Hydropower station runner chamber surface laser cladding repairing method

Abstract

The invention discloses a hydropower station runner chamber surface laser cladding repair method which comprises the following steps: S1, acquiring three-dimensional point cloud data of a wear area in a runner chamber, identifying the contour of the wear area based on the three-dimensional point cloud data, and establishing a three-dimensional wear area model; s2, on the basis of wear characteristics of the wear region contour contained in the three-dimensional wear region model, decomposing the wear region contour to obtain multiple sections of standard arcs; s3, generating a fusion path based on multiple sections of standard arcs; s4, the mechanical arm is controlled to move along the fusion path, and the laser nozzle is kept perpendicular to the abrasion area; and S5, the laser power is adjusted based on the movement speed of the mechanical arm and the real-time temperature to form a cladding layer. According to the method, the multiple standard arcs are constructed by decomposing the outline of the abrasion area, the limitation of a traditional single arc is broken through, geometric features are converted into mathematical weights in combination with a weighting function, the mechanical limitation of fixed parameter fusion is broken through, terrain self-adaptive path planning is achieved, the situation that a spray head suddenly turns is avoided, and the cladding layer quality is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laser cladding repair of runner chamber, and particularly relates to a laser cladding repair method for surface of runner chamber of hydropower station. BACKGROUND

[0002] The runner chamber of hydropower station is the operation interval of the water turbine. Due to the influence of the surface of the steel plate inside the runner chamber by the erosion of silt in the water flow, cavitation erosion, hydraulic vibration and the like for a long time, the surface of the steel plate is seriously worn. The gap between the runner chamber and the runner blade is precisely designed, and the wear will increase the gap. After the gap is increased, the water flow from the high-pressure side (outlet) of the runner to the low-pressure side (inlet) is increased, a 'volume loss' is formed, and the efficiency of the water turbine is directly reduced. Under the same water head and flow, the output power of the unit is reduced, the power generation is reduced, and economic losses are directly caused. Severe wear will change the profile of the runner chamber, destroy the optimal hydraulic flow passage, and affect the load carrying capacity of the power station. Therefore, the wear area needs to be repaired during each shutdown of the runner chamber. However, due to the particularity of the work of the runner chamber of the hydropower station, it is also impossible to frequently stop the machine to repair the surface of the steel plate of the runner chamber. Therefore, the repair cladding method is used to process the runner chamber in the prior art.

[0003] However, the laser cladding method currently applied to the inner wall repair of the runner chamber still has the following technical problems: (1) in the traditional method, when the robot moves along a single circular arc trajectory, the nozzle will move too fast on the flat section, causing the molten layer to be too thin, and the nozzle will make a sharp turn on the steep section, causing material accumulation and cracking, affecting the cladding quality. As shown in patent CN104988497B, the fixed scanning trajectory causes uneven thickness of the cladding layer, resulting in poor repair effect, thermal damage and thermal deformation; (2) the shape of the wear area of the runner chamber is irregular, which causes complex trajectory planning calculation of the robot arm and the robot arm cannot be dynamically adjusted in time, and the movement response of the robot arm is slow, so that the cladding head cannot be vertically positioned on the repair surface in real time, resulting in poor repair accuracy; (3) when continuously repairing, the laser head works continuously, and the continuous movement of the heat source in the same area will form a high temperature area. In the high temperature area, the heat input is much larger than the heat dissipation, which easily causes deterioration of the material organization and rapid increase of stress, and easily forms thermal cracks on the low melting point liquid film at the grain boundary. SUMMARY

[0004] To solve the technical problems in the background art, the present application provides a laser cladding repair method for the surface of the runner chamber of the hydropower station.

[0005] The laser cladding repair method for the surface of the runner chamber of the hydropower station provided by the present application comprises the following steps.

[0006] S1, three-dimensional point cloud data of the wear area of the runner chamber is obtained, and a three-dimensional wear area model is established based on the three-dimensional point cloud data and the wear area profile is identified.

[0007] S2. Based on the wear features of the wear region contour contained in the three-dimensional wear region model, decompose the wear region contour to obtain multiple standard circular arcs;

[0008] S3. Generate a fusion path based on multiple standard circular arcs;

[0009] S4. Control the robotic arm to move along the fusion path and keep the laser nozzle perpendicular to the wear area;

[0010] S5. Adjust the laser power based on the robotic arm's movement speed and real-time temperature to form the cladding layer;

[0011] S6. Verify the quality of the cladding layer and polish the surface of the cladding layer.

[0012] Preferably, S2 specifically comprises: S21, setting preset boundary markers at the boundary of the wear area in advance; S22, calculating the curvature of each point on the wear area contour, identifying the curvature extrema, and determining key dividing points in combination with the preset boundary markers; S23, dividing the wear area contour into multiple sub-intervals based on the key dividing points; S24, performing arc fitting on each sub-interval using the constrained least squares method to obtain standard arcs and outputting arc parameters representing each standard arc segment, wherein the arc parameters are specifically R. i , φ, Z0, k i α i and β i , where R i Let φ be the radius of curvature of the i-th standard arc, φ be the central angle parameter of the arc, Z0 be the axial reference height, and k be the radius of curvature of the arc. i α is the axial torsion coefficient. i ,β i The starting and ending angles of the arc.

[0013] Preferably, S2 further includes S25, which specifically involves: determining the maximum wear depth within the wear boundary sub-interval; determining the radius of the first arc and the radius of the second arc based on the maximum wear depth and the design radius, wherein the radius of the first arc is greater than the design radius and the radius of the second arc is less than the design radius; fitting the first arc and the second arc so that the first arc and the second arc are tangent and the tangent direction of the connection point is consistent with the trend of the wear area contour.

[0014] Preferably, S3 specifically includes: S31, constructing a parameterized equation for the wear arc, inputting the arc parameters of the standard arc into the parameterized equation for the wear arc, and obtaining a continuous standard arc strip composed of multiple segments of standard arcs connected end to end; S32, identifying the curvature difference between adjacent standard arcs; S33, generating a single smooth robot movement path without inflection points based on a single continuous standard arc strip using a curvature difference-driven Gaussian weight function and combined with Laplace smoothing constraints.

[0015] Preferably, the wear arc parameterization equation includes a first equation and a second equation, wherein the first equation is Γ i (φ)=[R i cosφ,R i sinφ, Z0-k i φ], the second equation is Γ i (φ)=[R i cosφ,R i sinφ, Z0-k i (n) φ], where φ∈[α] i ,β i ], k i (n) Dynamically updated based on the number of cladding layers; based on dynamically updated k i (n) After each cladding layer is completed, the laser displacement sensor scans the cladding layer thickness h. meas The system calculates the thickness deviation ratio. If |δ|>0.05, then k i (n) The algorithm updates according to the formula and automatically corrects the movement path of the robot on the next layer.

[0016] Preferably, S32 specifically involves: calculating the curvature difference between each adjacent standard arc; comparing each curvature difference with a preset curvature difference threshold; if all curvature differences are less than or equal to the first preset threshold, then S33 is not executed and the continuous standard arc strip is directly used as the robot's movement path; if any curvature difference is greater than the first preset threshold, then S33 is executed to process the continuous standard arc strip.

[0017] Preferably, step S33 specifically involves: calculating a curvature difference driving weight function based on the curvature difference, wherein the curvature difference driving weight function is... Where R i R i+1 The curvature radius of adjacent measurement points is denoted by σ, and the Gaussian kernel width is dynamically adjusted. The Laplacian smoothing constraint factor λ is dynamically determined based on the magnitude of the curvature difference. Using the curvature difference-driven weighting function and the Laplacian smoothing constraint factor, the fusion path point L between adjacent standard circular arcs is calculated using the path trajectory formula. f Generate a single smooth path without inflection points for each arc segment, wherein the path trajectory formula is: Where Γ i , Γ i+1 Let ω be the coordinate vector of the curvature center of adjacent measurement segments. i ω i+1 λ is the weighting factor, and λ is the Laplace smoothing constraint factor. It is a second-order differential operator for merging path points.

[0018] Preferably, the step of executing S33 to process the continuous standard arc strip when any curvature difference is greater than the first preset threshold specifically means: when any curvature difference is greater than or equal to the first preset threshold and less than the second preset threshold, executing S33 to automatically generate L. f ≤5 fusion path points; if any curvature difference is greater than or equal to the second preset threshold, execute S33 for automatic generation. One fusion path point.

[0019] Preferably, S4 specifically involves: real-time acquisition of surface state data of the wear area of ​​the rotary chamber, the surface state data including measurement data from a coaxial vision sensor and a laser displacement sensor; reconstructing a surface model of the wear area of ​​the rotary chamber based on the surface state data; calculating the normal vector of each point on the surface model; calculating the robot arm posture adjustment parameters based on the normal vector using a quaternion update algorithm; and controlling the robot arm to move along the fusion path and keep the laser nozzle perpendicularly pointing towards the wear area using a proportional-derivative controller according to the posture adjustment parameters.

[0020] Preferably, S5 specifically involves: acquiring real-time data on the robotic arm's movement speed and the temperature of the part to be repaired in the rotary chamber; and calculating the laser power adjustment amount based on the robotic arm's movement speed and the temperature of the part to be repaired in the rotary chamber using a dynamic power compensation formula, wherein the dynamic power compensation formula is... Where P0 is the reference power, ∈ is the heat dissipation coefficient, and T max T represents the highest temperature of the wear area during cladding, as fed back in real time by an infrared thermal imager. i The temperature of the wear area is fed back in real time by an infrared thermal imager; the laser output power is controlled according to the laser power adjustment amount to form a cladding layer.

[0021] This invention decomposes the wear area contour inside the turbine chamber of a hydropower station, constructing multiple interconnected standard arcs. This overcomes the limitations of traditional single arcs. By combining a weighting function in multi-curvature trajectory fusion to transform geometric features into mathematical weights, it overcomes the mechanical limitations of fixed-parameter fusion, achieving precise control of "strong fusion for large curvature differences and weak intervention for small curvature differences." This enables "terrain-adaptive" path planning, preventing sharp turns by the laser nozzle and ensuring the quality of the cladding layer. This invention can control the robotic arm to move along the fusion path while maintaining the laser nozzle perpendicular to the wear area, achieving precise control of the cladding area and improving repair accuracy. This invention can convert path smoothness into quantifiable thermodynamic control parameters, eliminating thermal stress through real-time temperature field feedback to prevent thermal cracking. Simultaneously, it utilizes the second derivative of curvature to suppress path jitter and combines a dynamic power compensation formula to reduce cladding thermal stress, avoiding the formation of high-temperature zones and thus helping to prevent thermal cracking. Attached Figure Description

[0022] Figure 1 This is a flowchart of the laser cladding repair method for the surface of the turbine chamber of a hydropower station proposed in this invention. Detailed Implementation

[0023] Reference Figure 1 The present invention proposes a laser cladding repair method for the surface of a hydropower station runner chamber, the specific steps of which are as follows:

[0024] S1. Obtain 3D point cloud data of the wear area inside the turbine chamber of the hydropower station, and use laser scanning to obtain a point cloud density of ≥50 points / cm². 2 It automatically identifies the contour of the wear area and builds a three-dimensional wear area model;

[0025] S2. Based on the wear features of the wear region contour contained in the three-dimensional wear region model, decompose the wear region contour to obtain multiple standard circular arcs;

[0026] Specifically, S21, set preset boundary markers in advance at the boundary of the wear area. The preset boundary markers are usually the starting point of the original contour (such as the design 0° position), the precise starting point of the wear area, the precise ending point of the wear area, and the ending point of the original contour (such as the 360° position).

[0027] S22. Calculate the curvature of each point on the contour of the wear area using mathematical methods, identify the extreme points of curvature, that is, the points where the curvature suddenly increases or decreases; then combine the preset boundary markers to determine the key dividing points. The key dividing points include the preset boundary markers and the extreme points of curvature. The preset boundary markers and the extreme points of curvature are combined together to form the key dividing points, and then they are arranged in order (angular order) on the contour.

[0028] S23. Between every two adjacent critical boundary points, an independent sub-interval is formed, dividing the wear area contour into multiple sub-intervals based on the critical boundary points. For example, the interval between the wear start point (A) and the first prominent curvature point (B) is [A->B]; the interval between point B and the next feature point (C) is [B->C]; and so on, until the wear end point.

[0029] S24. For each sub-interval, perform circular arc fitting using constrained least squares. Specifically, mathematically, continuously adjust the center position (xc, yc) and radius (r) of a virtual circle to minimize the sum of squared distances between points on this circle and all actual contour points within the sub-interval. During this calculation, some key constraints are applied, such as radius limitations. It must be ensured that the fitted radius r does not deviate excessively from the design radius R of the turbine chamber. des This constraint ensures that the fitting results meet basic engineering requirements (e.g., requiring |rR|). des |≤5%×Rdes Secondly, the endpoint of the fitted arc must be smoothly connected to the starting point of the next arc (i.e., at that point, the tangent directions of the two arcs cannot differ too much, for example, the angle difference should not exceed 2°). This constraint ensures that the entire repair trajectory is continuous and smooth, avoiding abrupt turns when the laser head moves.

[0030] It is known that the profile may change drastically at the boundary between the intact area and the worn area (such as near the wear start point and wear end point). Therefore, in order to prevent the robot's movement path from changing drastically at this location, it is necessary to process it using the following step S25.

[0031] S25. Determine the maximum wear depth within the wear boundary sub-interval; based on the maximum wear depth and the design radius, determine the radius of the first arc and the radius of the second arc, wherein the radius of the first arc is greater than the design radius and the radius of the second arc is less than the design radius; fit the first arc and the second arc so that the first arc and the second arc are tangent and the tangent direction of the connection point is consistent with the trend of the wear area contour.

[0032] The design rules for the first and second circular arcs are as follows:

[0033] Design rules:

[0034] (1) The radius of the arc segment near the intact area will be slightly larger than the design radius R. des (For example, r1 = R) des +0.7×Δh, where Δh is the maximum wear depth at this location). The radius of the arc segment near the wear zone will be slightly smaller than the design radius R. des (For example, r2 = R) des -0.3×Δh).

[0035] (2) The first and second arcs must be tangent, and the direction of the tangent at the connection point should be as consistent as possible with the overall contour trend.

[0036] After the above steps S21-S25, a standard arc is finally obtained, and the arc parameters representing each standard arc segment are output. Specifically, the arc parameters are R... i , φ, Z0, k i α i and β i , where R i Let be the radius of curvature of the i-th arc segment, φ be the central angle parameter of the arc, Z0 be the axial reference height, and k be the radius of curvature of the arc segment. i α is the axial torsion coefficient. i ,β i The starting and ending angles of the arc.

[0037] This step can decompose complex wear areas into multiple standard arcs based on the wear characteristics of the inner surface of the turbine runner, breaking through the traditional assumption of a single arc.

[0038] S3. Generate a fusion path based on multiple standard circular arcs;

[0039] Specifically:

[0040] S31. Construct the parameterized equation of the wear arc, input the arc parameters of the standard arc into the parameterized equation of the wear arc, and obtain a continuous standard arc strip composed of multiple standard arcs connected end to end.

[0041] The wear arc parameterization equation includes a first equation and a second equation, wherein the first equation is Γ. i (φ)=[R i cosφ,R i sinφ, Z0-k i φ], the first equation is mainly applied to cylindrical curved surfaces;

[0042] The second equation is Γ i (φ)=[R i cosφ,R i sinφ, Z0-k i (n) φ], the second equation is particularly applicable to conical surfaces;

[0043] Where: φ∈[α] i ,β i ];R i The radius of curvature of the i-th arc segment is represented by φ, the central angle parameter of the arc is Z0, and the axial reference height is k. i α is the axial torsion coefficient. i ,β i R is the start and end angle of the arc; i cosφ,R i sinφ describes the circular motion of an arc; Z0-k i φ solves the problem of axial height variation (especially for conical impeller chambers).

[0044] k i (n) Dynamically updated based on the number of cladding layers:

[0045]

[0046] h meas : Measured cladding layer thickness using a laser displacement sensor (unit: mm);

[0047] h theory : Target thickness set in the process (unit: mm);

[0048] Specifically, the robot moves along a fixed k i During the motion, the nth cladding layer will generate Δz. n =0.02nm cumulative misalignment, using the above-mentioned dynamic compensation strategy, after each cladding layer is completed, the laser displacement sensor scans the thickness h. meas The system calculates the thickness deviation ratio. If |δ|>0.05, then k i (n) The algorithm updates according to the formula and automatically corrects the movement path of the robot on the next layer.

[0049] The above-mentioned correction settings help prevent the cladding layer from accumulating on the conical surface and ensure that the cladding layer thickness is uniform.

[0050] S32. Identify the curvature difference between adjacent standard arcs: Calculate the curvature difference between each adjacent standard arc; compare each curvature difference with a preset curvature difference threshold;

[0051] When all curvature differences are less than or equal to the first preset threshold (0.3m≤ΔR<1.0m), S33 is not executed and the continuous standard arc strip is directly used as the robot's movement path, because when there is no sudden change in curvature, the robot can move directly along a single arc path (such as uniform wear of a φ500mm roller surface).

[0052] If any curvature difference is greater than the first preset threshold (ΔR≥1.0m), then process the continuous standard arc strip using S33; if any curvature difference is greater than or equal to the first preset threshold and less than the second preset threshold, then automatically generate L using S33. f ≤5 fusion path points; if any curvature difference is greater than or equal to the second preset threshold, execute S33 for automatic generation. One fusion path point.

[0053] S33. Based on a single continuous standard circular arc strip, a Gaussian weight function driven by curvature difference is used, combined with Laplace smoothing constraint to generate a single smooth robot movement path without inflection points;

[0054] Specifically:

[0055] A curvature difference-driven weighting function is calculated based on the curvature difference, and a Laplace smoothing constraint factor λ is dynamically determined based on the magnitude of the curvature difference. Using the curvature difference-driven weighting function and the Laplace smoothing constraint factor, the fusion path point L between adjacent standard circular arcs is calculated using the path trajectory formula. f Generate a single smooth path without inflection points for each arc segment;

[0056] The formula for the path trajectory is as follows:

[0057]

[0058] Wherein: Γ i , Γ i+1 The coordinate vectors of the curvature centers of adjacent measurement segments;

[0059] ω i ω i+1 For weighting factors;

[0060] λ is the Laplace smoothing constraint factor;

[0061] For the second-order differential operator of the fusion path points;

[0062] The curvature difference-driven weighting function is as follows:

[0063]

[0064] in:

[0065] R i R i+1 The radius of curvature (in meters) of adjacent measurement points;

[0066] σ is the dynamic adjustment parameter for the Gaussian kernel width, with a base value of 0.3m.

[0067] To ensure a smooth transition of the laser cladding arm from one arc segment to another, a Laplacian smoothing constraint mechanism is employed, in which:

[0068]

[0069] λ is dynamically configured based on the intensity of curvature abrupt changes, and the specific configuration rules are shown in Table 1 below:

[0070] Table 1

[0071]

[0072] When the curvature is too large, the weight will decay exponentially. According to the formula, it can be seen that it will automatically amplify the influence of the later stage. The weight factor will become smaller and the vector coordinates will be closer. It is equivalent to building a transition ramp between the sharp bend and the gentle slope.

[0073] When ΔR < 0.3m, ω i If the radius is approximately 1, then the trajectory is directly "stitched together" to maintain the same radius. The weak intervention operation of "stitching together the trajectory" can be summarized as "native geometric connection based on curvature difference threshold + micro-smoothing", and its complete operation chain is as follows: when the curvature difference between adjacent arc segments decomposed in step two is |R i -R j When | < 0.3m, the system invokes the splicing mode (weak intervention mechanism). At this time, the path generation module will assign the coordinates of the endpoint of the first arc segment (x... i ,y i) and the coordinates of the starting point of the second arc (x j ,y j The points that directly coincide with each other are connection points C0 (which must satisfy |C0-O) i |=R i And |C0-O j |=R j Geometric constraints are applied; if the original data does not satisfy them, O is corrected using the least squares method. j (Coordinates), and then the Laplacian smoothing factor λ is set to 0.25 to perform a single gradient smoothing operation on the 5 path points surrounding the connection point: This smoothing only eliminates sensor noise (amplitude N = N). i +N j (No new points added) At the same time, a weak intervention identifier is sent to the process control module, triggering the following linkage response: the laser power is maintained at the reference value P0, the axial flow velocity is maintained at 120mm / s, the robot arm posture update rate is reduced to 60Hz, and the final output is the processing trajectory directly spliced ​​by two original arcs (without adding any transition curves).

[0074] When ΔR > 1.0m, ω i If the value approaches 0, a powerful fusion mode is activated, generating a transition function L. trans =k·∣R i -R j | (k = 30 points / m), automatically extend the transition zone length (up to 30% of the original segment) to avoid "sharp turns and rollovers". At this time, the path generation unit performs the following chain of operations:

[0075] (1) Dynamic weight calculation: The center O of the first arc is calculated using the hyperbolic secant function w = sech(2.7Δκ). i The weights (at this point w→0) are adjusted, and the Laplace constraint factor λ is simultaneously increased to 0.92;

[0076] (2) Transition trajectory synthesis: through vector operations Generate 17 to 23 dense transition path points and automatically extend the transition segment length to 32% of the original arc length (extension coefficient μ = 0.32);

[0077] (3) Real-time fault-tolerant correction: Activate the closed-loop monitoring module when the trajectory curvature change rate |dκ / ds|>0.8m -1 Automatically insert curvature buffer point at / mm. ( (The normal vector) ultimately resulted in only 0.4 hot cracks / meter in the rim repair with a curvature difference > 1.8m (compared to > 5 cracks / meter in traditional methods).

[0078] The following example demonstrates the process of generating the blending path, using standard arc A and standard arc B as examples:

[0079] The relevant parameters of standard circular arc A and standard circular arc B are shown in Table 2 below:

[0080] Table 2

[0081]

[0082] The large curvature difference ΔR = 2.69m was calculated;

[0083] System response:

[0084] If ΔR > 1.0m, activate the strong fusion mode: calculate σ and ω. i And λ.

[0085] The formula for calculating σ is: σ = γ base +k reg ·ΔR,

[0086] Where, γ base =0.25 is the baseline adjustment parameter; k reg =0.1 is the curvature difference influence coefficient (calibrated through process testing).

[0087] Substituting ΔR = 2.69m into the formula for calculating σ, we can obtain:

[0088] σ = 0.25 + 0.1 × 2.69 = 0.519

[0089]

[0090] λ = 0.93

[0091] The obtained σ, ω i Substituting λ into the path trajectory formula, 82 fusion path points are generated, resulting in a transition zone length of 1.72m.

[0092] It is worth noting that the value of λ is initially determined based on the configuration rules in Table 1 above, where ΔR>1.0m, and is initially set to ≥0.9; however, excessively high λ will weaken the geometric features and needs to be considered in conjunction with the weight ω. i =0.96 Synergy: Under strong smoothing constraints, 96% of the original curvature characteristics are still retained to avoid cladding layer collapse. Finally, the intermediate value is taken to determine λ=0.93.

[0093] S4. Control the robotic arm to move along the fusion path and keep the laser nozzle perpendicular to the wear area;

[0094] The specific process is as follows: S41, real-time acquisition of surface condition data of the wear area of ​​the impeller chamber, the surface condition data including measurement data from coaxial vision sensor and laser displacement sensor;

[0095] S42. Reconstruct the surface model of the wear area of ​​the turbine chamber based on surface condition data;

[0096] The NURBS surface is fitted using the Progressive Iterative Approximation (PIA) algorithm:

[0097]

[0098] Here, u and v refer to surface parameters, which are two independent variables, typically in the range [0, 1].

[0099] The range of values ​​can be varied. By changing the values ​​of u and v, all points on the entire surface can be traversed.

[0100] m and n refer to the upper limit of the control point index: the index of i ranges from 0 to m, so there are m+1 control points in the u-parameter direction; the index of j ranges from 0 to n, so there are n+1 control points in the v-parameter direction. Therefore, the entire control point grid is an (m+1)×(n+1) matrix.

[0101] N (i，p) (u) is the p-th B-spline basis function. It is a function of parameter u, defining the influence of the i-th control point on the surface at parameter u. Its value is determined by the node vectors and the degree p using a recursive formula (Cox-de-Boor algorithm). It forms the "skeleton" of the entire surface, providing local support and ensuring smoothness.

[0102] N (j，q) (v) is the q-th B-spline basis function. It defines the influence of the j-th control point in the v direction.

[0103] P (i，j) These are control points. Each point is a three-dimensional coordinate point (x, y, z) that forms a control grid. These points act like "magnets," attracting the surface and determining its approximate shape and extent. P(i, j) is the control point in the i-th row and j-th column.

[0104] w (i，j) It is the weight, relative to the control point P. (i，j) The associated positive real number. It represents the "strength" of the attraction of the control point to the surface; the larger the weight, the closer the surface is pulled to the control point. When all weights are 1, the NURBS surface degenerates into a regular nonrational B-spline surface.

[0105] (1) Dynamic data acquisition, as shown in Table 3 below:

[0106] Table 3

[0107]

[0108] (2) Thermal deformation fusion calculation

[0109] Establish a geometry-temperature coupling model:

[0110]

[0111] in:

[0112] α = 1.2 × 10 -5 / ℃ (coefficient of thermal expansion)

[0113] β=0.07mm 2 (Material creep factor)

[0114] K: Workpiece stiffness matrix.

[0115] S43. Calculate the normal vector of each point on the surface model;

[0116] Formula for calculating normal vector:

[0117] in, : refers to the tangent vector of the surface along the u direction; this is a partial derivative that measures the change in surface point S when the parameter v is fixed and only u changes slightly. (u，v) The direction and speed of motion;

[0118] : The tangent vector of the surface along the v direction; it is also a partial derivative, which measures the value of a point S on the surface when the parameter u is fixed and only v changes slightly. (u，v) The direction and speed of motion.

[0119] S44. Calculate the robot arm posture adjustment parameters based on the normal vector using a quaternion update algorithm;

[0120] Quaternion update formula:

[0121] Where, q k : Represents the attitude (i.e., orientation) of the robotic arm's laser nozzle at the current moment (the kth control cycle);

[0122] q k+1 : Represents the new posture that the robotic arm's laser nozzle should adjust to in the next moment (the (k+1)th control cycle);

[0123] Δθ represents the "dynamic molten pool tilt angle compensation value," which is essentially a real-time correction angle between the laser beam incident direction (theoretical normal) and the actual flow direction of the molten pool. Its function is shown in Table 4 below:

[0124] Table 4

[0125]

[0126] The specific process is as follows:

[0127] Input: Deviation between current state and target;

[0128] Current stance: q k (Known);

[0129] Required rotation angle: Δθ (dynamic molten pool tilt angle compensation value, calculated from the table above);

[0130] The axis around which rotation is required: (That is, the unit normal vector perpendicular to the surface calculated in S43);

[0131] Calculate incremental rotation: around the axis Rotate by an angle Δθ, precisely to correct the nozzle from its current direction to a direction that is perfectly aligned with the surface normal.

[0132] Execution Update: Quaternion Multiplication symbol This represents quaternion multiplication, which is not ordinary multiplication but is used to combine rotations.

[0133] Perform a quaternion multiplication on the current pose qk and the incremental rotation to obtain the result q. k+1 It incorporates the new posture after this rotation.

[0134] S45. Based on the attitude adjustment parameters, the robotic arm is controlled by a proportional-derivative controller to move along the fusion path and keep the laser nozzle perpendicularly pointing towards the wear area. Dynamic response optimization is performed during this process.

[0135]

[0136] Where J: Jacobian matrix of the robotic arm;

[0137] K p K d PD gain (0.6 / 0.15);

[0138] e: Normal vector deviation angle.

[0139] The main process of S4 involves real-time surface reconstruction. The key technology for real-time surface reconstruction is point cloud feature extraction, which is mainly achieved using a 1200Hz line laser array with a point density of ≥50 points / cm². 2 Then, normal vector calculation is performed. The key technology is to use differential geometry and GPU for parallel computation. The response index is <3ms latency. Finally, quaternion update is performed, attitude interpolation is applied, and the laser head is controlled according to the response index of 120Hz update rate through SO(3) manifold optimization.

[0140] Compared to existing Euler angle control technologies, this invention avoids gimbal lock-up and reduces computational load by 40% compared to rotation matrix control. Calculations show that the attitude error of this invention is 0.02 rad in a circular arc segment with R = 1.8 m, while the attitude error of the traditional Euler angle method in the same segment is 0.12 rad. This demonstrates that the invention provides significantly smaller errors in laser head attitude control.

[0141] S5. Adjust the laser power based on the robotic arm's movement speed and real-time temperature to form the cladding layer;

[0142] Specifically: S51, real-time acquisition of robotic arm movement speed and temperature data of the part to be repaired in the rotary chamber;

[0143] (1) Data Acquisition

[0144] Coaxial infrared pyrometer: sampling rate 1kHz, spatial resolution 0.2mm;

[0145] Real-time capture of the bright temperature field T(x,y) on the surface of the molten pool;

[0146] (2) Peak Extraction

[0147]

[0148] Ω pool Dynamic region of the molten pool (short axis 1.5–4 mm);

[0149] ε: Material emissivity.

[0150] T max The coupling rules with other temperatures are shown in Table 5 below:

[0151] Table 5

[0152]

[0153] Among them, T reheat : Represents the reheat temperature; in multilayer cladding, when the laser head begins to clad a new layer, its heat will reheat a specific area of ​​the already solidified previous layer (or base material) to the same temperature; mainly used to control "melt pool stability". If T max With T reheat An excessively high ratio indicates that the secondary heating of the underlying material is too intense, potentially leading to over-tempering, coarse grains, or even localized remelting of the underlying structure, thus damaging its properties. By monitoring this ratio, the system can adjust the power to ensure a good metallurgical bond between the new cladding layer and the underlying layer without damaging the repaired area.

[0154] T max_HAZThe heat-affected zone (HAZ) represents the highest temperature within the heat-affected zone. This zone is located next to the molten pool, in the unmelted area of ​​the base material where mechanical properties and metallographic structure change due to heat. It is primarily used to control the "metallurgical bond quality." Molten pool temperature (T) max The temperature must be significantly higher than the HAZ to ensure complete melting of the powder and a strong metallurgical bond with the base material. If the temperature difference (T) max- T max_HAZ If the temperature difference is too small, it means that the heat in the molten pool is insufficient, which may result in "adhesion" rather than "fusion" and poor bonding strength. The system ensures the quality of repair by maintaining a sufficiently large positive temperature difference.

[0155] T base : Represents the base temperature; refers to the average temperature of the workpiece body (base material of the rotating chamber) far away from the molten pool and heat-affected zone. It can be regarded as the "ambient" temperature or background temperature of the repair area.

[0156] h: Represents a characteristic distance, i.e., from the center of the molten pool (temperature approximately T). max ) to the matrix (temperature approximately T) base The equivalent thermal conduction distance between ( ) is a thermodynamic control parameter. It is used to calculate the temperature gradient ΔT. A large temperature gradient is a major cause of thermal stress and hot cracking. The system indirectly controls this temperature gradient by controlling Tmax through a dynamic power compensation formula, keeping it within the material's tolerance range to suppress residual stress and hot cracking.

[0157] S52. Based on the robotic arm's movement speed and the temperature data of the part to be repaired in the rotary chamber, the laser power adjustment is calculated using a dynamic power compensation formula, which is as follows:

[0158]

[0159] Where: P0 is the reference power, ∈ is the heat dissipation coefficient, and T max T represents the highest temperature of the wear area during cladding, as fed back in real time by an infrared thermal imager. i ν represents the temperature of the wear area as fed back in real time by the infrared thermal imager; v represents the instantaneous linear velocity of the robotic arm end effector (laser nozzle) relative to the wear surface of the rotary chamber.

[0160] ξ is a dimensionless process coefficient that describes the coupling relationship between the robotic arm's motion speed v and the local surface curvature 1 / Ri on laser power. This coefficient encapsulates complex heat conduction physics. When the robotic arm moves in a high-curvature (small radius Ri) region, such as a small bend or the edge of a pit, even if the movement speed v remains constant, the trajectory of the laser heat source becomes more "compact," potentially leading to localized heat accumulation and temperature increases. ξ's role is to reconcile the conflict between speed and curvature. It determines the sensitivity of power adjustment required to cope with a specific v / Ri ratio.

[0161] ξ is not a fixed physical constant, but an empirical parameter that needs to be calibrated through process experiments. Its optimal value depends on:

[0162] 1. Material properties: Thermophysical properties of cladding powder and rotor chamber base material (such as thermal conductivity and specific heat capacity).

[0163] 2. Laser spot characteristics: size and energy distribution of the laser spot.

[0164] 3. Cladding process window: desired cladding depth, cladding width, etc.

[0165] S53. The laser output power is controlled according to the laser power adjustment amount to form a cladding layer. After the above control, when R1 = 1.2m and R2 = 4.5m, the laser head can adaptively reduce its speed by 80mm / s in the R1-R2 transition zone, and the attitude adjustment frequency is 120Hz, which is better than the 60Hz of the prior art (publication number CN113981434B). It can also reduce the power in the heat accumulation zone (T>800℃) to 70%.

[0166] This invention utilizes a dual-factor constraint mechanism of geometric and thermodynamic constraints to establish a cross-domain control chain of λ∝ΔR∝P(v), achieving dual control through λ, thus realizing dual control at both the mechanical and thermal ends. Mechanical end: limiting joint angular acceleration α ≤ 0.3 rad / s². 2 This transforms path smoothness into quantifiable thermodynamic control parameters. On the thermal side: it triggers closed-loop laser power adjustment, eliminates thermal stress through real-time temperature field feedback, and utilizes the second derivative of curvature. By suppressing path jitter and combining it with the subsequent dynamic power compensation formula to reduce cladding thermal stress, the depth of the heat-affected zone (HAZ) of the cladding layer can be reduced from 0.42 mm to 0.17 mm.

[0167] The following describes the steps for adjusting the laser power along the path, using the fusion path generated by standard arcs A and B in step S3 as an example:

[0168] First, the path is geometrically determined that the robotic arm must reduce its speed to v = 0.08 m / s when passing through the transition zone from standard circular arc A to standard circular arc B to prevent sharp turns and material accumulation.

[0169] Next, by using sensors such as infrared thermal imagers to monitor in real time, the temperature data of the current wear area was obtained, for example: T max =2100℃, T i =1850℃;

[0170] Then, the system calls a complete set of parameters that have been pre-calibrated through numerous process experiments, for example: P0 = 1500W; ξ = 0.5; ∈ = 0.001. Simultaneously, it is assumed that the current laser head is located at a position with a radius of curvature of Ri ≈ 2.0m (for example, this is an equivalent radius of curvature on the fusion path);

[0171] Finally, by using the known R i , λ, v, T max T i Substituting P0, ξ, and ∈ into the dynamic power compensation formula above, we finally calculated P = 1426W. The calculation process is shown below:

[0172]

[0173] P = 1500 × exp(0.08) × [1 + 0.25]

[0174] P≈1500×1.083×1.25

[0175] P≈1426W

[0176] S6. Verify the quality of the cladding layer and polish the surface of the cladding layer to obtain a more ideal cladding layer.

[0177] The present invention was compared with other methods in terms of the variation in the thickness of the prepared cladding layer, the number of hot cracks, and the attitude delay time during robot movement. The comparison results are shown in Table 6 below:

[0178] Table 6

[0179]

[0180] As can be seen from the table above, compared with traditional ring scanning technology and NURBS planning technology, the cladding layer prepared by this invention has a more uniform thickness, produces fewer thermal cracks, and has a shorter delay during robot movement.

[0181] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for repairing the surface of a hydroelectric power station runner chamber by laser cladding, characterized in that, The method comprises the following steps: S1, obtaining three-dimensional point cloud data of the wear area inside the runner chamber of a hydropower station, identifying the wear area profile based on the three-dimensional point cloud data, and establishing a three-dimensional wear area model; S2, decomposing the wear area profile based on the wear features of the wear area profile contained in the three-dimensional wear area model to obtain a plurality of standard circular arcs; S3, generating a fusion path based on the plurality of standard circular arcs; S4, controlling the mechanical arm to move along the fusion path and keeping the laser nozzle perpendicular to the wear area; S5, adjusting the laser power based on the movement speed of the mechanical arm and the real-time temperature to form a cladding layer; S6, verifying the quality of the cladding layer and polishing the surface of the cladding layer.

2. The method of claim 1, wherein the method further comprises: S2 is specifically: S21, setting preset boundary marker points in advance at the wear region boundary; S22, calculating the curvature of each point on the wear region profile, identifying the curvature extreme points, and determining the key demarcation points in combination with the preset boundary marker points; S23, dividing the wear region profile into multiple subintervals according to the key demarcation points; S24, performing circular arc fitting on each subinterval by using a least square method with constraints to obtain standard circular arcs and output circular arc parameters representing each standard circular arc, the circular arc parameters being specifically R i , φ, Z0, k i , α i , and β i , wherein R i is the curvature radius of the ith standard circular arc, φ is a circular arc center angle parameter, Z0 is an axial reference height, k i is an axial distortion coefficient, α i , β i are circular arc start and end angles.

3. The method of claim 2, wherein the laser cladding is performed on the surface of the runner chamber of the hydroelectric power station. S2 further comprises S25, which is specifically: determining the maximum wear depth in the wear boundary subinterval; based on the maximum wear depth and the design radius, determining the radius of the first circular arc and the radius of the second circular arc, wherein the radius of the first circular arc is greater than the design radius, and the radius of the second circular arc is less than the design radius; fitting the first circular arc and the second circular arc so that the first circular arc and the second circular arc are tangent and the tangent direction of the connection point is consistent with the trend of the wear area profile.

4. The method of claim 2, wherein the method further comprises: S3 is specifically: S31, constructing a wear circular arc parameterization equation, inputting the circular arc parameters of the standard circular arc into the wear circular arc parameterization equation to obtain a continuous standard circular arc strip composed of a plurality of standard circular arcs connected end to end; S32, identifying the curvature difference between adjacent standard circular arcs; S33, based on a single continuous standard circular arc strip, using the curvature difference to drive a Gaussian weight function and jointly generating a robot movement path with no inflection points using Laplace smoothing constraints.

5. The method of claim 4, wherein the method further comprises: The wear circle arc parameterization equation includes a first equation and a second equation, the first equation is Γ i (φ) = R i cosφ, R i sinφ, Z0-k i φ], and the second equation is Γ i (φ) = [R i cosφ, R i sinφ, Z0-k i (n) φ], wherein φ∈[α i ,β i ], k i (n) Dynamically update according to the number of cladding layers; based on the dynamically updated k i (n) After completing one layer of cladding, the laser displacement sensor scans the cladding layer thickness h meas , the system calculates the thickness deviation ratio If |δ|>0.05, then k i (n) Update according to the formula, automatically correct the next layer of robot movement path.

6. The method of claim 4, wherein the method further comprises: S32 is specifically: calculating the curvature difference between each adjacent standard circular arc; comparing each curvature difference with a preset curvature difference threshold; when all curvature differences are less than or equal to a first preset threshold, S33 is not executed and the continuous standard circular arc strip is directly taken as the robot movement path; when any curvature difference is greater than the first preset threshold, S33 is executed to process the continuous standard circular arc strip.

7. The method of claim 4, wherein the laser cladding is performed on the surface of the runner chamber of the hydroelectric power station. The S33 is specifically: calculating a curvature difference driving weight function based on the curvature difference, the curvature difference driving weight function is wherein R i , R i+1 is the curvature radius of adjacent measurement points, and σ is a Gaussian kernel width dynamic adjustment parameter; a Laplace smoothing constraint factor λ is dynamically determined based on the size of the curvature difference; The fusion path point L between adjacent standard circular arcs is calculated by a path trajectory formula using a curvature difference driving weight function and a Laplace smoothing constraint factor f , to generate a single smooth path without inflection points for each arc segment, wherein the path trajectory formula is wherein Γ i , Γ i+1 is the curvature center coordinate vector of adjacent measurement segments, ω i , ω i+1 is a weight factor, λ is a Laplace smoothing constraint factor, is a second-order differential operator of the fusion path point.

8. The method of claim 7, wherein the method further comprises: The step S33 of processing the continuous standard circular arc strip when any curvature difference is greater than the first preset threshold value specifically comprises: when any curvature difference is greater than or equal to the first preset threshold value and less than the second preset threshold value, automatically generating L f ≤5 fusion path points; and when any curvature difference is greater than or equal to the second preset threshold value, automatically generating fusion path points.

9. The method of claim 1, wherein the method further comprises: S4 is specifically: real-time acquisition of the surface state data of the wear area of the runner chamber, the surface state data including the measurement data of the coaxial vision sensor and the laser displacement sensor; reconstructing the curved surface model of the wear area of the runner chamber based on the surface state data; calculating the normal vector of each point on the curved surface model; calculating the mechanical arm posture adjustment parameter based on the normal vector through a quaternion update algorithm; controlling the mechanical arm to move along the fusion path and keep the laser nozzle pointing vertically to the wear area through a proportional-derivative controller according to the posture adjustment parameter.

10. The method of claim 1, wherein, S5 specifically is: real-time acquisition of the mechanical arm movement speed and the runner chamber part to be repaired temperature data; based on the mechanical arm movement speed and the runner chamber part to be repaired temperature data, the laser power adjustment amount is calculated through a dynamic power compensation formula, the dynamic power compensation formula is Wherein P0 is the reference power, ∈ is the heat dissipation coefficient, T max is the highest temperature of the wear area during cladding, which is fed back by the infrared thermal imager in real time, T i is the wear area temperature fed back by the infrared thermal imager in real time. The laser output power is controlled according to the laser power adjustment amount to form a cladding layer.

Images