Aero-engine rotor point cloud matching method and device based on adaptive weight

Abstract

The invention discloses an aero-engine rotor point cloud matching method and device based on adaptive weight, belongs to the technical field of aero-engine rotor stacking, and particularly relates to point cloud matching in virtual assembly in aero-engine rotor stacking. The problem of rigid body interference existing in a traditional ICP matching algorithm is solved; the method comprises an iterative optimization step: in the calculation process of the weighted Euclidean distance, a self-adaptive weight coefficient is introduced for introducing a constraint condition to suppress the generation of an interference phenomenon. The aero-engine rotor point cloud matching method and device based on the adaptive weight are suitable for completing aero-engine rotor stacking large-scale point cloud matching and guiding engine rotor stacking.

Description

Technical Field

[0001] The invention relates to the technical field of aircraft engine rotor stacking, and in particular to point cloud matching in virtual assembly of aircraft engine rotor stacking. Background Art

[0002] (Aviation) Engine rotors have the advantages of large meshing area and transmission torque, strong load-bearing capacity, high positioning accuracy, and good stability. When assembling engine rotors, the contact area between multi-stage rotors is large and the entire tooth surface is stressed, so it is necessary to precisely measure and evaluate the three-dimensional profile of the engine rotor.

[0003] With the increasing development of digital manufacturing technology, virtual assembly can use computer technology to simulate the entire process of actual assembly based on the three-dimensional model of the actual workpiece, and then realize the prediction of assembly results and assembly error diagnosis.

[0004] 3D shape measurement technology can provide massive, high-precision 3D point cloud coordinate data for virtual assembly by quickly measuring the target workpiece, and is the best choice for obtaining 3D information of workpieces. In 1992, Besl proposed the most classic iterative closest point algorithm, namely the ICP algorithm, which can iteratively match two parts of point clouds with different postures and establish the posture transformation relationship between the two parts of the point clouds. Today, the ICP matching algorithm is widely used in 3D scanning measurement, unifying two or more groups of point clouds scanned from different perspectives into the same world coordinate system.

[0005] However, when the traditional ICP matching algorithm matches the point clouds of two mating surfaces with machining errors, since the algorithm itself is an unconstrained matching algorithm and the core of the algorithm is to minimize the average value of the Euclidean distance between the corresponding points, there will be mutual interference between the mating surfaces after matching, that is, the mating surface point clouds will penetrate each other. However, in actual assembly, each part can be regarded as a rigid body, and there will be no penetration between the mating surfaces. There is a large difference between the matching results of the traditional ICP algorithm and the actual assembly results, and it cannot be directly applied to the calculation of assembly posture.

[0006] Therefore, there is an urgent need to explore a large-scale point cloud matching method suitable for the stacking points of the rotor of the core machine of an aircraft engine to guide the precise assembly of high-speed rotating components such as the engine. Summary of the invention

[0007] The present invention provides a point cloud matching method and device for an aero-engine rotor based on adaptive weights, which solves the problem existing in the traditional ICP matching algorithm. When matching the point clouds of two mating surfaces with machining errors, since the algorithm itself is an unconstrained matching algorithm and the core of the algorithm is to minimize the average Euclidean distance between corresponding points, mutual interference will occur between the mating surfaces after matching, that is, the point clouds of the mating surfaces penetrate each other, and there is a large difference between the matching result and the actual assembly result, so it cannot be directly applied to the calculation of the assembly posture.

[0008] The point cloud matching method for an aero-engine rotor based on adaptive weights according to the present invention includes the following steps:

[0009] Point set setting step: Obtain the three-dimensional point cloud models of the convex stop and concave stop of the rotor component of the aero-engine to be matched respectively. Take one of the three-dimensional point cloud models as the source data point set and the other three-dimensional point cloud model as the target data point set;

[0010] Initial matrix setting step: Set the initial rotation transformation matrix and the initial translation transformation matrix respectively;

[0011] Iterative optimization step: Based on the initial rotation transformation matrix and the initial translation transformation matrix, perform iterative optimization to minimize the weighted Euclidean distance between the source data point set and the target data point set, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration; wherein, in the calculation process of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce constraint conditions to suppress the generation of interference phenomena:

[0012] For two corresponding points that interfere in the source data point set and the target data point set in each iteration, they have a larger weight in the calculation process of the weighted Euclidean distance;

[0013] For two corresponding points that do not interfere in the source data point set and the target data point set in each iteration, they have a smaller weight in the calculation process of the weighted Euclidean distance;

[0014] Matching completion step: According to the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration, obtain the final pose transformation matrix; Use the final pose transformation matrix to align the convex stop and the concave stop to complete the matching of the three-dimensional point cloud models of the convex stop and the concave stop.

[0015] Further, a preferred implementation manner is provided. In the point set setting step:

[0016] Take the three-dimensional point cloud model of the convex stop of the rotor component of the aero-engine to be matched as the source data point set;

[0017] Take the 3D point cloud model of the concave notch of the rotor component of the aero-engine to be matched as the target data point set.

[0018] Further, a preferred embodiment is provided. Each iteration process of the iterative optimization step includes: a rotation and translation transformation step, a point cloud fine matching step, an accuracy error judgment step, and an iteration end judgment step.

[0019] Rotation and translation transformation step: Use the optimal rotation transformation matrix and the optimal translation transformation matrix of the previous iteration to perform a pose transformation on the source data point set of the previous iteration to obtain the source data point set of this iteration; wherein, the initial rotation transformation matrix is the optimal rotation transformation matrix of the 0th iteration, the initial translation transformation matrix is the optimal translation transformation matrix of the 0th iteration, and the source data point set obtained in the point set setting step is the source data point set of the 0th iteration.

[0020] Point cloud fine matching step: Calculate the weighted Euclidean distance between the source data point set of this iteration and the target data point set, minimize the weighted Euclidean distance, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix of this iteration; the optimal rotation transformation matrix and the optimal translation transformation matrix of this iteration form the pose transformation matrix of this iteration.

[0021] Accuracy error judgment step: Calculate the mean value of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of this iteration as the accuracy error judgment mean value.

[0022] Iteration end judgment step: Judge whether the number of iterations has reached the given maximum number of iterations, and judge whether the accuracy error judgment mean value obtained in this iteration is greater than the given error threshold:

[0023] If any one of the above two judgments is yes, the iteration ends; otherwise, the iteration continues.

[0024] Further, a preferred embodiment is provided. The point cloud fine matching step is as follows:

[0025]

[0026] Wherein:

[0027] p i is the coordinate value of the i-th data point in the source data point set of this iteration; q i is the coordinate value of the i-th data point in the target data point set; is the normal vector at the data point q i in the target data point set; is the vector pointing from the data point q i in the target data point set to the data point p i in the source data point set; Ai <0 means and the included angle between them is greater than 90°, p i and q i interfere; A i ≥0 means and the included angle between them is less than or equal to 90°, p i and q i do not interfere; K i is the adaptive weight coefficient of p i and q i ; N is the total number of matched data points; minf(R k , T k ) means to minimize the weighted Euclidean distance between the source data point set and the target data point set of this iteration to obtain the optimal rotation transformation matrix R k and the optimal translation transformation matrix T k ; k represents the current iteration number.

[0028] Furthermore, a preferred embodiment is provided, and the accuracy error judgment step is as follows:

[0029]

[0030] Wherein:

[0031] p j is the coordinate value of the j-th data point in the source data point set of this iteration; q j is the coordinate value of the j-th data point in the target data point set; e k is the mean value of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of this iteration, as the mean value for accuracy error judgment.

[0032] Furthermore, a preferred embodiment is provided, and the initial matrix setting step includes a point cloud rough matching step:

[0033] According to the prior knowledge of on-site machining of the rotor components of aero-engines, perform point cloud rough matching on the three-dimensional point cloud models of the convex and concave stop ports to obtain the initial rotation transformation matrix and the initial translation transformation matrix.

[0034] The present invention also proposes an aero-engine rotor point cloud matching device based on adaptive weights, and the device includes the following modules:

[0035] Point set setting module: respectively obtain the three-dimensional point cloud models of the convex and concave stop ports of the rotor components of the aero-engine to be matched, and use one of the three-dimensional point cloud models as the source data point set and the other three-dimensional point cloud model as the target data point set;

[0036] Initial matrix setting module: Set the initial rotation transformation matrix and the initial translation transformation matrix respectively;

[0037] Iterative optimization module: Based on the initial rotation transformation matrix and the initial translation transformation matrix, perform iterative optimization to minimize the weighted Euclidean distance between the source data point set and the target data point set, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration; wherein, during the calculation of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce constraint conditions to suppress the generation of interference phenomena:

[0038] For two corresponding points that interfere in the source data point set and the target data point set during each iteration, they have a larger weight during the calculation of the weighted Euclidean distance;

[0039] For two corresponding points that do not interfere in the source data point set and the target data point set during each iteration, they have a smaller weight during the calculation of the weighted Euclidean distance;

[0040] Matching completion module: According to the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration, obtain the final pose transformation matrix; Use the final pose transformation matrix to align the convex rabbet and the concave rabbet, and complete the matching of the three-dimensional point cloud models of the convex rabbet and the concave rabbet.

[0041] The present invention also proposes a computer device, including: a processor and a memory, the memory is used to store the executable instructions of the processor, and the processor is configured to execute the above-mentioned adaptive weight-based aero-engine rotor point cloud matching method according to any one of the above by executing the executable instructions.

[0042] The present invention also proposes a computer storage medium, in which a computer program is stored, and when the computer program runs, it executes the above-mentioned adaptive weight-based aero-engine rotor point cloud matching method according to any one of the above.

[0043] The present invention also proposes a computer program product, including computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned adaptive weight-based aero-engine rotor point cloud matching method according to any one of the above are implemented.

[0044] The present invention has the following beneficial effects:

[0045] 1. The adaptive weight-based aero-engine rotor point cloud matching method of the present invention solves the problem of rigid body interference generated during the matching of the engine rotor by adding an adaptive weight coefficient.

[0046] 2. The adaptive weight-based aero-engine rotor point cloud matching method of the present invention performs rough matching with prior knowledge, and solves the problem of local optimization that is prone to occur during conventional matching.

[0047] The method and device for matching point clouds of an aero-engine rotor based on adaptive weights according to the present invention are applicable to completing the matching of large-scale point clouds of an aero-engine rotor stack and guiding the stacking of the engine rotor. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments 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. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 It is a flowchart of the method for matching point clouds of an aero-engine rotor based on adaptive weights in an embodiment of the present invention. Detailed Embodiments

[0050] To describe the technical solutions and advantages of the present invention more clearly, the detailed embodiments of the present invention will be further described in detail and completely below in conjunction with the drawings. The described embodiments are only some preferred embodiments of the present invention, rather than all the implementation manners; the described embodiments are intended to explain the present invention and should not be construed as a limitation to the present invention; the reasonable combination of the technical features defined in each embodiment of the present invention, and all other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts, fall within the scope of protection of the present invention.

[0051] In one embodiment, a method for matching point clouds of an aero-engine rotor based on adaptive weights is provided, and the method includes the following steps:

[0052] Point set setting step: respectively obtain the three-dimensional point cloud models of the convex and concave stops of the rotor components of the aero-engine to be matched, and use one of the three-dimensional point cloud models as the source data point set and the other three-dimensional point cloud model as the target data point set;

[0053] Initial matrix setting step: respectively set the initial rotation transformation matrix and the initial translation transformation matrix;

[0054] Iterative optimization step: perform iterative optimization based on the initial rotation transformation matrix and the initial translation transformation matrix to minimize the weighted Euclidean distance between the source data point set and the target data point set, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration; wherein, in the calculation process of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce a constraint condition to suppress the generation of interference phenomena:

[0055] For two corresponding points that interfere in the source data point set and the target data point set in each iteration, they have a relatively large weight in the calculation process of the weighted Euclidean distance;

[0056] For two corresponding points that do not interfere in the source data point set and the target data point set in each iteration, they have a relatively small weight in the calculation process of the weighted Euclidean distance;

[0057] Matching completion step: According to the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration, obtain the final pose transformation matrix; Use the final pose transformation matrix to align the male rabbet and the female rabbet, and complete the matching of the three-dimensional point cloud models of the male rabbet and the female rabbet.

[0058] In this embodiment, the matching result of the three-dimensional point cloud models of the male rabbet and the female rabbet can be used to guide the stacking of the engine rotor.

[0059] In this embodiment, aiming at the interference problem (rigid body) caused by the traditional ICP algorithm, the generation of interference phenomena is suppressed by introducing constraint conditions. The specific principle is: By introducing an adaptive weight coefficient in the calculation process of the weighted Euclidean distance:

[0060] That is, for two corresponding points that interfere, their weights account for a relatively large proportion in the calculation process of the weighted Euclidean distance;

[0061] For two corresponding points that do not interfere, their weights account for a relatively small proportion in the calculation process of the weighted Euclidean distance.

[0062] In this embodiment, the introduction of the adaptive weight coefficient can force the rotation and translation transformation matrix of the target data point set (or the target point cloud model) to iteratively move in the direction of less interference until the generation of interference phenomena is completely suppressed, so that the final matching result can accurately predict the actual assembly pose.

[0063] In this embodiment, the male rabbet is a typical geometric structure in the rotor component, and is usually used to cooperate with other components (such as the female rabbet) to ensure the accuracy and stability of the assembly.

[0064] In this embodiment, the three-dimensional point cloud model is the surface geometric data of the male rabbet or the female rabbet obtained by three-dimensional scanning or measurement technology, which contains a large number of discrete point coordinates and is used to describe its spatial shape.

[0065] In addition, in one embodiment, in the step of setting the point set:

[0066] Take the three-dimensional point cloud model of the male rabbet of the rotor component of the aeroengine to be matched as the source data point set;

[0067] Take the three-dimensional point cloud model of the female rabbet of the rotor component of the aeroengine to be matched as the target data point set.

[0068] In this embodiment, the three-dimensional point cloud model of the convex rabbet is used as the source data point set, which is the object to be matched. The three-dimensional point cloud model of the concave rabbet is used as the target data point set. The matching objective is to transform the source data point set to the position that best aligns with the target data point set through a rotation transformation matrix and a translation transformation matrix.

[0069] In this embodiment, the precise matching of the convex rabbet and the concave rabbet is the key to ensuring the assembly accuracy of the rotor stack.

[0070] In this embodiment, through the adjustment of the adaptive weight coefficient, the geometric interference between the convex rabbet and the concave rabbet during the matching process is avoided, ensuring that the matching result meets the actual assembly requirements.

[0071] In addition, in one embodiment, each iteration process of the iterative optimization step includes: a rotation and translation transformation step, a precise point cloud matching step, an accuracy error judgment step, and an iteration end judgment step;

[0072] Rotation and translation transformation step: Perform a pose transformation on the source data point set of the previous iteration using the optimal rotation transformation matrix and the optimal translation transformation matrix of the previous iteration to obtain the source data point set of the current iteration; wherein, the initial rotation transformation matrix is the optimal rotation transformation matrix of the 0th iteration, the initial translation transformation matrix is the optimal translation transformation matrix of the 0th iteration, and the source data point set obtained in the point set setting step is the source data point set of the 0th iteration;

[0073] Precise point cloud matching step: Calculate the weighted Euclidean distance between the source data point set of the current iteration and the target data point set, minimize the weighted Euclidean distance, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix of the current iteration; the optimal rotation transformation matrix and the optimal translation transformation matrix of the current iteration form the pose transformation matrix of the current iteration;

[0074] Accuracy error judgment step: Calculate the mean value of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of the current iteration as the mean value for accuracy error judgment;

[0075] Iteration end judgment step: Judge whether the number of iterations has reached the given maximum number of iterations, and judge whether the mean value of the accuracy error judgment obtained in the current iteration is greater than the given error threshold:

[0076] If any one of the above two judgments is yes, the iteration ends; otherwise, the iteration continues.

[0077] In this embodiment, in the rotation and translation transformation step:

[0078] Let the optimal rotation transformation matrix of the previous iteration be R k-1, the optimal translation transformation matrix of the previous iteration is T k-1 , the source data point set of the previous iteration is P k-1 , the source data point set of this iteration is P k , then:

[0079] P k =R k-1 P k-1 +T k-1 。

[0080] In this embodiment, if the optimal rotation transformation matrix of this iteration is R k , the optimal translation transformation matrix of this iteration is T k , then the pose transformation matrix of this iteration is [R k ,T k 。

[0081] In this embodiment, the function of the pose transformation matrix:

[0082] Rotation transformation matrix: Describes the rotation change of the source point set in each iteration.

[0083] Translation transformation matrix: Describes the translation change of the source point set in each iteration.

[0084] Pose transformation matrix: Combines rotation and translation transformations, representing the overall pose change of the source point set in each iteration.

[0085] In this embodiment, in each iteration, the algorithm gradually adjusts the pose of the source point set to align it with the target point set by optimizing the rotation transformation matrix and the translation transformation matrix.

[0086] In addition, in one embodiment, in the matching completion step, according to the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration, the final pose transformation matrix is obtained as follows:

[0087] Multiply the pose transformation matrices of each iteration to obtain the final pose transformation matrix.

[0088] The final pose transformation matrix is represented as [R,T].

[0089] In this embodiment, the final pose transformation matrix represents the cumulative transformation from the initial pose to the final matching pose:

[0090] Pose accumulation: By multiplying the transformation matrices of each iteration, accumulate the rotation and translation changes in all iterations to ensure the accuracy of the final pose.

[0091] Global optimization: The final pose transformation matrix reflects the global optimization process of the source point set from the initial pose to the final matching pose.

[0092] Actual assembly guidance: The final pose transformation matrix can be directly used for actual assembly to guide the precise alignment of the rotor stack.

[0093] In addition, in one embodiment, the point cloud fine matching steps are as follows:

[0094]

[0095] Where:

[0096] p i is the coordinate value of the i-th data point in the source data point set of this iteration; q i is the coordinate value of the i-th data point in the target data point set; is the normal vector at the data point q i in the target data point set; is the vector pointing from the data point q i in the target data point set to the data point p i in the source data point set; A i <0 means and the included angle between them is greater than 90°, and p i and q i interfere; A i ≥0 means and the included angle between them is less than or equal to 90°, and p i and q i do not interfere; K i is the adaptive weight coefficient of p i and q i ; N is the total number of matched data points; minf(R k , T k ) means to minimize the weighted Euclidean distance between the source data point set and the target data point set of this iteration to obtain the optimal rotation transformation matrix R k and the optimal translation transformation matrix T k ; k represents the current iteration number.

[0097] In this embodiment, in the point cloud fine matching step:

[0098] When A i <0, it means and the included angle between them is greater than 90°, that is, p i and q i interfere. Therefore, the value of the adaptive weight coefficient K i takes a number greater than 1 and undergoes adaptive matching with the virtual assembly result later;

[0099] When A iWhen it is ≥ 0, it means and the included angle between them is less than or equal to 90°, that is, p i and q i do not interfere with each other. Therefore, the adaptive weight coefficient K i has a value of 1.

[0100] In this embodiment, in the point cloud fine matching step, it is crucial to reasonably select the value of the adaptive weight coefficient K i : when the value of the adaptive weight coefficient K i is too small, the interference phenomenon during matching cannot be well suppressed; while when the value of the adaptive weight coefficient K i is too large, the matching accuracy will be reduced, and even the iteration cannot converge.

[0101] To address this problem, an adaptive weight coefficient assembly simulation was performed on the point clouds of the two engine rotors, setting the range of the adaptive weight coefficient, and obtaining the simulation results under different adaptive weight coefficients.

[0102] The simulation results show that as the adaptive weight coefficient increases, the interference phenomenon between the two mating surfaces in the matching result is gradually suppressed, and only a very small number of points interfere.

[0103] Thus, it can be seen that by adding an adaptive weight coefficient to the objective function (i.e., minf(R k , T k )) on the basis of the traditional ICP matching algorithm (furthermore, combining the rough matching of point clouds through prior knowledge), the problem of mutual interference between the mating surfaces after matching can be solved, making the matching result closer to the true contact state between parts.

[0104] In this embodiment, minimizing the weighted Euclidean distance means that the alignment error between the source point set and the target point set is minimized, thereby improving the matching accuracy.

[0105] In addition, in one embodiment, the accuracy error judgment step is as follows:

[0106]

[0107] Where:

[0108] p j is the coordinate value of the jth data point in the source data point set of this iteration; q j is the coordinate value of the jth data point in the target data point set; e k is the average value of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of this iteration, as the mean value for accuracy error judgment.

[0109] In this embodiment, the mean of the precision error judgment is used to evaluate the matching error between the source data point set and the target data point set in the current iteration.

[0110] In addition, in one embodiment, the initial matrix setting step includes a rough point cloud matching step:

[0111] According to the prior knowledge of on-site machining of the rotor components of an aero-engine, a rough point cloud matching is performed on the three-dimensional point cloud models of the convex and concave stop ports to obtain an initial rotation transformation matrix and an initial translation transformation matrix.

[0112] It should be noted that in the virtual assembly process, since it is the point cloud matching between actual engine rotors, it is often locally optimal due to its non-linearity. Therefore, in this embodiment, when importing the point cloud model, the prior knowledge of on-site machining is considered, and a rough point cloud matching is performed on it to avoid falling into the local optimum and achieve the global optimum as much as possible.

[0113] In this embodiment, considering the prior knowledge of on-site machining, a rough point cloud matching is performed on it to obtain an initial rotation transformation matrix and an initial translation transformation matrix, which mainly play the following key roles in the subsequent steps:

[0114] Provide an initial starting point for algorithm optimization and accelerate convergence: The iterative process of the traditional ICP algorithm is sensitive to the initial value. If the initial value is far from the true pose, it may lead to slow convergence or even failure. By providing a reasonable initial pose estimate through prior knowledge, the number of iterations of subsequent fine matching is significantly reduced, and the computational efficiency is improved.

[0115] Avoid falling into the local optimum and enhance global convergence: In complex non-linear scenarios (such as the stacking of aero-engine rotors), the traditional ICP algorithm is prone to falling into the local optimum solution due to the deviation of the initial value. By pre-aligning the point cloud through rough matching, the pose search space is reduced, and the algorithm is guided to iterate in the direction of the global optimum, avoiding incorrect convergence caused by the initial deviation.

[0116] Suppress rigid body interference and improve matching accuracy: This method adjusts the objective function through an adaptive weight coefficient, but the accuracy of the initial pose directly affects the interference suppression effect. Reasonable initial values of the rotation transformation matrix and the translation transformation matrix can reduce the initial interference area, making the weighted optimization process more efficiently correct the remaining small deviations, and finally achieve high-precision interference-free matching.

[0117] Support the engineering feasibility of actual assembly: In actual assembly, the physical pose of the rotor is usually within a certain range (such as machining tolerance limits). The initial rotation transformation matrix and translation transformation matrix obtained by rough matching with prior knowledge ensure that the algorithm starts from an engineering feasible pose range, avoiding generating an assembly scheme that is theoretically optimal but actually unachievable.

[0118] Summary: The initial rotation transformation matrix and the initial translation transformation matrix obtained by considering the prior knowledge of on-site processing are not only the starting point for algorithm optimization, but also the bridge connecting prior knowledge and refined matching. By reducing the problem complexity and enhancing the convergence stability, they ultimately achieve efficient and high-precision virtual assembly of aero-engine rotors.

[0119] In addition, in one embodiment, there is provided a point cloud matching device for aero-engine rotors based on adaptive weights. The device includes the following modules:

[0120] Point set setting module: respectively obtain the three-dimensional point cloud models of the convex and concave stops of the rotor components of the aero-engine to be matched, and use one of the three-dimensional point cloud models as the source data point set and the other as the target data point set;

[0121] Initial matrix setting module: respectively set the initial rotation transformation matrix and the initial translation transformation matrix;

[0122] Iterative optimization module: perform iterative optimization based on the initial rotation transformation matrix and the initial translation transformation matrix to minimize the weighted Euclidean distance between the source data point set and the target data point set, and obtain the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration; wherein, during the calculation of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce constraint conditions to suppress the generation of interference phenomena:

[0123] For two corresponding points that interfere in the source data point set and the target data point set during each iteration, they have a relatively large weight during the calculation of the weighted Euclidean distance;

[0124] For two corresponding points that do not interfere in the source data point set and the target data point set during each iteration, they have a relatively small weight during the calculation of the weighted Euclidean distance;

[0125] Matching completion module: obtain the final pose transformation matrix according to the optimal rotation transformation matrix and the optimal translation transformation matrix for each iteration; use the final pose transformation matrix to align the convex and concave stops, and complete the matching of the three-dimensional point cloud models of the convex and concave stops.

[0126] The above further describes the technical solutions provided by the present invention through several specific embodiments to highlight the advantages and beneficial effects of the technical solutions provided by the present invention. However, the above several specific embodiments are not used as limitations to the present invention. Any reasonable modifications and improvements to the present invention, reasonable combinations of implementation manners, and equivalent replacements within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An aero-engine rotor point cloud matching method based on adaptive weights, characterized in that: The method comprises the following steps: Point set setting step: respectively obtaining three-dimensional point cloud models of the convex stop and the concave stop of the rotor component of the aircraft engine to be matched, taking one of the three-dimensional point cloud models as the source data point set, and the other three-dimensional point cloud model as the target data point set; Initial matrix setting steps: set the initial rotation transformation matrix and the initial translation transformation matrix respectively; Iterative optimization steps: Iterative optimization is performed based on the initial rotation transformation matrix and the initial translation transformation matrix to minimize the weighted Euclidean distance between the source data point set and the target data point set, and the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration are obtained; in the calculation process of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce constraints to suppress the occurrence of interference: For two corresponding points that interfere with the source data point set and the target data point set in each iteration, they have a larger weight in the weighted Euclidean distance calculation process; For two corresponding points in the source data point set and the target data point set that do not interfere with each other in each iteration, they have a smaller weight in the weighted Euclidean distance calculation process; Matching completion steps: According to the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration, the final posture transformation matrix is ​​obtained; the final posture transformation matrix is ​​used to align the convex stop and the concave stop, and the matching of the three-dimensional point cloud models of the convex stop and the concave stop is completed.

2. The method for matching aero-engine rotor point cloud based on adaptive weights according to claim 1, characterized in that: In the point set setting step: The three-dimensional point cloud model of the convex stop of the rotor component of the aircraft engine to be matched is used as a source data point set; The three-dimensional point cloud model of the concave stop of the rotor component of the aircraft engine to be matched is used as the target data point set.

3. The method for matching aero-engine rotor point cloud based on adaptive weights according to claim 1, characterized in that: Each iteration process of the iterative optimization step includes: a rotation and translation transformation step, a point cloud precise matching step, a precision error judgment step, and an iteration end judgment step; Rotation and translation transformation step: use the optimal rotation transformation matrix and optimal translation transformation matrix of the previous iteration to transform the pose of the source data point set of the previous iteration to obtain the source data point set of this iteration; wherein, the initial rotation transformation matrix is ​​the optimal rotation transformation matrix of the 0th iteration, the initial translation transformation matrix is ​​the optimal translation transformation matrix of the 0th iteration, and the source data point set obtained in the point set setting step is the source data point set of the 0th iteration; Point cloud precise matching steps: Calculate the weighted Euclidean distance between the source data point set and the target data point set of this iteration, minimize the weighted Euclidean distance, and obtain the optimal rotation transformation matrix and optimal translation transformation matrix of this iteration; the optimal rotation transformation matrix and optimal translation transformation matrix of this iteration constitute the pose transformation matrix of this iteration; Precision error judgment step: Calculate the mean of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of this iteration as the precision error judgment mean; Iteration end judgment step: judge whether the number of iterations reaches the given maximum number of iterations, and judge whether the mean value of the accuracy error obtained in this iteration is greater than the given error threshold: If the result of any of the above two judgments is yes, the iteration ends; otherwise, the iteration continues.

4. The method for matching aero-engine rotor point cloud based on adaptive weights according to claim 3, characterized in that: The point cloud precise matching steps are as follows: in: p i is the coordinate value of the i-th data point in the source data point set of this iteration; q i is the coordinate value of the i-th data point in the target data point set; Set data point q as the target data point i The normal vector at ; is the data point q in the target data point set i Points to data point p in the source data point set i A vector; i <0 means and The angle between them is greater than 90°, p i and q i Interference occurs; A i ≥0 means and The angle between them is less than or equal to 90°, p i and q i No interference occurred; K i For p i and q i The adaptive weight coefficient; N is the total number of matched data points; minf(R k ,T k ) represents minimizing the weighted Euclidean distance between the source data point set and the target data point set of this iteration to obtain the optimal rotation transformation matrix R of this iteration k and the optimal translation transformation matrix T k ; k represents the current iteration number.

5. The method for matching aero-engine rotor point cloud based on adaptive weights according to claim 3, characterized in that: The steps of determining the accuracy error are as follows: in: p j is the coordinate value of the jth data point in the source data point set of this iteration; q j is the coordinate value of the jth data point in the target data point set; e k The mean of the Euclidean distances between each point in the target data point set and its nearest point in the source data point set of this iteration is used as the mean value for accuracy error judgment.

6. The method for matching aero-engine rotor point cloud based on adaptive weights according to claim 1, characterized in that: The initial matrix setting step includes a point cloud rough matching step: According to the prior knowledge of on-site processing of rotor parts of aircraft engines, the three-dimensional point cloud models of the convex stop and the concave stop are roughly matched to obtain the initial rotation transformation matrix and the initial translation transformation matrix.

7. The aircraft engine rotor point cloud matching device based on adaptive weights is characterized by: The device comprises the following modules: Point set setting module: respectively obtain the 3D point cloud models of the convex stop and the concave stop of the rotor component of the aircraft engine to be matched, use one of the 3D point cloud models as the source data point set, and the other 3D point cloud model as the target data point set; Initial matrix setting module: set the initial rotation transformation matrix and initial translation transformation matrix respectively; Iterative optimization module: Iterative optimization is performed based on the initial rotation transformation matrix and the initial translation transformation matrix to minimize the weighted Euclidean distance between the source data point set and the target data point set, and the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration are obtained; in the calculation process of the weighted Euclidean distance, an adaptive weight coefficient is introduced to introduce constraints to suppress the occurrence of interference phenomena: For two corresponding points that interfere with the source data point set and the target data point set in each iteration, they have a larger weight in the weighted Euclidean distance calculation process; For two corresponding points in the source data point set and the target data point set that do not interfere with each other in each iteration, they have a smaller weight in the weighted Euclidean distance calculation process; Matching completion module: obtain the final posture transformation matrix based on the optimal rotation transformation matrix and the optimal translation transformation matrix of each iteration; use the final posture transformation matrix to align the convex stop and the concave stop to complete the matching of the three-dimensional point cloud models of the convex stop and the concave stop.

8. A computer device comprising: A processor and a memory, characterized in that the memory is used to store executable instructions of the processor, and the processor is configured to execute the adaptive weight-based aero-engine rotor point cloud matching method described in any one of claims 1 to 6 by executing the executable instructions.

9. A computer storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is run, the method for matching aircraft engine rotor point clouds based on adaptive weights as described in any one of claims 1 to 6 is executed.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for matching aero-engine rotor point cloud based on adaptive weights as described in any one of claims 1 to 6 are implemented.

Images