NURBS surface, the actual normal vector of the blade is obtained, and the attitude of the three-axis measuring instrument is adjusted using the actual normal vector;

[0007] A convolutional neural network is constructed, and the convolutional neural network is trained based on the MBD model to obtain a blade surface measurement point prediction model, and the measurement points are planned using the blade surface measurement point prediction model;

[0008] Calculate the comprehensive deviation index, adjust the measurement speed and measurement point spacing of the three-axis measuring instrument based on the comprehensive deviation index, calculate the similarity between measurement points, and perform process parameter feedback based on the similarity.

[0009] Preferably, the actual normal vector includes:

[0010]

[0011] Wherein, S 实际 represents the actual surface curve, and u and v represent the horizontal and vertical coordinates of the parameter points.

[0012] Preferably, the method for adjusting the attitude of the three-axis measuring instrument using the actual normal vector includes:

[0013] Calculate the corrected normal vector:

[0014] n 修正 = n 理论 + Δn

[0015] Δn = α·(n 理论 - n 实际 ) + β·(κ 理论 - κ 实际 )·t

[0016] Wherein, n 修正 represents the corrected normal vector, Δn represents the normal vector deviation, n 理论 represents the theoretical normal vector calculated from the MBD model, κ 理论 represents the theoretical local curvature provided by the MBD model, κ 实际 represents the actual local curvature calculated by fitting the actual surface, α represents the normal vector compensation weight, β represents the curvature difference compensation coefficient, and t represents the surface tangent vector;

[0017] Adjust the attitude of the three-axis measuring instrument based on the corrected normal vector.

[0018] Preferably, the method for training the convolutional neural network includes:

[0019] Parse the STEP file of the MBD model into NURBS surfaces, discretize them into point cloud data, and generate a two-dimensional grayscale image through orthogonal projection;

[0020] Filter and denoise the two-dimensional grayscale image, and extract the surface boundary features through an edge detection algorithm;

[0021] Input the denoised image, the extracted boundary features, and the corresponding historical measurement data into the convolutional neural network for training to obtain the blade surface measurement point prediction model.

[0022] Preferably, the method for adjusting the measurement speed and measurement point spacing of a three-axis measuring instrument includes:

[0023] Calculating the comprehensive deviation index:

[0024]

[0025] where δ x represents the coordinate deviation of the measurement point in the x direction, δ y represents the coordinate deviation of the measurement point in the y direction, δ z represents the coordinate deviation of the measurement point in the z direction, γ represents the local torsion angle of the curved surface, and κ represents the local curvature difference weight coefficient;

[0026] If δ 综合 > 0.1 mm, the measurement speed v is reduced to 70% of the original value, and the measurement point spacing d is adjusted to d 新 = d 原 × 0.5;

[0027] If δ 综合 > 0.05 mm, the measurement speed v is increased to 120% of the original value, and the measurement point spacing d is adjusted to d 新 = d 原 × 1.2.

[0028] Preferably, the method for performing process parameter feedback includes:

[0029] For the measurement point p i , calculate the curvature covariance matrix C within its 10 × 10 neighborhood, with eigenvalues λ1 and λ2, and construct the eigenvector F = [λ1, λ2, n];

[0030] Calculating the similarity based on the eigenvector:

[0031]

[0032] where F 实测 represents the measured eigenvector, F 理论 represents the theoretical eigenvector, and w represents the weight vector;

[0033] If s < 0.85, it is determined that the matching is abnormal, an alarm signal is generated, the measurement of the current batch is paused, an adjustment instruction is sent to the production system through the OPCUA protocol, and the abnormal data is recorded in the SQL database.

[0034] The present invention also provides a blade measurement system based on the fusion of three-axis measurement and the MBD model. The measurement system applies the measurement method described in any one of the above, and includes: a measuring instrument attitude control module, a measurement point planning module, and an intelligent adjustment module;

[0035] The attitude control module of the measuring instrument preliminarily measures the blade through a three-axis measuring instrument and fits it into the first NURBS surface, obtains the actual normal vector of the blade based on the first NURBS surface, and adjusts the attitude of the three-axis measuring instrument by using the actual normal vector;

[0036] The measuring point planning module is used to construct a convolutional neural network, train the convolutional neural network based on the MBD model to obtain a measuring point prediction model for the blade surface, and plan the measuring points by using the measuring point prediction model for the blade surface;

[0037] The intelligent adjustment module is used to calculate a comprehensive deviation index, adjust the measuring speed and the measuring point spacing of the three-axis measuring instrument based on the comprehensive deviation index, calculate the similarity between measuring points, and perform process parameter feedback based on the similarity.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] The present invention solves the problems of low measuring accuracy, low efficiency, poor fusion effect between measuring data and the MBD model, and difficulty in batch measurement in the existing blade measurement technology. Through innovative three-axis measurement path planning, intelligent data fusion algorithms, real-time feedback optimization mechanisms, and integrated measurement platforms, high-precision and high-efficiency batch measurement of complex blade surfaces is achieved, and the differences between actual blades and the MBD model are accurately analyzed, providing strong support for quality control and design optimization in blade production. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0041] Figure 1 It is a schematic flowchart of the method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0043] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0044] Example 1

[0045] In this embodiment, as Figure 1 shown, a blade measurement method based on the fusion of three-axis measurement and MBD model includes the following steps:

[0046] S1. Initially measure the blade with a three-axis measuring instrument and fit it into the first NURBS surface. Obtain the actual normal vector of the blade based on the first NURBS surface, and use the actual normal vector to adjust the attitude of the three-axis measuring instrument.

[0047] In this embodiment, a modular positioning tooling is adopted. Each tooling unit is equipped with a high-precision pneumatic fixture and a magnetic base, supporting the simultaneous fixation of 4-8 blades (the specific quantity is adjusted according to the blade size). The tooling reference surface is aligned with the coordinate system of the three-axis measuring instrument through laser calibration to ensure that the measurement references of all blades are consistent. The surface point cloud data of the blades that have been measured in the same batch (stored in the MBD database) is collected by the three-axis measuring instrument and fitted into the NURBS surface S 实际 (u, v). Align the actual surface data with the theoretical surface S 理论 (u, v) of the MBD model through the parameter domain (u, v) to ensure that the same parameter point corresponds to the same position.

[0048] The actual normal vector includes:

[0049]

[0050] Among them, S 实际 represents the actual surface curve, and u and v represent the abscissa and ordinate of the parameter point. Each time a blade is measured, the actual surface database is automatically updated for subsequent normal vector compensation of the blade.

[0051] The method of using the actual normal vector to adjust the attitude of the three-axis measuring instrument includes: calculating the corrected normal vector:

[0052] n 修正 = n 理论 + Δn

[0053] Δn = α·(n 理论 - n 实际 ) + β·(κ 理论 - κ 实际 )·t

[0054] Among them, n 修正 represents the corrected normal vector, Δn represents the normal vector deviation, n 理论 represents the theoretical normal vector calculated from the MBD model, κ 理论 represents the theoretical local curvature provided by the MBD model, κ 实际The actual local curvature representing the fitting calculation of the actual surface, α represents the normal vector compensation weight, which is output by a backpropagation neural network (BPNN) trained with historical data and reflects the error sensitivity of the current area, β represents the curvature difference compensation coefficient, fixed at 0.05, and t represents the surface tangent vector. Adjust the attitude of the three-axis measuring instrument based on the corrected normal vector to ensure that the measurement path is always perpendicular to the actual surface.

[0055] S2. Construct a convolutional neural network, train the convolutional neural network based on the MBD model to obtain a blade surface measurement point prediction model, and use the blade surface measurement point prediction model to plan the measurement points.

[0056] In this embodiment, the network architecture of the convolutional neural network includes: (1) Convolutional layer: 4 layers of convolution, the convolutional kernel sizes are 5×5 (number of channels 32), 3×3 (number of channels 64), 3×3 (number of channels 128), 1×1 (number of channels 256) respectively, and the activation function is ReLU; (2) Pooling layer: Max pooling (2×2 window), stride 2; (3) Fully connected layer: The output layer has 3 neurons, corresponding to the curvature κ, the normal vector deviation Δn, and the recommended measurement point density N respectively. 新增 。

[0057] The method for training the convolutional neural network includes: parsing the STEP file of the MBD model into a NURBS surface, discretizing it into point cloud data, and generating a two-dimensional grayscale image through orthogonal projection; filtering and denoising the two-dimensional grayscale image, and extracting the surface boundary features through an edge detection algorithm; inputting the denoised image, the extracted boundary features, and the corresponding historical measurement data (including curvature, normal vector, actual measurement error) into the convolutional neural network for training to obtain a blade surface measurement point prediction model, and the loss function is:

[0058] L = λ1L MSE +λ2L 余弦相似度

[0059]

[0060] Among them, λ1 and λ2 represent eigenvalues. In this embodiment, λ1 can take 0.7 and λ2 can take 0.3.

[0061] S3. Calculate the deviation comprehensive index, adjust the measurement speed and measurement point spacing of the three-axis measuring instrument based on the deviation comprehensive index, calculate the similarity between measurement points, and perform process parameter feedback based on the similarity.

[0062] The method for adjusting the measurement speed and measurement point spacing of the three-axis measuring instrument includes: calculating the deviation comprehensive index:

[0063]

[0064] Among them, δ x represents the coordinate deviation in the x - direction of the measurement point (the difference between the measured coordinate and the theoretical coordinate), δ y represents the coordinate deviation in the y - direction of the measurement point, δ z represents the coordinate deviation in the z - direction of the measurement point, γ represents the local torsion angle of the surface (reflecting the degree of surface distortion), κ represents the local curvature difference weight coefficient (used to balance the influence of curvature on the comprehensive deviation); if δ 综合 > 0.1mm, then reduce the measurement speed v to 70% of the original value, and adjust the measurement point spacing d to d 新 = d 原 × 0.5; if δ 综合 > 0.05mm, then increase the measurement speed v to 120% of the original value, and adjust the measurement point spacing d to d 新 = d 原 × 1.2.

[0065] The method for process parameter feedback includes: for the measurement point p i , calculate the curvature covariance matrix C within its 10×10 neighborhood, the eigenvalues are λ1 and λ2, and construct the eigenvector F = [λ1, λ2, n]; calculate the similarity based on the eigenvector:

[0066]

[0067] Among them, F 实测 represents the measured eigenvector (constructed from the eigenvalues of the curvature covariance matrix), F 理论 represents the theoretical eigenvector (constructed from the eigenvalues of the curvature covariance matrix of the corresponding point of the MBD model), w represents the weight vector (weights assigned according to curvature sensitivity); if s < 0.85, then it is determined that the matching is abnormal, generate an alarm signal, pause the measurement of the current batch, and send an adjustment instruction to the production system through the OPCUA protocol, and at the same time record the abnormal data into the SQL database.

[0068] Example 2

[0069] In this embodiment, a blade measurement system based on the fusion of three - axis measurement and MBD model includes: a measuring instrument attitude control module, a measurement point planning module, and an intelligent adjustment module.

[0070] The measuring instrument attitude control module conducts a preliminary measurement on the blade through a three - axis measuring instrument and fits it into the first NURBS surface, obtains the actual normal vector of the blade based on the first NURBS surface, and adjusts the attitude of the three - axis measuring instrument using the actual normal vector.

[0071] The measurement point planning module is used to construct a convolutional neural network, train the convolutional neural network based on the MBD model to obtain a blade surface measurement point prediction model, and use the blade surface measurement point prediction model to plan the measurement points.

[0072] The intelligent adjustment module is used to calculate the comprehensive deviation index, adjust the measurement speed and measurement point spacing of the three-axis measuring instrument based on the comprehensive deviation index, calculate the similarity between measurement points, and perform process parameter feedback based on the similarity.

[0073] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A blade measurement method based on the fusion of three-axis measurement and MBD model, characterized in that, It includes the following steps: Preliminarily measure the blade with a three-axis measuring instrument and fit it into the first NURBS surface, obtain the actual normal vector of the blade based on the first NURBS surface, and adjust the attitude of the three-axis measuring instrument by using the actual normal vector; Construct a convolutional neural network, train the convolutional neural network based on the MBD model to obtain a blade surface measurement point prediction model, and use the blade surface measurement point prediction model to plan the measurement points; Calculate the deviation comprehensive index, adjust the measurement speed and measurement point spacing of the three-axis measuring instrument based on the deviation comprehensive index, calculate the similarity between measurement points, and perform process parameter feedback based on the similarity; 2. The blade measurement method based on the fusion of three-axis measurement and MBD model according to claim 1, wherein The actual normal vector includes: Among them, S 实际 represents the actual surface curve, and u and v represent the horizontal and vertical coordinates of the parameter points.

3. The leaf measurement method based on the fusion of three-axis measurement and MBD model according to claim 2, characterized in that The method for adjusting the attitude of the three-axis measuring instrument by using the actual normal vector includes: Calculate the corrected normal vector: n 修正 = n 理论 + Δn Δn = α·(n 理论 - n 实际 ) + β·(κ 理论 - κ 实际 )·t where n 修正 represents the corrected normal vector, Δn represents the normal vector deviation, and n 理论 represents the theoretical normal vector calculated from the MBD model, κ 理论 represents the theoretical local curvature provided by the MBD model, κ 实际 represents the actual local curvature calculated by fitting the actual surface, α represents the normal vector compensation weight, β represents the curvature difference compensation coefficient, and t represents the surface tangent vector; Adjust the attitude of the three-axis measuring instrument based on the corrected normal vector.

4. The blade measurement method based on the fusion of three-axis measurement and MBD model according to claim 3, wherein The method for training the convolutional neural network includes: Parse the STEP file of the MBD model into a NURBS surface, discretize it into point cloud data, and generate a two-dimensional grayscale image through orthogonal projection; Filter and denoise the two-dimensional grayscale image, and extract the surface boundary features through an edge detection algorithm; Input the denoised image, the extracted boundary features, and the corresponding historical measurement data into the convolutional neural network for training to obtain the blade surface measurement point prediction model.

5. A blade measurement method based on the fusion of three-axis measurement and MBD model according to claim 4, characterized in that, The method for adjusting the measurement speed and measurement point spacing of the three-axis measuring instrument includes: Calculate the deviation comprehensive index: Among them, δ x represents the coordinate deviation in the x-direction of the measurement point, and δ y represents the coordinate deviation in the y-direction of the measurement point, and δ z represents the coordinate deviation in the z-direction of the measurement point. γ represents the local twist angle of the surface, and κ represents the local curvature difference weight coefficient; If δ 综合 > 0.1 mm, the measurement speed v is reduced to 70% of the original value, and the measuring point spacing d is adjusted to d 新 = d 原 × 0.5; If δ 综合 > 0.05 mm, the measurement speed v is increased to 120% of the original value, and the measuring point spacing d is adjusted to d 新 = d 原 × 1.

2.

6. The leaf measurement method based on the fusion of three-axis measurement and MBD model according to claim 5, characterized in that The method for performing process parameter feedback includes: For measurement point p i , calculate the curvature covariance matrix C within its 10×10 neighborhood. The eigenvalues are λ1 and λ2, and construct the eigenvector F = [λ1, λ2, n]; Calculate the similarity based on the feature vector: Among them, F 实测 represents the measured eigenvector, and F 理论 represents the theoretical eigenvector, and w represents the weight vector; If s < 0.85, it is determined that the matching is abnormal, an alarm signal is generated, the measurement of the current batch is paused, an adjustment instruction is sent to the production system through the OPCUA protocol, and the abnormal data is recorded in the SQL database.

7. A blade measurement system based on the fusion of three-axis measurement and MBD model, the measurement system applying the measurement method according to any one of claims 1-6, characterized in that It includes: A measuring instrument attitude control module, a measurement point planning module, and an intelligent adjustment module; The measuring instrument attitude control module preliminarily measures the blade with a three-axis measuring instrument and fits it into the first NURBS surface, obtains the actual normal vector of the blade based on the first NURBS surface, and adjusts the attitude of the three-axis measuring instrument by using the actual normal vector; The measurement point planning module is used to construct a convolutional neural network, train the convolutional neural network based on the MBD model to obtain a blade surface measurement point prediction model, and use the blade surface measurement point prediction model to plan the measurement points; The intelligent adjustment module is used to calculate the deviation comprehensive index, adjust the measurement speed and measurement point spacing of the three-axis measuring instrument based on the deviation comprehensive index, calculate the similarity between measurement points, and perform process parameter feedback based on the similarity.

Images