Robot posture optimization and follow-up control method and system for free-form surface repair

Abstract

The present invention provides a robot posture optimization and follow-up control method and system for free-form surface repair. The system includes a wear surface reverse modeling module, a wear surface initial normal vector calculation module, an initial processing trajectory and normal vector extraction module, a normal vector rotation angle polynomial regression noise reduction module, a processing trajectory position and posture NURBS fitting module, a position and posture trajectory curvature adaptive segmentation module, a segmented trajectory position and posture synchronous acceleration and deceleration look-ahead module, and a robot motion and laser process collaborative control module. The present invention realizes posture optimization and follow-up control during robot cladding process to ensure the processing quality of laser repair.

Description

Technical Field

[0001] The present invention belongs to the technical field of precision measurement, and in particular relates to a robot posture optimization and follow-up control method and system for free-form surface repair. Background Art

[0002] Free-form surfaces have smooth shapes and good aerodynamic properties, and are often used as special functional surfaces in key core equipment, such as aviation turbine engines, automobile engines, and precision molds. Such core functional components usually work in harsh environments with high temperature, high pressure, overload, high impact, vibration, and fatigue. After long-term service, they are prone to wear and failure, resulting in performance degradation or even scrapping of core equipment. Since special engineering materials and relatively complex processing techniques are often required in the manufacturing process of such parts, compared to remanufacturing new workpieces, it is energy-saving, material-saving, time-saving, economical, and environmentally friendly to remanufacture defective parts through robot laser cladding to restore their functions and performance based on the faulty parts.

[0003] In order to achieve a stable distribution of spot energy during the cladding process, the laser cladding nozzle is usually required to be dynamically perpendicular to the surface to be processed. However, since the surface morphology of the free-form surface after wear has changed significantly compared to the standard design model, the actual workpiece has undergone a significant change in shape and size at a macro level. The surface normal vector of the standard design model cannot accurately describe the distribution of the surface normal vector of the actual workpiece, and the actual workpiece lacks an accurate reference model to guide trajectory planning during the repair process; at a micro level, the micro-morphology and roughness of the actual workpiece surface have changed significantly, and the worn surface is distributed with a large number of tiny pits and protrusions. Strictly adhering to the laser cladding nozzle being perpendicular to the processing surface will cause the robot to have severe posture jitter during the laser cladding process, seriously affecting the smoothness and stability of the repair processing movement, and ultimately causing a serious decline in processing quality. In addition, due to the complex surface geometry of the free-form surface workpiece, the curvature of the position and posture trajectory during the repair process varies greatly. The traditional constant processing speed can easily cause mechanical shock in areas with excessive local curvature, which often causes the robot to shake, vibrate, or even be damaged. The traditional constant speed interpolation control method cannot adapt to the free-form surface repair scenario. Therefore, the dynamic adjustment and smooth control of the posture of the laser cladding nozzle during the repair process is one of the key difficulties in the robot free-form surface repair process.

[0004] The existing method proposes a robot processing path generation method and system for laser cladding. This method extracts the robot position and posture information by parsing the G-code data, and then uses the straight line command to generate the robot processing trajectory. However, this invention does not take into account the free-form surface wear repair scenario. Compared with the standard design model, the shape, size and surface morphology of the actual workpiece have undergone tremendous changes, and the original design model cannot accurately guide the repair processing of the workpiece after wear. In addition, the straight line type processing trajectory used by this method has the disadvantages of low precision, poor continuity, and insufficient smoothness, which cannot meet the precise tracking of the surface of complex surface models in the free-form surface repair scenario. At the same time, this method does not fully consider the problem of mechanical shock that is prone to occur in areas with excessive local curvature at a constant processing speed, and it is difficult to ensure the smoothness and stability of the position and posture movement during the repair process.

[0005] In summary, in the free-form surface laser repair scenario, since the actual workpiece surface undergoes significant changes in shape, size and surface morphology compared to the standard design model, and the existing methods do not fully consider the jitter problem of the normal vector of the worn workpiece surface, there is currently no good robot cladding posture optimization and follow-up control method to achieve dynamic posture adjustment and smooth motion control of the laser cladding nozzle. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art, the present invention provides a robot posture optimization and follow-up control method and system for free-form surface repair in order to realize dynamic posture adjustment and smooth motion control of the laser cladding nozzle in the free-form surface repair scenario.

[0007] The technical solution of the present invention is as follows:

[0008] A robot posture optimization and follow-up control method for free-form surface repair, the method comprising the following steps:

[0009] Step 1: Collect the contour of the worn free-form surface and reverse model it to obtain the size, shape and surface micro-morphology characteristics of the actual workpiece surface;

[0010] Step 2: De-noise and sparsely process the reverse reconstructed wear surface, and then obtain the original normal vector distribution of the reverse surface through statistical calculation;

[0011] Step 3: Based on the actual model of the reverse reconstructed worn surface, the repair machining trajectory is designed, and according to the distribution of the original normal vector, the designed discrete trajectory point position information and the initial normal vector information are combined to form 6-dimensional machining position and normal vector trajectory data;

[0012] Step 4: Establish the trajectory plane of each processing trajectory, introduce the normal vector rotation angle and use the polynomial fitting denoising method with a sliding window to obtain the optimized processing position and normal vector trajectory data;

[0013] Step 5: Introduce 6-dimensional NURBS curves to fit discrete robot position and posture data to generate high-precision, smooth, and synchronized robot motion position and posture trajectories;

[0014] Step 6: Based on the curvature of the machining trajectory in the position space and the posture space, the machining position and posture trajectory is adaptively divided into dangerous segments and safe segments, and an appropriate machining speed is allocated;

[0015] Step 7: Based on the S-curve acceleration and deceleration law, the position and attitude synchronization look-ahead algorithm calculates the node speed of each segment through reverse deceleration iteration and forward acceleration iteration, so as to achieve a fast and smooth transition of different processing speeds between the safe segment and the dangerous segment in the position space and attitude space;

[0016] Step 8: During the laser repair process, the actual movement speed of the robot is dynamically monitored and the laser cladding process parameters are adjusted online to ensure the thermodynamic stability of the laser cladding molten pool during the processing process and improve the quality of laser cladding repair.

[0017] Furthermore, the calculation process of the original normal vector is specifically as follows:

[0018] Firstly, the inverse reconstructed wear surface is subjected to denoising and sparse processing. The three-dimensional covariance matrix of the adjacent point set at each discrete point on the inverse model surface is statistically calculated, and then the original normal vector of each discrete point on the inverse surface is obtained through singular value SVD decomposition.

[0019] Furthermore, the designed discrete trajectory point position information and initial normal vector information are combined to form the 6-dimensional processing position and normal vector trajectory data as follows:

[0020] Based on the actual model of the reverse reconstructed worn surface, a free-form surface repair machining trajectory is designed, and then all discrete points on the machining trajectory are traversed, the original normal vector set corresponding to the neighboring point set of each trajectory point is extracted, and the unit average vector of the original normal vector set is calculated, which is used as the initial normal vector of the current discrete point; finally, the planned discrete trajectory point position information and the initial normal vector information are combined to form a 6-dimensional machining position and normal vector trajectory data.

[0021] Furthermore, the introduction of the normal vector rotation angle and the noise reduction by polynomial fitting with a sliding window are specifically as follows:

[0022] Based on the 6-dimensional processing position and normal vector trajectory data, the processing plane of each processing trajectory is established, and then the initial normal vector on the processing trajectory is projected on the plane. The normal vector rotation angle is introduced and the normal vector rotation angle at each discrete point is calculated. The initial normal vector rotation angle is fitted and denoised, and the jitter of the normal vector caused by surface microscopic pits and protrusions is removed. Then, the optimized normal vector rotation angle after denoising is inversely resolved and regressed to obtain the optimized processing position and normal vector trajectory data.

[0023] Furthermore, the generation of the robot motion position and posture trajectory is specifically as follows:

[0024] By processing the optimized processing position and normal vector trajectory data, the discrete position and posture information of the robot motion control is obtained, and then the 6-dimensional NURBS curve is introduced to fit the discrete robot position and posture points to generate a high-precision, smooth and synchronous robot motion trajectory.

[0025] Furthermore, the adaptive segmentation of the processing position and posture trajectory into dangerous segments and safe segments is specifically as follows:

[0026] Based on the fitted position and posture machining trajectory, the curvature of the trajectory in the position space and posture space is calculated respectively. According to the given curvature threshold, the fitted NURBS machining trajectory is divided into dangerous segments and safe segments, and appropriate machining speeds are allocated to them respectively to avoid mechanical impact at parts with excessive curvature in the position space and posture space, so as to ensure the smoothness and stability of the position and posture movement during the repair machining process.

[0027] Furthermore, step 7 is specifically as follows: for the segmented position and posture trajectories in the trajectory buffer queue, the adjacent segments of the current processing segment are stored in the look-ahead window, and the forward acceleration and deceleration iterations are used to calculate and finally coordinate the forward results of the position trajectory and the posture trajectory to obtain the node speeds of each segment, so as to realize the fast and smooth transition of different speeds between the safe segment and the dangerous segment in the position space and the posture space, thereby ensuring the synchronous, fast and smooth speed switching of the position movement and the posture movement during the motion control interpolation process.

[0028] The present invention also provides a robot posture optimization and follow-up control system for free-form surface repair, including: a wear surface reverse modeling module, a wear surface initial normal vector calculation module, an initial processing trajectory and normal vector extraction module, a normal vector rotation angle polynomial regression denoising module, a processing trajectory position and posture NURBS fitting module, a position and posture trajectory curvature adaptive segmentation module, a segmented trajectory position and posture synchronous acceleration and deceleration look-ahead module and a robot motion and laser process collaborative control module;

[0029] in,

[0030] The wear surface reverse modeling module is used to collect the contour of the worn free-form surface and reverse model it to obtain the size, shape and surface microscopic morphology characteristics of the actual workpiece surface;

[0031] The wear surface initial normal vector calculation module is used to calculate the three-dimensional distribution characteristics of discrete points on the reverse surface by statistical calculation, and then calculate and obtain the original normal vector of each point;

[0032] The initial processing trajectory and normal vector extraction module is used to design and plan the processing trajectory for free-form surface repair according to the process requirements of laser repair, and generate 6-dimensional discrete processing position and normal vector trajectory data according to the original normal vector;

[0033] The normal vector rotation angle polynomial regression denoising module is used to introduce the normal vector rotation angle based on the 6-dimensional processing position and normal vector data and perform fitting and denoising on the original normal vector through a polynomial regression fitting method with a sliding window, thereby removing the jitter of the normal vector caused by microscopic pits and protrusions on the surface, and ensuring the accessibility and stability of the processing trajectory posture;

[0034] The processing trajectory position and posture NURBS fitting module is used to perform data fitting on the optimized discrete processing position and posture data by introducing a 6-dimensional NURBS curve fitting method, and generate a high-precision, smooth, and continuous robot position and posture motion trajectory, which is used to repair the precise and synchronous robot position and posture control during the processing;

[0035] The position and attitude trajectory curvature adaptive segmentation module is used to perform adaptive curvature segmentation on the fitting trajectory and allocate appropriate processing speeds respectively by calculating the curvature of the trajectory in the position space and the attitude space, so as to ensure the smoothness and stability of the position and attitude movement during the repair processing;

[0036] The segmented trajectory position and posture synchronous acceleration and deceleration look-ahead module is used to obtain the speed of each segment node through reverse deceleration iteration and forward acceleration iteration calculation, so as to realize the fast and smooth transition of different speeds between the safe segment and the dangerous segment, and ensure the fast and smooth speed switching of the position movement and the posture movement during the motion control interpolation process;

[0037] The robot motion and laser process coordinated control module calculates the scanning speed of the laser nozzle relative to the worn surface, dynamically adjusts the laser cladding process parameters online, ensures the thermodynamic stability of the laser cladding molten pool during the processing, and improves the quality of laser cladding repair.

[0038] Technical effects of the present invention:

[0039] The present invention solves the problem of lack of accurate model reference when the actual workpiece undergoes a large degree of macroscopic shape and size changes in the free-form surface wear repair scenario. At the same time, considering that the worn surface is distributed with a large number of tiny pits and protrusions, the problem of poor processing quality caused by fluctuations in the normal vector of the worn surface is solved. In addition, the problem of severe mechanical impact in areas with excessive local curvature caused by traditional constant processing speed is overcome, and the dynamic adjustment and smooth control of the posture of the laser cladding nozzle during the repair process is realized, thereby improving the surface quality of the repair process.

[0040] Compared with the prior art, the present invention obtains the size, shape and microscopic surface morphology of the actual workpiece by performing a reverse reconstruction method on the worn free-form surface, which more realistically reflects the actual state of the worn workpiece. At the same time, taking into full consideration that a large number of tiny pits and protrusions are distributed on the surface of the actual workpiece under the wear scenario, strictly adhering to the principle that the laser cladding nozzle is perpendicular to the processing surface will cause the robot to have a violent posture jitter during the laser cladding process, seriously affecting the smoothness and stability of the repair processing movement. A method of normal vector rotation angle polynomial regression denoising with a sliding window is proposed, which can overcome the problem of normal vector fluctuation caused by tiny pits and protrusions, and at the same time, ensure that the laser cladding nozzle is dynamically perpendicular to the processing surface to the greatest extent. In addition, considering that the curvature of the free-form surface processing trajectory in the position space and the attitude space is variable, the fitted NURBS processing trajectory is adaptively divided into safe segments and dangerous segments, and a suitable processing speed is allocated. Based on the S-curve acceleration and deceleration law, a fast and smooth speed connection between the segmented trajectories in the position space and the attitude space is achieved, ensuring the high precision, stability and speed of the robot laser cladding processing movement. At the same time, during the laser repair process, the scanning speed of the laser nozzle relative to the worn surface is calculated online, and the laser cladding process parameters are dynamically adjusted to ensure the thermodynamic stability of the laser cladding molten pool during the processing process, further improving the quality of laser cladding repair. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0042] Figure 1 The figure is a flow chart of a robot posture optimization and follow-up control method for free-form surface repair.

[0043] Figure 2 Structural diagram of the robotic laser cladding system for free-form surface repair.

[0044] Figure 3Schematic diagram of the worn surface.

[0045] Figure 4 Schematic diagram of initial normal vector calculation.

[0046] Figure 5 Schematic diagram of scanning trajectory position and normal vector.

[0047] Figure 6 Create a schematic diagram for the trajectory plane.

[0048] Figure 7 Schematic diagram of calculating the projection normal vector rotation angle.

[0049] Figure 8 Schematic diagram of polynomial regression denoising with sliding window.

[0050] Fig. 9 Schematic diagram before and after normal vector optimization.

[0051] Fig.10 Generate schematic diagrams for optimized poses of machining trajectories.

[0052] Fig.11 Schematic diagram of NURBS fitting of machining trajectory position and posture.

[0053] Fig.12 Schematic diagram of the curvature segmentation of the machining trajectory.

[0054] Fig.13 This is a schematic diagram of the acceleration and deceleration look-ahead for segmented trajectories.

[0055] Fig.14 Schematic diagram of the acceleration and deceleration look-ahead method based on the S-curve.

[0056] Fig.15 Scanning for worn gas turbine blades.

[0057] Fig.16 Structural diagram of laser repair equipment for worn blades

[0058] Fig.17 This is the experimental diagram of the original processing trajectory for blade repair.

[0059] Fig.18 This is the experimental diagram of the initial processing normal vector of blade repair.

[0060] Fig.19 Experimental diagram of regression fitting of the original normal rotation angle for blade restoration.

[0061] Fig. 20 Experimental diagram of optimizing normal vector for blade repair processing.

[0062] Fig.21 This is a segmented experimental diagram of the blade repair position and attitude trajectory curvature.

[0063] Fig. 22 Cartesian position and attitude result diagram for blade repair speed planning.

[0064] Fig.23 Feedback result diagram for the blade repair robot joint.

[0065] Fig.24 This is a comparison chart of the blade repair effects. DETAILED DESCRIPTION

[0066] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0067] The present invention proposes a robot posture optimization and follow-up control method for free-form surface repair to achieve dynamic adjustment and smooth control of the posture of the laser cladding nozzle during the repair process. Figure 1As shown, first, the robot is driven to carry a line laser sensor to collect three-dimensional data of the wear surface, and the three-dimensional point cloud model of the wear surface is reconstructed in reverse to obtain detailed macroscopic shape, size and microscopic surface appearance information of the actual wear surface. Further, according to the given adjacent radius, the adjacent point set of each discrete point on the reverse surface is obtained, and the original normal vector information of the reverse surface is obtained by statistically calculating the covariance matrix of the adjacent point set. Further, based on the original normal vector information of the reverse surface, the initial normal vector corresponding to each discrete point on the given scanning trajectory is obtained one by one, and the position information of the original discrete trajectory point and the obtained initial normal vector information are combined to obtain the 6-dimensional initial processing trajectory position and normal vector data group. Further, based on the position and normal vector data group of each processing trajectory, the working plane of the current scanning trajectory is fitted, and the normal vector rotation angle at each discrete point is calculated. By using the proposed polynomial regression method with sliding window, the fitting denoising of the normal vector rotation angle is realized to remove the jitter of the normal vector caused by the microscopic pits and protrusions on the surface. Furthermore, according to the optimized processing position and normal vector information, the discrete position and posture trajectory points are fitted into a smooth, continuous and synchronous position and posture processing trajectory using the 6-dimensional NURBS interpolation fitting method. Furthermore, according to the curvature of the processing trajectory in the position space and the posture space, the fitted position and posture trajectory is adaptively divided into dangerous segments and safe segments, and a suitable processing speed is allocated to avoid mechanical shock caused by excessive speed at the parts with excessive curvature in the position space and the posture space. Furthermore, by using the position and posture synchronization forward-looking algorithm based on S-curve acceleration and deceleration, the segmented processing position and posture speed are reversely decelerated and iterated and forward accelerated to achieve a fast and smooth transition of the processing position and posture speed between different segments. Furthermore, during the laser repair process, by dynamically monitoring the actual movement speed of the robot, the laser cladding process parameters are adjusted online to ensure the thermodynamic stability of the laser cladding molten pool during the processing process and improve the quality of laser cladding repair.

[0068] The specific steps of reverse modeling of the worn surface described in step 1 are:

[0069] like Figure 2 As shown in the figure, the robot system for free-form surface repair consists of an industrial controller, a line laser sensor, a robot servo motor, a robot body, an I / O device, a semiconductor laser, a powder feeder, a laser cladding head, a laser cladding nozzle and other auxiliary equipment. The line laser sensor and the laser cladding head are installed at the end of the industrial robot. The robot controller controls the status of the robot servo motor and the I / O device in real time through the EtherCAT protocol. The digital analog signal of the I / O device is further used to trigger the start and stop of the laser and the powder feeder and the change of the working parameters. At the same time, the robot controller establishes communication with the line laser sensor through the TCP / IP protocol to realize the online control of the line laser sensor and the real-time upload of the collected contour.

[0070] By driving the robot to carry the line laser sensor to scan the surface of the worn workpiece to be tested, and then by periodically collecting the surface contour of the workpiece, the reverse reconstruction of the worn workpiece is completed, providing model guidance for the later surface repair. Figure 3 As shown in the figure, the worn surface is composed of discrete point cloud data M. Due to the irregular wear in the service engineering, there are irregular pits and protrusions on the reverse surface. By using the line laser sensor online scanning method, the macroscopic shape and size parameters of the curved surface workpiece after wear can be obtained, and the information such as microscopic surface pits and protrusions can also be retained, which can more accurately describe the real characteristics of the workpiece after actual wear.

[0071] The specific steps for calculating the initial normal vector of the wear surface described in step 2 are:

[0072] With a given machining accuracy δ error , the actual wear surface point cloud reconstructed by reverse reconstruction is appropriately sparse and denoised, and the burden of subsequent initial normal vector calculation is reduced by simplifying the point cloud model. Then, according to the given adjacent radius r n , get each discrete point P on the reverse reconstructed point cloud i The adjacent point set {N j},like Figure 4 As shown. By using statistical analysis methods, the adjacent point set {N j The three-dimensional covariance matrix E of i , the specific calculation method is as follows:

[0073]

[0074] in,

[0075]

[0076] Furthermore, by j The three-dimensional covariance matrix E of i Perform singular value SVD decomposition, obtain the minimum eigenvector of the covariance matrix and normalize it, which is the initial normal vector a at the point i By iterating the above algorithm, the initial normal vectors {a i}.

[0077] The specific steps of initial scanning trajectory and normal vector extraction described in step 3 are:

[0078] Based on the actual point cloud model reconstructed by reverse scanning, the laser repair trajectory is designed and planned with reference to the processing parameters of free-form surface workpiece repair, such as Figure 5According to the position coordinate information of the planned processing trajectory points, each processing trajectory point P on the reverse reconstructed point cloud is obtained. i The adjacent point set {N k}The corresponding normal vector set {a k}. Take the normal vector set {a k} and normalize it to the current trajectory point P i The corresponding normal vector a' i , the specific calculation process is as follows:

[0079]

[0080] in, The position coordinates of the current point and the normal vector Combined to form 6-dimensional position normal vector coordinate data It is used as the position and normal vector information of the initial planning trajectory for subsequent trajectory fitting and interpolation.

[0081] The specific steps of normal vector rotation angle polynomial regression denoising described in step 4 are as follows:

[0082] In order to quantify the speed of change of the normal vector of the laser cladding nozzle on each processing trajectory during the laser repair process, the normal vector rotation angle is introduced, and then the sliding polynomial regression of the normal vector rotation angle is performed to achieve smooth noise reduction of the attitude angle and remove the jitter of the normal vector caused by surface micro pits and protrusions.

[0083] First, the working plane of each trajectory is established based on the position and normal vector information of the initial planned trajectory. i The initial normal vector a' along the current point i Translate l to obtain the offset point Q i ,like Figure 6 As shown, it can be calculated by the following formula:

[0084] Q i =P i +la' i

[0085] Furthermore, each scanning trajectory discrete point {P i} and its corresponding offset point set {Q i} combination as the object of plane fitting. Assume that the trajectory plane equation is Ax+By+Cz+D=0, and the current trajectory plane parameters A, B, C, D can be obtained by iterative calculation of the least squares method. Then, the initial processing normal vector is projected on the trajectory plane to obtain the normal vector a with plane constraint. i . Obtain the projected normal vector set {a”i} and normalize it, and use it as the reference principal normal vector a of the current processing trajectory main ,like Figure 7 As shown. Further, record the projection normal vector a at each discrete point of the current scanning trajectory" i With the principal normal vector a main The angle θ between i is the normal vector rotation angle, and the cumulative arc length S is established i Normal vector rotation angle θ i Two-dimensional data, the specific calculation method is as follows:

[0086]

[0087] By using the polynomial regression method with a sliding window, the least squares method is used to obtain the polynomial fitting curve parameters within the current sliding window, which are recorded as Furthermore, in order to prevent the problem that the robot processing position and posture cannot be reached due to excessive normal vector rotation angle, a linear constraint coefficient K is added to realize excessive posture change constraints. Finally, according to the optimized projection normal vector rotation angle result Inverse solution to obtain the optimized normal vector a"' i ,like Fig. 9 The method of normal vector rotation angle polynomial regression denoising can ensure that the laser cladding nozzle is perpendicular to the worn free-form surface during the laser repair process to the greatest extent, while constraining the maximum rotation angle of the laser cladding nozzle, achieving smooth denoising of the attitude angle, removing the jitter of the normal vector caused by microscopic pits and protrusions on the surface, and ensuring the accessibility and stability of the processing trajectory attitude.

[0088] The specific steps of the processing position and posture NURBS fitting method described in step 5 are as follows:

[0089] First, according to the optimized normal vector a' i and the planned discrete trajectory points P i , get the trajectory tangent vector N i , optimize the normal vector A i And the derived vector O i , and then calculate the robot posture matrix R i , which can be calculated by the following formula:

[0090]

[0091] Among them, the trajectory tangent vector Optimize the normal vector A i =a"' i , derive the vector O i =A i ×Ni ,like Fig.10 As shown. According to the relationship between the Euler angle and the attitude matrix in the above formula, the final optimized trajectory point position and attitude coordinates can be obtained.

[0092] Introduce 6-dimensional NURBS curve to optimize the final trajectory point position and posture coordinates Perform interpolation fitting to obtain the high-order continuous robot processing position and posture trajectory, which is calculated by the following formula:

[0093]

[0094] Among them, w i is the weight of each discrete point, D i is the position and posture coordinates of each discrete point, N i,k (u) is the k-order B-spline basis function, which is calculated by the following formula:

[0095]

[0096] Through 6-dimensional NURBS curve interpolation fitting, the original discrete processing points form a smooth and continuous processing trajectory, such as Fig.11 As shown, the description of the position and posture of the laser cladding nozzle with high precision and high synchronization during the free-form surface repair process is achieved.

[0097] The specific steps of step 6 for segmenting the curvature of the machining trajectory are as follows:

[0098] Since the worn free-form surface is relatively complex, in order to achieve high-precision tracking of the worn surface, the curvature of its repair trajectory in the position space and attitude space is variable, and there is a situation where the curvature is too large in some local areas. The traditional motion control method does not fully consider the problem of excessive local curvature in the position space and attitude space during the free-form surface repair process. It adopts a constant processing speed, which can easily cause the robot to be impacted and vibrated in the part with excessive curvature, seriously affecting the stability of the repair processing movement. Therefore, in order to reduce the mechanical impact of the robot movement, the present invention adopts an adaptive position and attitude trajectory segmentation method to divide the fitted position and attitude trajectory into a dangerous segment (excessive curvature) and a safe segment, and allocates a suitable processing speed to ensure the smoothness and stability of the position and attitude movement during the repair process.

[0099] like Fig.11 As shown, the Frenet frames are established in the position space and attitude space respectively, with the current position and attitude point [C 1 (u i ), C 2 (u i )] as the origin, establish the position and attitude tangent vector T1 (u i ), T 2 (u i ), position and attitude subnormal vector B 1 (u i ), B 2 (u i ) and the position and attitude principal normal vector N 1 (u i ), N 2 (u i ), where the position and attitude tangent vector T 1 (u i ), T 2 (u i ) and the position and attitude principal normal vector N 1 (u i ), N 2 (u i ) respectively constitute the position osculating plane and the attitude osculating plane. The specific calculation method for establishing the Frenet frame is as follows:

[0100]

[0101]

[0102]

[0103] Furthermore, according to the curvature calculation formula C″(u i )=s″(u i )T(u i )+κ[s′(u i )] 2 N(u i ), where s″(u i ), s′(u i ) is the machining path length s(u i ). After simplification, the position space and trajectory space curvature can be calculated by the following formula:

[0104]

[0105] According to the given position and attitude trajectory curvature thresholds, the fitted NURBS trajectory can be adaptively segmented into dangerous segments and safe segments in the position space and attitude space, respectively, as follows: Fig.12 Furthermore, appropriate processing speeds are allocated to dangerous sections and safe sections to avoid mechanical impacts at locations with large curvatures and at posture trajectories, thus ensuring the stability and smoothness of the processing.

[0106] The specific steps of the step 7 segmented trajectory acceleration and deceleration look-ahead are as follows:

[0107] Since the fitted position and posture NURBS trajectory is adaptively divided into dangerous segments and safe segments, and different processing speeds are assigned, the speed transition between different segments is a key issue. The present invention proposes a position and posture segmented synchronous forward-looking method based on the S-curve acceleration and deceleration law to achieve fast and smooth switching of different processing speeds between segments, thereby ensuring the stability and smoothness of laser cladding processing.

[0108] like Fig.13 As shown in the figure, the adjacent segments of the current processing segment are stored in the look-ahead window. As the look-ahead window is gradually refreshed from the beginning of the queue to the end of all processing trajectories, it is necessary to complete the processing of each segment. To ensure smooth and fast transition of the speed between different segments during the processing, reverse deceleration iteration and forward acceleration iteration are performed in the position space and attitude space respectively based on the S-curve acceleration and deceleration law. The results of the look-ahead of the position space and attitude space are synchronously coordinated to obtain the starting speed and ending speed between different segments.

[0109] The reverse deceleration iteration is to ensure the safety of the processing process and realize the safety requirement of decelerating the current position and posture movement speed to zero within the look-ahead window, so as to ensure that the equipment has sufficient deceleration performance to cope with sudden emergency stops. The reverse deceleration iteration starts from the end segment of the look-ahead window to the beginning segment to obtain the maximum reverse deceleration speed v allowed by the starting speed of the current processing segment. backward , to ensure that there is enough deceleration distance during the movement. According to the S-curve acceleration and deceleration law, such as Fig.14 As shown in (1), the reference distance s is calculated referrence It is used to determine the working mode of reverse deceleration in the current section. The specific calculation process is as follows:

[0110]

[0111] Among them, v end (i) is the current end speed. The end speed of the final segment of the initial look-ahead window is 0, and the subsequent end speed is the starting speed v obtained by the previous iteration. start (i+1). max is the maximum acceleration constraint of the current segment, j max is the maximum jerk constraint of the current segment.

[0112] Further, compare the actual length s of the current segment with the reference distance s referrence There are three different acceleration and deceleration modes: Mode 1, Mode 2 and Mode 3.

[0113] Mode 1: s(i) = s referrence , during the deceleration process of the current section, the acceleration starts from 0 and reaches the maximum acceleration - a max , and then with the maximum jerk jmax is reduced to 0. Therefore, the starting speed of the current segment can be calculated by the following formula:

[0114]

[0115] Mode 2: s>s referrence , during the deceleration process of the current section, the acceleration starts from 0 and reaches the maximum acceleration - a max After that, keep it for a while, then accelerate with the maximum jerk j max is reduced to 0. Therefore, the starting speed of the current segment can be calculated by the following formula:

[0116]

[0117] Mode 3: s(i) referrence , during the deceleration process of the current section, the acceleration starts from 0 and does not reach the maximum acceleration - a max When the maximum acceleration j max is reduced to 0. Therefore, the starting speed of the current segment can be calculated by the following formula:

[0118] v start (i) = v end +j max (t 3 ) 2

[0119] in,

[0120] According to the above three modes, the starting speed v can be obtained start (i). Then, by refreshing forward, calculate the starting speed v of the next segment start (i-1), the starting speed v of the previous segment start (i) As the end speed v of the next segment end (i-1), repeatedly iterate the above algorithm until the beginning of the look-ahead window, and the final maximum reverse deceleration speed v can be obtained backward .

[0121] The calculation method of forward acceleration iteration is similar to that of reverse deceleration iteration. The current actual running speed is used as the starting speed, and the maximum end speed that can be achieved in the look-ahead window is calculated to maximize the position and posture movement speed. The forward acceleration iteration starts from the starting segment of the look-ahead window and executes to the end segment to obtain the maximum forward acceleration speed v allowed by the end speed of the current processing segment. forward , there are also three acceleration modes, such as Fig.14 (2) as shown.

[0122] Furthermore, by comparing the forward acceleration speed v forward and the magnitude of the reverse deceleration speed v​backward Relationship, take the minimum as the final node speed v node . Calculate the position node velocity v in the position space and attitude space respectively 1_node and attitude node velocity Take the minimum as the final node velocity to achieve synchronization of position and posture motion v node By proposing a position and attitude segmented synchronous look-ahead method based on the S-curve acceleration and deceleration law, smooth and fast speed switching between dangerous segments and safe segments in position space and attitude space is achieved, ensuring synchronous, smooth and high-speed control interpolation of position and attitude motion of the free-form surface wear repair trajectory.

[0123] The specific steps of step 8 of coordinated control of robot position and posture movement and laser process are as follows:

[0124] Since the fitting processing trajectory is adaptively divided into safe and dangerous segments, and different processing speeds are assigned, and there are acceleration and deceleration transitions between different segments, the processing speed of the laser cladding nozzle relative to the worn surface changes dynamically during the repair process. In order to improve the quality of laser cladding repair and ensure the thermodynamic stability of the laser cladding molten pool during the processing, it is necessary to dynamically adjust the laser parameters according to the actual scanning speed changes.

[0125] By periodically monitoring the robot's motion state, the motion speed v of the laser cladding nozzle relative to the wear surface is obtained by online calculation. nozzel (t). Further, according to the power density function definition Among them, P e is the ideal laser power, V e is the ideal scanning speed, D is the spot diameter, and the laser power parameter P(t)=E is dynamically adjusted. s ·D·v nozzel (t). By synchronously sending robot position and posture control instructions and laser control parameters, the synchronization of free-form surface repair movement and process is achieved, ensuring the thermodynamic stability of the laser cladding molten pool during processing and improving the processing quality of laser cladding repair.

[0126] In order to further illustrate the effectiveness and practicality of the present invention, an application case of the present invention, namely, the repair of a steam turbine guide blade, is used for experimental verification. Fig.15 As shown, due to long-term operation in a high temperature, high pressure and high fatigue environment, the surface of the turbine engine blades is prone to irregular wear and deformation.

[0127] Fig.16The robot laser repair experimental platform used is composed of a robot, a line laser sensor, a laser, a water cooler, a powder feeder, a laser cladding head, and a coaxial powder feeding cladding nozzle. In addition, by installing a laser tracker target ball on the laser cladding head, the robot position information is dynamically collected during the actual processing process to verify the effectiveness and practicality of the robot posture optimization and follow-up control method for free-form surface repair proposed in the present invention.

[0128] The robot is driven to carry a line laser sensor to scan the surface of the blade to be tested for wear. Fig.15 As shown in the figure, the actual shape, size and microscopic morphology of the worn blade are obtained online, such as Fig.17 At the same time, according to the process requirements of blade cladding repair, the original scanning trajectory can be obtained, and the original normal vector of the point cloud can be further calculated, and the initial normal vector information at the discrete points on the initial processing trajectory can be obtained according to the planned processing trajectory, as shown in Fig.18 As shown, it is obvious that due to the large number of tiny pits and protrusions distributed after wear, the initial normal vector has jitter and roughness problems.

[0129] Furthermore, the normal vector rotation angle polynomial regression fitting method with a sliding window proposed in the present invention is applied. By establishing a trajectory plane to project the initial processing trajectory and the normal vector, the normal vector rotation angle is calculated based on the projected normal vector and the sliding polynomial fitting method is used to achieve smoothing and de-jittering of the normal vector in the processing process. Fig.19 The following is a comparison of the normal vector rotation effects before and after normal vector de-jittering for two processing trajectories. Fig. 20 This is a normal vector result diagram after optimization using the method described in the present invention. The results show that the method proposed in the present invention can ensure that the normal vector of the laser cladding nozzle is perpendicular to the surface to be processed to the greatest extent during the free-form surface repair process. At the same time, the normal vector jitter problem caused by tiny pits and protrusions on the worn surface is avoided. The optimized laser processing posture can ensure high-precision and smooth position posture trajectory during the processing process.

[0130] Furthermore, based on the optimized processing normal vector, the processing trajectory position and posture NURBS fitting method and processing trajectory curvature adaptive segmentation method described in the present invention are applied to fit the discrete position and posture information of the worn blade repair processing into a smooth and continuous NURBS trajectory, and at the same time, according to the curvature of the processing trajectory in the position space and posture space, the processing trajectory is adaptively segmented into safe segments and dangerous segments. Furthermore, the segmented trajectory position and posture synchronous acceleration and deceleration look-ahead method and the robot motion and laser process coordinated control method described in the present invention are used to finally achieve the smooth, smooth and synchronous control of the motion position, posture and laser process of the robot repair processing, and complete the laser cladding repair processing of the blade surface wear. Fig.21This is the experimental diagram of the curvature segmentation of the blade repair position and attitude trajectory. The results show that this method can achieve smooth and continuous cladding trajectory generation, and can accurately identify the parts with excessive curvature in the position space and attitude space, and adaptively segment the processing trajectory. Fig. 22 Cartesian position and attitude result diagram for blade repair speed planning, Fig.23 The blade repair robot joint feedback result diagram shows that compared with the method without segmented interpolation, the adaptive segmentation and appropriate speed distribution method can effectively reduce the Cartesian space position, posture and actual maximum acceleration of the robot joint space in the large curvature trajectory segment during the processing, avoid the robot's jitter, vibration and mechanical shock in the large curvature trajectory segment, and ensure high-precision, stable and smooth movement during the laser repair process. Fig.24 It is a comparison diagram of the processing effects of the method without posture control and the posture optimization and follow-up control method described in this article. The results show that the posture optimization and follow-up control method proposed in the invention can ensure that the processing posture is perpendicular to the surface to be processed to the greatest extent during the free-form surface repair process. At the same time, the normal vector jitter problem caused by tiny pits and protrusions on the worn surface is avoided. Compared with the method without posture control, the repaired surface has significant improvements in indicators such as roughness, waviness and thermal deformation, which fully verifies the effectiveness and practicality of the robot posture optimization and follow-up control method for free-form surface repair proposed in the present invention.

[0131] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A robot posture optimization and follow-up control method for free-form surface repair, characterized in that: The method comprises the following steps: Step 1: Collect the contour of the worn free-form surface and reverse model it to obtain the size, shape and surface micro-morphology characteristics of the actual workpiece surface; Step 2: De-noise and sparsely process the reverse reconstructed wear surface, and then obtain the original normal vector distribution of the reverse surface through statistical calculation; Step 3: Based on the actual model of the reverse reconstructed worn surface, the repair machining trajectory is designed, and according to the distribution of the original normal vector, the designed discrete trajectory point position information and the initial normal vector information are combined to form 6-dimensional machining position and normal vector trajectory data; Step 4: Establish the trajectory plane of each processing trajectory, introduce the normal vector rotation angle and use the polynomial fitting denoising method with a sliding window to obtain the optimized processing position and normal vector trajectory data; The method of introducing the normal vector rotation angle and reducing noise by polynomial fitting with a sliding window is specifically as follows: Based on the 6-dimensional processing position and normal vector trajectory data, the processing plane of each processing trajectory is established, and then the initial normal vector on the processing trajectory is projected on the plane, the normal vector rotation angle is introduced and the normal vector rotation angle at each discrete point is calculated, and the initial normal vector rotation angle is fitted and denoised to remove the jitter of the normal vector caused by microscopic pits and protrusions on the surface, and then the optimized normal vector rotation angle after inverse regression denoising is obtained to obtain the optimized processing position and normal vector trajectory data; Step 5: Introduce 6-dimensional NURBS curves to fit discrete robot position and posture data to generate high-precision, smooth, and synchronized robot motion position and posture trajectories; Step 6: Based on the curvature of the machining trajectory in the position space and the posture space, the machining position and posture trajectory is adaptively divided into dangerous segments and safe segments, and an appropriate machining speed is allocated; Step 7: Based on the S-curve acceleration and deceleration law, the position and attitude synchronization look-ahead algorithm calculates the node speed of each segment through reverse deceleration iteration and forward acceleration iteration, so as to achieve a fast and smooth transition of different processing speeds between the safe segment and the dangerous segment in the position space and attitude space; Step 8: During the laser repair process, the actual movement speed of the robot is dynamically monitored and the laser cladding process parameters are adjusted online to ensure the thermodynamic stability of the laser cladding molten pool during the processing process and improve the quality of laser cladding repair.

2. The robot posture optimization and follow-up control method for free-form surface repair according to claim 1 is characterized in that: The calculation process of the original normal vector is specifically as follows: Firstly, the inverse reconstructed wear surface is subjected to denoising and sparse processing. The three-dimensional covariance matrix of the adjacent point set at each discrete point on the inverse model surface is statistically calculated, and then the original normal vector of each discrete point on the inverse surface is obtained through singular value SVD decomposition.

3. The robot posture optimization and follow-up control method for free-form surface repair according to claim 1 is characterized in that: The combination of the designed discrete trajectory point position information and the initial normal vector information to form the 6-dimensional processing position and normal vector trajectory data is specifically: Based on the actual model of the reverse reconstructed worn surface, a free-form surface repair machining trajectory is designed, and then all discrete points on the machining trajectory are traversed to extract the original normal vector set corresponding to the neighboring point set of each trajectory point, and the unit average vector of the original normal vector set is calculated and used as the initial normal vector of the current discrete point; Finally, the planned discrete trajectory point position information and initial normal vector information are combined to form 6-dimensional processing position and normal vector trajectory data.

4. The robot posture optimization and follow-up control method for free-form surface repair according to claim 1 is characterized in that: The generation of the robot motion position and posture trajectory is specifically as follows: By processing the optimized processing position and normal vector trajectory data, the discrete position and posture information of the robot motion control is obtained, and then the 6-dimensional NURBS curve is introduced to fit the discrete robot position and posture points to generate a high-precision, smooth and synchronous robot motion trajectory.

5. The robot posture optimization and follow-up control method for free-form surface repair according to claim 1 is characterized in that: The adaptive segmentation of the machining position and posture trajectory into dangerous segments and safe segments is specifically as follows: Based on the fitted position and posture machining trajectory, the curvature of the trajectory in the position space and posture space is calculated respectively. According to the given curvature threshold, the fitted NURBS machining trajectory is divided into dangerous segments and safe segments, and appropriate machining speeds are allocated to them respectively to avoid mechanical impact at parts with excessive curvature in the position space and posture space, so as to ensure the smoothness and stability of the position and posture movement during the repair machining process.

6. The robot posture optimization and follow-up control method for free-form surface repair according to claim 1 is characterized in that: The step 7 is specifically as follows: for the segmented position and posture trajectories in the trajectory buffer queue, the adjacent segments of the current processing segment are stored in the look-ahead window, and the look-ahead results of the position trajectory and the posture trajectory are finally coordinated through reverse deceleration iteration and forward acceleration iteration calculation to obtain the node speed of each segment, so as to realize the fast and smooth transition of different speeds between the safe segment and the dangerous segment in the position space and the posture space, thereby ensuring the synchronous, fast and smooth speed switching of the position movement and the posture movement during the motion control interpolation process.

7. A robot posture optimization and follow-up control system for free-form surface repair, characterized in that: include: Wear surface reverse modeling module, wear surface initial normal vector calculation module, initial processing trajectory and normal vector extraction module, normal vector rotation angle polynomial regression denoising module, processing trajectory position and posture NURBS fitting module, position and posture trajectory curvature adaptive segmentation module, segmented trajectory position and posture synchronous acceleration and deceleration look-ahead module and robot motion and laser process collaborative control module; in, The wear surface reverse modeling module is used to collect the contour of the worn free-form surface and reverse model it to obtain the size, shape and surface microscopic morphology characteristics of the actual workpiece surface; The wear surface initial normal vector calculation module is used to calculate the three-dimensional distribution characteristics of discrete points on the reverse surface by statistical calculation, and then calculate and obtain the original normal vector of each point; The initial processing trajectory and normal vector extraction module is used to design and plan the processing trajectory for free-form surface repair according to the process requirements of laser repair, and generate 6-dimensional discrete processing position and normal vector trajectory data according to the original normal vector; The normal vector rotation angle polynomial regression denoising module is used to introduce the normal vector rotation angle based on the 6-dimensional processing position and normal vector data and perform fitting and denoising on the original normal vector through a polynomial regression fitting denoising method with a sliding window, thereby removing the jitter of the normal vector caused by microscopic pits and protrusions on the surface and ensuring the accessibility and stability of the processing trajectory posture; The method of introducing the normal vector rotation angle and reducing noise by polynomial fitting with a sliding window is specifically as follows: Based on the 6-dimensional processing position and normal vector trajectory data, the processing plane of each processing trajectory is established, and then the initial normal vector on the processing trajectory is projected on the plane, the normal vector rotation angle is introduced and the normal vector rotation angle at each discrete point is calculated, and the initial normal vector rotation angle is fitted and denoised to remove the jitter of the normal vector caused by microscopic pits and protrusions on the surface, and then the optimized normal vector rotation angle after inverse regression denoising is obtained to obtain the optimized processing position and normal vector trajectory data; The processing trajectory position and posture NURBS fitting module is used to perform data fitting on the optimized discrete processing position and posture data by introducing a 6-dimensional NURBS curve fitting method, so as to generate a high-precision, smooth, and continuous robot position and posture motion trajectory, which is used to repair the precise and synchronous robot position and posture control during the processing; The position and attitude trajectory curvature adaptive segmentation module is used to perform adaptive curvature segmentation on the fitting trajectory and allocate appropriate processing speeds respectively by calculating the curvature of the trajectory in the position space and the attitude space, so as to ensure the smoothness and stability of the position and attitude movement during the repair processing; The segmented trajectory position and posture synchronous acceleration and deceleration look-ahead module is used to obtain the speed of each segment node through reverse deceleration iteration and forward acceleration iteration calculation, so as to realize the fast and smooth transition of different speeds between the safe segment and the dangerous segment, and ensure the fast and smooth speed switching of the position movement and the posture movement during the motion control interpolation process; The robot motion and laser process coordinated control module calculates the scanning speed of the laser nozzle relative to the worn surface, dynamically adjusts the laser cladding process parameters online, ensures the thermodynamic stability of the laser cladding molten pool during the processing, and improves the quality of laser cladding repair.

Images