A spatial parameter measurement system based on three-coordinate measuring machine

Abstract

The present invention belongs to the field of three-coordinate measurement technology, and specifically discloses a spatial parameter measurement system based on a three-coordinate measuring machine, including a measuring platform, a support arm, an X-axis drive module, a Y-axis drive module, a Z-axis drive module and a measuring component. A motion controller is provided on the top of the control box to control the movement of each drive module. The box integrates data acquisition, dynamic compensation, parameter fusion and execution control modules. The data acquisition module obtains multi-source and environmental parameters, the dynamic compensation module outputs compensation instructions based on them, the parameter fusion module outputs fused point cloud coordinates, and the execution control module sends instructions to the motion controller to drive the three axes to move according to a preset path. The present invention constructs a multi-module collaborative architecture, combines contact and non-contact measurement methods, solves the problem of precision attenuation of traditional three-coordinate measuring machines, realizes dynamic error correction, and improves the measurement adaptability to high-precision workpieces.

Description

Technical Field

[0001] The invention belongs to the technical field of three-coordinate measurement, and in particular relates to a spatial parameter measurement system based on a three-coordinate measuring machine. Background Art

[0002] As core equipment in the field of precision measurement, coordinate measuring machines (CMMs) utilize orthogonal motion systems in the X, Y, and Z directions, working in conjunction with measuring components, to achieve high-precision inspection of spatial parameters such as workpiece dimensions, form and position tolerances. Traditional CMMs primarily rely on contact probes or non-contact sensors to acquire point cloud data, achieving parameter measurement through coordinate system transformation and algorithmic processing. However, the shift toward higher precision and greater sophistication in manufacturing is placing higher demands on measuring equipment's environmental adaptability and dynamic measurement accuracy.

[0003] Existing coordinate measuring machines (CMMs) generally lack dynamic error compensation mechanisms during measurement. On the one hand, when the probe contacts the workpiece, changes in contact force, movement speed, and path can cause elastic deformation or mechanical vibration of the probe, resulting in coordinate acquisition errors. On the other hand, fluctuations in ambient temperature can cause mechanical structures such as the measuring platform and support arm to expand and contract, further amplifying measurement errors. Furthermore, the point cloud data from non-contact scanning units is not reliable enough for reflective surfaces and steep feature areas. Existing systems fail to effectively integrate multi-source data for error correction, making them unable to meet the measurement requirements of high-precision workpieces in aerospace, precision instrumentation, and other applications. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects in the prior art and provide a spatial parameter measurement system based on a three-coordinate measuring machine.

[0005] The present invention provides a spatial parameter measurement system based on a three-dimensional coordinate measuring machine, comprising a measuring platform arranged on the top of a control box and a support arm arranged on one side of the top of the control box, an X-axis drive module, a Y-axis drive module and a Z-axis drive module arranged on the top of the support arm, and a measuring assembly arranged on the bottom of the Z-axis drive module;

[0006] A motion controller is also provided on the top of the control box, and the motion controller is used to control the X-axis drive module to drive the Y-axis drive module to move in the X-axis direction; control the Y-axis drive module to drive the Z-axis drive module to move in the Y-axis direction; and control the Z-axis drive module to drive the measuring assembly to move in the Z-axis direction;

[0007] The control box is internally integrated with:

[0008] Data acquisition module, used to collect multi-source parameter information and environmental parameter information of the workpiece to be tested;

[0009] A dynamic compensation module, connected to the data acquisition module, outputs a dynamic compensation instruction based on the multi-source parameter information and the environmental parameter information;

[0010] The execution control module is connected to the dynamic compensation module, and sends the dynamic compensation instruction to the motion controller to control the X-axis drive module, the Y-axis drive module and the Z-axis drive module to move according to a preset path.

[0011] A further solution is that the measuring component includes a sliding frame arranged at the output end of the Z-axis drive module and slidably connected to the Z-axis drive module, a rotating motor is arranged inside the sliding frame, a probe is arranged at the output end of the rotating motor, a laser scanning module is arranged on one side of the probe, a connecting block is arranged at the bottom of the sliding frame, the probe passes through the connecting block, and binocular structured light cameras are arranged on both sides of the connecting block.

[0012] A further solution is that the data acquisition module includes a contact acquisition unit, a non-contact scanning unit and an environmental parameter acquisition unit;

[0013] The contact acquisition unit is a probe, which is used to obtain the three-dimensional coordinates (X, Y, Z) of the probe, the probe velocity (Vx, Vy, Vz) and the probe contact force Fz;

[0014] The non-contact scanning unit is composed of a laser scanning module and a binocular structured light camera, and is used to obtain surface point cloud data of the workpiece to be measured;

[0015] The environmental parameter acquisition unit includes a temperature sensor for acquiring the environmental temperature T.

[0016] A further solution is that the dynamic compensation module includes:

[0017] Multi-source data input unit, used to receive the probe's three-dimensional coordinates (X, Y, Z), probe velocity (Vx, Vy, Vz), probe contact force Fz, and ambient temperature T in real time;

[0018] The error prediction model uses a hybrid architecture of depthwise separable convolutional layers and gated recurrent units, where:

[0019] The input is 8-channel time series data (Xr, Yr, Zr, Vx, Vy, Vz, Fz, T) consisting of the probe's three-dimensional coordinates (X, Y, Z), probe velocity (Vx, Vy, Vz), probe contact force Fz, and ambient temperature T;

[0020] The depthwise separable convolution layer includes 8 independent 3×3 convolution kernels to extract spatial features channel by channel and output 8×N×N feature maps. After extracting spatial features, the channel information is fused through 1×1 point-by-point convolution to generate an 8-channel feature map, where N is the size of the time window.

[0021] The hidden layer dimension of the gated recurrent unit is 128, and the 8-channel feature map generated by the depthwise separable convolutional layer is flattened into an 8×N×N dimensional vector as input, and the hidden state is output;

[0022] The fully connected mapping layer is used to concatenate the global pooling results of the depthwise separable convolutional layer with the hidden state of the gated recurrent unit, and map the generated compensation parameters ΔR and (ΔX, ΔY, ΔZ). ΔR is the probe radius compensation coefficient, and ΔX, ΔY, and ΔZ are coordinate offsets.

[0023] A compensation instruction generation unit, used to convert the compensation parameters into analog signals recognizable by the motion controller through the FPGA acceleration unit;

[0024] The compensation instruction generating unit is connected to the execution control module and is used to send the analog signal to the motion controller in real time.

[0025] A further solution is that the error prediction model is constructed as follows:

[0026] Collect the three-dimensional coordinates (X, Y, Z), velocity (Vx, Vy, Vz), probe contact force Fz and ambient temperature T of several sets of probes under different measurement conditions;

[0027] The continuous time series data is divided into a set time window size N to form an 8×N×N feature map. High-precision measurement equipment is used to obtain real measurement values, which are compared with the collected data to obtain the true compensation parameters corresponding to each sample. Manual experts mark the feature map and the true compensation parameters to obtain paired data samples.

[0028] The paired data samples are input into the model and trained iteratively for multiple times to obtain an error prediction model whose inputs are the probe's three-dimensional coordinates (X, Y, Z), velocity (Vx, Vy, Vz), probe contact force Fz, and ambient temperature T, and whose outputs are compensation parameters ΔR and (ΔX, ΔY, ΔZ).

[0029] A further solution is that a parameter fusion module is integrated into the control box and connected to the data acquisition module for fusing multi-source parameter information to output fused point cloud coordinates; the parameter fusion module includes:

[0030] a multi-source data acquisition unit configured to synchronously acquire laser scanning point cloud data and probe contact point cloud data; the laser scanning point cloud data includes three-dimensional coordinates (X1, Y1, Z1) and signal strength; the probe contact point cloud data includes three-dimensional coordinates (X2, Y2, Z2);

[0031] The data preprocessing unit is used to calculate the normal vector of the laser point cloud data, obtain the normal vector of each point, and calculate the signal-to-noise ratio of the laser point cloud; The dynamic weight allocation unit is configured as follows:

[0032] The initial weight of the laser point cloud is set to W = 0.5;

[0033] Adjust the laser point cloud weight W1 for areas that meet the following conditions:

[0034] Planar continuity area: When the angle between the normal vectors of adjacent points is less than 5°, the laser point cloud weight W1 is multiplied by the gain factor α, where α=1.2±0.1;

[0035] Reflective area: When the signal-to-noise ratio is less than the preset threshold, the laser point cloud weight W1 is multiplied by the attenuation factor β, and the probe point cloud weight is W2=1-W1, where β=0.5±0.1;

[0036] The point cloud fusion unit obtains the fused point cloud coordinates based on the dynamic weight allocation unit. The fused point cloud coordinates are (Xr, Yr, Zr), where:

[0037] ;

[0038] The point cloud fusion unit is connected to the error prediction model, and the fused point cloud coordinates (Xr, Yr, Zr) are input into the error prediction model to output compensation parameters ΔR, (ΔX, ΔY, ΔZ).

[0039] A further solution is that the measuring platform includes a support base arranged on the top of the control box, a clamp fixing plate is arranged on the top of the support base, a circular hole is opened on the clamp fixing plate to form a measuring area, a fixed clamping part and two sliding clamping parts are arranged on the clamp fixing plate along the circumference of the measuring area, a first guide rail is provided at the bottom of the sliding clamping part, and the sliding clamping part is slidably connected to the first guide rail; a second guide rail is also provided on the clamp fixing plate, the second guide rail is located between the two sliding clamping parts, a Y-shaped push rod is slidably provided on the second guide rail, a slide groove is provided in the middle of the sliding clamping part, and the Y-shaped push rod is slidably connected to the slide groove; a clamping cylinder is also provided on the clamp fixing plate, and the output end of the clamping cylinder is fixedly connected to the Y-shaped push rod.

[0040] A further solution is that the angle between the two sliding clamping parts is 120°, and the end of the Y-shaped push rod away from the clamping cylinder is two oblique rods with an angle of 60°, and the oblique rods are slidably connected to the slide groove, so that when the Y-shaped push rod slides along the second guide rail, the two oblique rods respectively drive the two sliding clamping parts to slide along the first guide rail.

[0041] A further solution is that the X-axis drive module and the Y-axis drive module have the same structure, both of which are screw transmission mechanisms, and the Z-axis drive module includes a Z-axis fixed plate slidably connected to the Y-axis drive module, a Z-axis guide rail is provided on one side of the Z-axis fixed plate, and a sliding frame is provided on the Z-axis guide rail, and the sliding frame is slidably connected to the Z-axis guide rail;

[0042] A Z-axis cylinder is provided on the top of the Z-axis fixed plate, and the output end of the Z-axis cylinder is fixedly connected to the sliding frame; a rotating motor is provided inside the sliding frame for driving the binocular structured light camera and the laser scanning module to rotate.

[0043] A further solution is that the control box is further connected to an external obstacle avoidance module;

[0044] The obstacle avoidance module includes a workpiece model building unit, a fixture model building unit, a model superposition unit and a boundary extraction unit;

[0045] The workpiece model building unit builds a three-dimensional model of the workpiece to be measured based on the workpiece scanning point cloud of the non-contact scanning unit;

[0046] The fixture model building unit builds a three-dimensional model of the fixture based on the fixture scanning point cloud of the non-contact scanning unit;

[0047] The model superposition unit is used to superimpose the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture in the same coordinate system;

[0048] The boundary extraction unit is used to traverse the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture and identify boundary points, connect the boundary points into a closed loop, and form an edge path of the overlapping area;

[0049] The boundary extraction unit is connected to the execution control module, and sends the edge path to the motion controller to control the X-axis drive module, the Y-axis drive module and the Z-axis drive module to move according to the edge path.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present invention solves the problem of precision degradation in traditional three-dimensional coordinate measuring machines due to changes in working conditions and environmental interference by constructing an integrated multi-module collaborative measurement system. Specifically, the data acquisition module acquires multi-source parameters such as the probe's three-dimensional coordinates, probe contact force, and ambient temperature in real time. The dynamic compensation module generates real-time compensation instructions, which are combined with the high-precision point cloud data from the parameter fusion module. Ultimately, the execution control module drives the three axes to move along a preset path, achieving dynamic correction of measurement errors. By combining the high-precision single-point measurement of a contact probe with non-contact laser scanning and large-area rapid scanning using binocular structured light, the system can not only meet the submicron precision measurement requirements for key features, but also efficiently complete full-area scanning of complex surfaces, significantly improving the measurement adaptability of high-precision workpieces.

[0052] The dynamic compensation module of the present invention adopts a hybrid architecture of deep separable convolutional layers and gated recurrent units to perform spatiotemporal feature fusion on 8-channel time series data consisting of the probe's three-dimensional coordinates, velocity, contact force and ambient temperature, and outputs the probe radius compensation coefficient and coordinate offset in real time. Compared with traditional empirical formula compensation, this model is trained with massive working condition data and can accurately capture nonlinear error sources such as mechanical deformation and thermal drift, reduce coordinate errors caused by temperature changes, and meet the measurement requirements of aerospace components in variable temperature environments.

[0053] The present invention utilizes an FPGA acceleration unit to achieve high-speed conversion of compensation parameters to analog signals, and cooperates with the three-axis linkage control of the motion controller to reduce the dynamic positioning error of the system during high-speed measurement, effectively solving the trajectory deviation problem caused by compensation lag in traditional equipment.

[0054] The parameter fusion module of this invention significantly improves blind spot measurement on complex surfaces by dynamically adjusting the weights of the laser point cloud and the probe contact point cloud. By using a temperature sensor to monitor ambient temperature in real time and embedding an error prediction model, the system automatically compensates for mechanical displacement caused by thermal deformation, eliminating the need for frequent manual calibration and significantly improving the stability of long-term continuous measurements. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The following drawings are merely provided for illustrative purposes only and are not intended to limit the scope of the present invention.

[0056] Figure 1 : Schematic diagram of the structure of the present invention;

[0057] Figure 2 : Schematic diagram of the three-axis drive structure of the present invention;

[0058] Figure 3 : Figure 2 Enlarged view of part A in the middle;

[0059] Figure 4: Schematic diagram of the measurement platform structure;

[0060] Figure 5 : Block diagram of the collaborative control principle of each module of the control box of the present invention;

[0061] Figure 6 : Structural connection diagram of the error budget model of the present invention;

[0062] In the figure: 1. Control box; 2. Measuring platform; 3. X-axis drive module; 4. Y-axis drive module; 5. Z-axis drive module; 6. Measuring component; 7. Motion controller; 8. Support seat; 9. Fixture fixing plate; 10. Measuring area; 11. Fixed clamping part; 12. First guide rail; 13. Sliding clamping part; 14. Y-shaped push rod; 15. Second guide rail; 16. Clamping cylinder; 17. Z-axis fixed plate; 18. Z-axis guide rail; 19. Sliding frame; 20. Rotating motor; 21. Z-axis cylinder; 22. Connecting block; 23. Arc frame; 24. Binocular structured light camera; 25. Fixed tube; 26. Laser scanning module; 27. Probe; 28. Data acquisition module; 29. ​​Dynamic compensation module; 30. Execution control Module; 31. Parameter fusion module; 32. Obstacle avoidance module; 33. Multi-source data input unit; 34. Error prediction model; 35. Compensation instruction generation unit; 36. Contact acquisition unit; 37. Non-contact scanning unit; 38. Environmental parameter acquisition unit; 39. Multi-source data acquisition unit; 40. Data preprocessing unit; 41. Dynamic weight allocation unit; 42. Point cloud fusion unit; 43. Workpiece model construction unit; 44. Fixture model construction unit; 45. Model superposition unit; 46. Boundary extraction unit; 47. Depth convolution layer; 48. 3×3 convolution kernel; 49. Point-by-point convolution layer; 50. Gated recurrent unit; 51. Global average pooling layer; 52. Splicing layer; 53. Fully connected mapping layer. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution, design method and advantages of the present invention more clear, the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] like Figure 1 and Figure 5As shown, the present invention provides a spatial parameter measurement system based on a three-dimensional coordinate measuring machine, which supports synchronous acquisition and real-time compensation of multi-source parameters, and solves the problem of precision attenuation caused by environmental changes and mechanical errors in traditional equipment. It includes a measuring platform 2 arranged on the top of a control box 1 and a support arm arranged on one side of the top of the control box 1, an X-axis drive module 3, a Y-axis drive module 4 and a Z-axis drive module 5 are arranged on the top of the support arm, and a measuring component 6 is arranged at the bottom of the Z-axis drive module 5; a motion controller 7 is also arranged on the top of the control box 1, and the motion controller 7 is used to control the X-axis drive module 3 to drive the Y-axis drive module 4 to move in the X-axis direction; and control the Y-axis drive module 4 to drive the Z-axis drive module 5 to move in the Y-axis direction; and control the Z-axis drive module 5 to drive the measuring component 6 to move in the Z-axis direction; Integrated inside the control box 1 are: a data acquisition module 28, which is used to collect multi-source parameter information and environmental parameter information of the workpiece to be measured; a dynamic compensation module 29, which is connected to the data acquisition module 28 and outputs dynamic compensation instructions based on the multi-source parameter information and environmental parameter information; a parameter fusion module 31, which is connected to the data acquisition module 28 and is used to fuse the multi-source parameter information to output the fused point cloud coordinates; an execution control module 30, which is connected to the dynamic compensation module 29 and sends the dynamic compensation instructions to the motion controller 7 to control the X-axis drive module 3, the Y-axis drive module 4 and the Z-axis drive module 5 to move according to the preset path.

[0065] like Figure 4As shown, in order to ensure the fixation of the workpiece to be measured, in this embodiment, the measuring platform 2 includes a support base 8 arranged on the top of the control box 1, and a fixture fixing plate 9 is arranged on the top of the support base 8. A circular hole is opened on the fixture fixing plate 9 to form a measuring area 10 of the workpiece to be measured, and a fixed clamping portion 11 and two sliding clamping portions 13 are arranged on the fixture fixing plate 9 along the side of the measuring area 10. A first guide rail 12 is provided at the bottom of the sliding clamping portion 13, and the sliding clamping portion 13 is slidably connected to the first guide rail 12; a second guide rail 15 is also provided on the fixture fixing plate 9, and the second guide rail 15 is located between the two sliding clamping portions 13. A Y-shaped push rod 14 is slidably provided on the second guide rail 15, and a slide groove is provided in the middle of the sliding clamping portion 13, and the Y-shaped push rod 14 is slidably connected to the slide groove; a clamping cylinder 16 is also provided on the fixture fixing plate 9, and the output end of the clamping cylinder 16 is fixedly connected to the Y-shaped push rod 14. The included angle between the two sliding clamping parts 13 is 120°, and the included angle between each sliding clamping part 13 and the fixed clamping part 11 is also 120°. The end of the Y-shaped push rod 14 away from the clamping cylinder 16 is composed of two inclined rods with an included angle of 60°. The inclined rods are slidably connected to the slide groove, so that when the Y-shaped push rod 14 slides along the second guide rail 15, the two inclined rods respectively drive the two sliding clamping parts 13 to slide along the first guide rail 12. The measuring platform 2 adopts a linkage design between the Y-shaped push rod 14 and the sliding clamping part 13. Driven by the clamping cylinder 16, it realizes rapid positioning and adaptive clamping of the workpiece, and evenly distributes the clamping force, thereby improving the workpiece clamping efficiency.

[0066] like Figure 2 As shown, in this embodiment, the X-axis drive module 3 and the Y-axis drive module 4 have the same structure, both of which are screw transmission mechanisms. The Z-axis drive module 5 includes a Z-axis fixed plate 17 that is slidably connected to the Y-axis drive module 4. A Z-axis guide rail 18 is provided on one side of the Z-axis fixed plate 17. A sliding frame 19 is provided on the Z-axis guide rail 18, and the sliding frame 19 is slidably connected to the Z-axis guide rail 18; a Z-axis cylinder 21 is provided on the top of the Z-axis fixed plate 17, and the output end of the Z-axis cylinder 21 is fixedly connected to the sliding frame 19; and a rotating motor 20 is provided inside the sliding frame 19. Figure 3As shown, the measurement assembly 6 includes a sliding frame 19 disposed at the output end of the Z-axis drive module 5 and slidably connected to the Z-axis drive module 5. A rotary motor 20 is disposed within the sliding frame 19, and a probe 27 is disposed at the output end of the rotary motor 20. A fixed tube 25 is sleeved on the outside of the probe 27, and a laser scanning module 26 is disposed on the fixed tube 25. A connecting block 22 is disposed at the bottom of the sliding frame 19, and the probe 27 passes through the connecting block 22. An arc frame 23 is disposed on both sides of the connecting block 22, and a binocular structured light camera 24 is disposed on the arc frame 23. The rotating motor 20 drives the binocular structured light camera 24 and the laser scanning module 26 to rotate. In conjunction with the coordinated control of the Z-axis cylinder 21 and the Z-axis guide rail 18, the binocular structured light camera 24 and the laser scanning module 26 can be adjusted at multiple angles, thereby expanding the detection range of the measurement assembly 6. The device is particularly suitable for non-contact scanning of complex structures such as deep holes and cavities.

[0067] In this embodiment, the data acquisition module 28 includes a contact acquisition unit 36, a non-contact scanning unit 37, and an environmental parameter acquisition unit 38. The contact acquisition unit 36 ​​is the probe 27, and is used to acquire the three-dimensional coordinates (X, Y, Z) of the probe 27, the velocity (Vx, Vy, Vz) of the probe 27, and the probe contact force Fz. The probe contact force Fz is monitored in real time by a force-sensitive sensor integrated within the probe 27, which is used to sense the normal force when the probe 27 contacts the workpiece. The non-contact scanning unit 37, consisting of the laser scanning module 26 and the binocular structured light camera 24, is used to acquire surface point cloud data of the workpiece to be measured. The environmental parameter acquisition unit 38 includes a temperature sensor for acquiring the ambient temperature T. The temperature sensor is located inside the control box 1 or near the measurement platform 2 to ensure real-time monitoring of temperature changes in the measurement environment and avoid thermal expansion and contraction errors caused by temperature fluctuations in the support arm, drive module, and probe 27. This enables comprehensive acquisition of the workpiece's three-dimensional coordinates, velocity, contact force, surface point cloud data, and ambient temperature. This multi-source data synchronization acquisition mechanism provides rich input information for subsequent error compensation and point cloud fusion, resolving the information-missing issues inherent in traditional single-modal measurement equipment. The contact acquisition unit 36 ​​acquires highly accurate single-point coordinates and contact force data, making it suitable for precise measurement of key features. The non-contact scanning unit 37 rapidly acquires large-area surface point clouds, making it suitable for measuring complex curved surfaces or delicate workpieces. The temperature sensor monitors ambient temperature in real time, providing a basis for thermal deformation error compensation. Combined, these three elements enable the system to balance accuracy, efficiency, and adaptability, meeting measurement requirements under diverse working conditions.

[0068] In this embodiment, the dynamic compensation module 29 includes: a multi-source data input unit 33, an error prediction model 34 and a compensation instruction generation unit 35; wherein, the multi-source data input unit 33 is used to receive the three-dimensional coordinates (X, Y, Z) of the probe 27, the speed of the probe 27 (Vx, Vy, Vz), the probe contact force Fz, and the ambient temperature T in real time; the error prediction model 34 adopts a hybrid architecture of a depth-separable convolutional layer and a gated recurrent unit 50, which can effectively extract spatial features and time-dependent features in time series data. Specifically: the input is 8-channel time series data (Xr, Yr, Zr, Vx, Vy, Vz, Fz, T) consisting of the three-dimensional coordinates (X, Y, Z) of the probe 27, the speed of the probe 27 (Vx, Vy, Vz), the probe contact force Fz and the ambient temperature T; the depthwise separable convolution layer includes 8 independent 3×3 convolution kernels 48 to extract spatial features channel by channel and output an 8×N×N feature map. After extracting the spatial features, the channel information is fused through a 1×1 point-by-point convolution layer 49 to generate an 8-channel feature map. Where N is the size of the time window; the hidden layer dimension of the gated recurrent unit is 128, and the 8-channel feature map generated by the depthwise separable convolutional layer is flattened into an 8×N×N dimensional vector as input, and the hidden state is output; the fully connected mapping layer 53 is used to concatenate the global pooling result of the depthwise separable convolutional layer with the hidden state of the gated recurrent unit, and map to generate compensation parameters ΔR, (ΔX, ΔY, ΔZ); wherein ΔR is the probe 27 radius compensation coefficient, and ΔX, ΔY, ΔZ are coordinate offsets; through 8 The model inputs channel timing data (3D coordinates, 3D velocity, probe contact force, and temperature), and outputs the probe 27 radius compensation coefficient ΔR and coordinate offsets (ΔX, ΔY, ΔZ) in real time. This dynamically compensates for mechanical motion errors, thermal deformation errors, and probe 27 deformation errors caused by contact force, addressing the accuracy degradation problem of traditional CMMs caused by environmental changes and fluctuating operating conditions. A compensation instruction generation unit 35 converts the compensation parameters into analog signals recognizable by the motion controller 7 via the FPGA acceleration unit. This unit is connected to the execution control module 30 and transmits the analog signals to the motion controller 7 in real time. In this process, the compensation parameters ΔR and (ΔX, ΔY, ΔZ) are critical data for compensating for errors generated during CMM measurement. The FPGA acceleration unit rapidly converts these digital compensation parameters into analog signals recognizable by the motion controller 7. After receiving the analog signal, the motion controller 7 drives the motors based on the information carried by the signal, thereby controlling the X-axis drive module 3, the Y-axis drive module 4, and the Z-axis drive module 5 to achieve precise movement to achieve error compensation. Specifically, the motion controller 7 performs motion planning based on the compensation parameters obtained through analysis, combined with the current measurement task and the current position of the three axes. It determines parameters such as the target position, movement speed, and acceleration of each axis to ensure that the three axes can accurately move to the compensated position.

[0069] The construction process of the error prediction model 34 is as follows:

[0070] The three-dimensional coordinates (X, Y, Z), velocity (Vx, Vy, Vz), probe contact force Fz, and ambient temperature T of several sets of probes 27 are collected under different measurement conditions. The collected data are normalized and mapped to the interval [0, 1] or [-1, 1] to eliminate the dimensional influence of different features and accelerate the convergence of the model.

[0071] Continuous time series data is divided into a set time window size N to form an 8×N×N feature map. Where N is the time window. In this embodiment, N is 50, that is, the data of every 50 time points is regarded as a sample. Each sample contains data of 8 channels (X, Y, Z, Vx, Vy, Vz, Fz, T), thereby forming an 8×N×N feature map. High-precision measurement equipment is used to obtain real measurement values, which are compared with the collected data to obtain the true compensation parameters corresponding to each sample. Manual experts mark the feature map and the true compensation parameters to obtain paired data samples.

[0072] The paired data samples are input into the model and trained iteratively for multiple times to obtain an error prediction model 34 whose inputs are the three-dimensional coordinates (X, Y, Z), velocity (Vx, Vy, Vz), probe contact force Fz and ambient temperature T of the probe 27, and whose outputs are compensation parameters ΔR and (ΔX, ΔY, ΔZ).

[0073] like Figure 6As shown above, the depthwise separable convolution layer includes a depthwise convolution layer 47 and a pointwise convolution layer 49. The depthwise convolution layer 47 applies eight independent 3×3 convolution kernels 48 to the input 8×N×N feature map, performing convolution operations on each channel, extracting spatial features channel by channel. For example, the spatial variation features of temperature data or the spatial variation features of probe 27's X coordinates can be extracted separately, and the output is still an 8×N×N feature map. The pointwise convolution layer 49 convolves the feature map output by the depthwise convolution with a 1×1 convolution kernel, fusing the features of each channel to generate a new 8-channel feature map. The gated recurrent unit 50 flattens the 8-channel feature map output by the depthwise separable convolution layer into an 8×N×N dimensional vector, which serves as the input to the gated recurrent unit. The gated recurrent unit 50 outputs a 128-dimensional hidden state vector. Fully connected mapping layer 53: The feature map output by the depthwise separable convolution layer is subjected to a global average pooling 51 operation through the global average pooling layer to be compressed into an 8-dimensional vector. The 8-dimensional vector is then concatenated with the 128-dimensional hidden state vector output by the gated recurrent unit through the concatenation layer 52 to obtain a 136-dimensional vector. The concatenated 136-dimensional vector is input to the fully connected mapping layer 53 and mapped to a 4-dimensional compensation parameter ΔR (ΔX, ΔY, ΔZ) through a linear transformation.

[0074] Continue to refer to Figure 5The parameter fusion module 31 includes a multi-source data acquisition unit 39, a data pre-processing unit 40, a dynamic weight allocation unit 41, and a point cloud fusion unit 42. Specifically, the multi-source data acquisition unit 39 is configured to synchronously acquire laser scanning point cloud data and probe 27 contact point cloud data; the laser scanning point cloud data includes three-dimensional coordinates (X1, Y1, Z1) and signal intensity; the probe 27 contact point cloud data includes three-dimensional coordinates (X2, Y2, Z2); the data pre-processing unit 40 is used to calculate the normal vector of the laser point cloud data, obtain the normal vector of each point, and calculate the laser point cloud signal-to-noise ratio; The dynamic weight allocation unit 41 is configured as follows: the initial weight of the laser point cloud is set to W=0.5; the laser point cloud weight W1 is adjusted for areas that meet the following conditions: plane continuity area: when the angle between the normal vectors of adjacent points is less than 5°, the laser point cloud weight W1 is multiplied by the gain factor α, where α=1.2±0.1; reflective area: when the signal-to-noise ratio is ≤20dB, the laser point cloud weight W1 is multiplied by the attenuation factor β, and the probe 27 point cloud weight is W2=1-W1, where β=0.5±0.1; for the plane continuity area (the angle between the normal vectors of adjacent points is less than 5°), the weight of the laser scanning point cloud data is increased by the gain factor α, because laser scanning has the characteristics of high efficiency and high precision in large-area plane measurement. For reflective areas (SNR ≤ 20dB), the weight of the laser scanning point cloud data is reduced by the attenuation factor β, and the weight of the probe 27 contact point cloud data is increased, because the probe 27 contact measurement in the reflective area can provide more accurate measurement results; the point cloud fusion unit 42 obtains the fused point cloud coordinates according to the weighted average method. The fused point cloud coordinates are (Xr, Yr, Zr), where

[0075] .

[0076] The point cloud fusion unit 42 is connected to the error prediction model 34 and inputs the fused point cloud coordinates (Xr, Yr, Zr) into the error prediction model 34 to output compensation parameters ΔR and (ΔX, ΔY, ΔZ). By fusing the coordinates, the accuracy of the compensation parameters ΔR and (ΔX, ΔY, ΔZ) is further improved.

[0077] In this embodiment, the control box 1 is also connected to an external obstacle avoidance module 32; the obstacle avoidance module 32 includes a workpiece model construction unit 43, a fixture model construction unit 44, a model superposition unit 45 and a boundary extraction unit 46; wherein, the workpiece model construction unit 43 constructs a three-dimensional model of the workpiece to be measured based on the workpiece scanning point cloud of the non-contact scanning unit 37. Specifically, when the non-contact scanning unit 37 scans the workpiece, the laser scanning module 26 emits a laser beam to the workpiece surface, and obtains the depth information of the workpiece surface by measuring the time or phase difference of the laser reflected light. The binocular structured light camera 24 uses two cameras to shoot the structured light pattern on the workpiece surface from different angles, and calculates the three-dimensional coordinates of the workpiece surface point by the triangulation principle; the fixture model construction unit 44 is similar to the workpiece model construction unit 43. The non-contact scanning unit 37 scans the fixture to obtain the fixture's The point cloud coordinates are then used to construct a three-dimensional model of the fixture; the model superposition unit 45 is used to superimpose the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture in the same coordinate system. It is necessary to determine a unified global coordinate system and convert the three-dimensional model of the workpiece and the three-dimensional model of the fixture into the coordinate system. In this embodiment, a fixed reference point or reference plane can be selected, and the coordinates of the two models can be converted using a coordinate transformation matrix. In a unified coordinate system, the three-dimensional model of the workpiece and the three-dimensional model of the fixture are superimposed, and the coordinate data of the two models are merged into the same data structure for subsequent boundary extraction; the boundary extraction unit 46 is used to traverse the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture and identify boundary points, connect the boundary points into a closed loop, and form an edge path of the overlapping area. In this embodiment, a sixteen-neighborhood method is used to determine the boundary position of the model. For example, if there are a large number of points in the neighborhood of a point that do not belong to the same model, or the distance between the point and its neighboring points varies greatly, the point can be identified as a boundary point. When the boundary points are connected, starting from one boundary point, the nearest unconnected boundary points are searched in sequence until a closed loop is formed; the boundary extraction unit 46 is connected to the execution control module 30, and sends the edge path to the motion controller 7 in the form of a digital signal to control the X-axis drive module 3, the Y-axis drive module 4 and the Z-axis drive module 5 to move according to the edge path.

[0078] While various embodiments of the present invention have been described above, the above descriptions are intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A spatial parameter measurement system based on a three-coordinate measuring machine, characterized in that: It includes a measuring platform arranged on the top of the control box and a support arm arranged on one side of the top of the control box, an X-axis drive module, a Y-axis drive module and a Z-axis drive module are arranged on the top of the support arm, and a measuring component is arranged at the bottom of the Z-axis drive module; A motion controller is also provided on the top of the control box, and the motion controller is used to control the X-axis drive module to drive the Y-axis drive module to move in the X-axis direction; and controlling the Y-axis driving module to drive the Z-axis driving module to move in the Y-axis direction; and controlling the Z-axis driving module to drive the measuring component to move in the Z-axis direction; The control box is internally integrated with: Data acquisition module, used to collect multi-source parameter information and environmental parameter information of the workpiece to be tested; A dynamic compensation module is connected to the data acquisition module, and outputs compensation parameters through an error prediction model based on the multi-source parameter information and the environmental parameter information, thereby generating a dynamic compensation instruction; a parameter fusion module, connected to the data acquisition module, for fusing multi-source parameter information to output fused point cloud coordinates; and optimizing the compensation parameters by using the fused point cloud coordinates as an input parameter of the error prediction model; An execution control module is connected to the dynamic compensation module, and sends the dynamic compensation instruction to the motion controller to control the X-axis drive module, the Y-axis drive module, and the Z-axis drive module to move along a preset path; The dynamic compensation module includes: Multi-source data input unit, used to receive parameter information of the data acquisition module in real time; The error prediction model is configured to take parameter information as input and output compensation parameters ΔR and (ΔX, ΔY, ΔZ). ΔR is the probe radius compensation coefficient, and (ΔX, ΔY, ΔZ) is the coordinate offset. A compensation instruction generating unit, connected to the execution control module, for sending compensation parameters to the motion controller in real time; The control box is also connected to an external obstacle avoidance module; the obstacle avoidance module includes a workpiece model construction unit, a fixture model construction unit, a model superposition unit and a boundary extraction unit; The workpiece model building unit builds a three-dimensional model of the workpiece to be measured based on the workpiece scanning point cloud of the non-contact scanning unit; The fixture model building unit builds a three-dimensional model of the fixture based on the fixture scanning point cloud of the non-contact scanning unit; The model superposition unit is used to superimpose the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture in the same coordinate system; The boundary extraction unit is used to traverse the three-dimensional model of the workpiece to be measured and the three-dimensional model of the fixture and identify boundary points, connect the boundary points into a closed loop, and form an edge path of the overlapping area; The boundary extraction unit is connected to the execution control module, and sends the edge path to the motion controller to control the X-axis drive module, the Y-axis drive module and the Z-axis drive module to move according to the edge path.

2. The spatial parameter measurement system based on a three-coordinate measuring machine according to claim 1, characterized in that: The measuring component includes a sliding frame arranged at the output end of the Z-axis drive module and slidably connected to the Z-axis drive module, a rotating motor is arranged inside the sliding frame, a probe is arranged at the output end of the rotating motor, a laser scanning module is arranged on one side of the probe, a connecting block is arranged at the bottom of the sliding frame, the probe passes through the connecting block, and binocular structured light cameras are arranged on both sides of the connecting block.

3. The spatial parameter measurement system based on a three-coordinate measuring machine according to claim 2, characterized in that: The parameter fusion module includes: a multi-source data acquisition unit configured to synchronously acquire laser scanning point cloud data and probe contact point cloud data; The dynamic weight allocation unit is configured as follows: Adjust the laser point cloud weight W1 for areas that meet the following conditions, and the probe point cloud weight W2 = 1-W1: Plane continuity area: When the angle between the normal vectors of adjacent points is less than the preset angle, the laser point cloud weight W1 is multiplied by the gain factor; Reflective area: When the signal-to-noise ratio is less than the preset threshold, the laser point cloud weight W1 is multiplied by the attenuation factor; The point cloud fusion unit obtains the fused point cloud coordinates by weighted averaging based on the laser point cloud weight W1 and the probe point cloud weight W2; the point cloud fusion unit is also configured to input the fused point cloud coordinates into the error prediction model to output compensation parameters ΔR, (ΔX, ΔY, ΔZ).

4. The spatial parameter measurement system based on a three-coordinate measuring machine according to claim 3, characterized in that: The measuring platform includes a support base arranged on the top of the control box, a clamp fixing plate is arranged on the top of the support base, a circular hole is opened on the clamp fixing plate to form a measuring area, a fixed clamping part and two sliding clamping parts are arranged on the clamp fixing plate along the circumference of the measuring area, a first guide rail is provided at the bottom of the sliding clamping part, and the sliding clamping part is slidably connected to the first guide rail; a second guide rail is also provided on the clamp fixing plate, the second guide rail is located between the two sliding clamping parts, a Y-shaped push rod is slidably provided on the second guide rail, a slide groove is provided in the middle of the sliding clamping part, and the Y-shaped push rod is slidably connected to the slide groove; a clamping cylinder is also provided on the clamp fixing plate, and the output end of the clamping cylinder is fixedly connected to the Y-shaped push rod.

5. The spatial parameter measurement system based on a three-coordinate measuring machine according to claim 4, characterized in that: The included angle between the two sliding clamping parts is 120°, and the end of the Y-shaped push rod away from the clamping cylinder is two oblique rods with an included angle of 60°. The oblique rods are slidably connected to the slide groove so that when the Y-shaped push rod slides along the second guide rail, the two oblique rods respectively drive the two sliding clamping parts to slide along the first guide rail.

6. The spatial parameter measurement system based on a three-dimensional coordinate measuring machine according to claim 5, characterized in that: The X-axis drive module and the Y-axis drive module have the same structure, both of which are screw transmission mechanisms. The Z-axis drive module includes a Z-axis fixed plate slidably connected to the Y-axis drive module, a Z-axis guide rail is provided on one side of the Z-axis fixed plate, and a sliding frame is provided on the Z-axis guide rail, and the sliding frame is slidably connected to the Z-axis guide rail; A Z-axis cylinder is provided on the top of the Z-axis fixed plate, and the output end of the Z-axis cylinder is fixedly connected to the sliding frame; a rotating motor is provided inside the sliding frame for driving the binocular structured light camera and the laser scanning module to rotate.

Images