Terahertz-based curved surface thermal barrier coating nondestructive thickness measuring system and method

Abstract

The invention discloses a terahertz-based curved surface thermal barrier coating nondestructive thickness measurement system and method, and the method comprises the steps: firstly carrying out the coordinate calibration and reverse reconstruction of a sample piece and a mechanical arm through an optical scanning and trinocular vision measurement system, and building a virtual space; then matching the virtual space with the actual space, extracting a normal vector of the curved surface of the sample piece, performing path planning on the mechanical arms to generate a control script, and controlling the double mechanical arms to carry a terahertz measuring head to scan the sample piece in a manner of being vertical to the surface of the sample piece to obtain a terahertz signal; and finally, constructing a terahertz signal analytical model, optimizing terahertz simulation signal parameters through a self-adaptive teaching and learning optimization algorithm, and fitting a detection signal until an optimal solution is obtained, namely the thickness of the thermal barrier coating. According to the invention, through cooperation of the two mechanical arms and path planning, rapid detection of a complex curved surface structure is adapted; by establishing an analytical model and combining with an optimization algorithm, the detection precision is improved, and high-precision nondestructive detection suitable for a complex curved surface structure is realized.

Description

Technical Field

[0001] The present invention relates to the fields of terahertz thickness measurement technology and curved surface detection technology, and in particular to a terahertz-based non-destructive thickness measurement system and method for curved surface thermal barrier coatings. Background Art

[0002] The engine is regarded as the heart of modern aerospace aircraft, and the thermal barrier coating is regarded as the cover of the aircraft engine blades. The thermal barrier coating is an important thermal insulation component of current aircraft engines, which provides a guarantee for the safe and stable operation of turbine blades in high temperature environments. The thickness of the ceramic layer is an important factor in determining its service performance, and its thickness has strict indicators.

[0003] In recent years, non-destructive testing methods have developed rapidly and have been widely used in industrial production. For example, eddy current, ultrasound, infrared thermal imaging, microwaves and radiation can all achieve non-destructive testing. However, these traditional testing methods often have limitations such as low accuracy, difficulty in qualitative and quantitative evaluation, corrosion of coatings, and harm to the human body. As a new type of non-destructive testing technology, terahertz has the characteristics of coherence, strong penetration, low energy, transient, characteristic fingerprint, etc., coupled with the breakthrough of terahertz time-domain spectroscopy technology, terahertz technology has become a new star in the field of non-destructive testing and analysis.

[0004] The above advantages have made terahertz nondestructive testing technology increasingly widely used in ceramic layer thickness detection. The main measurement methods are based on time of flight, based on the standard TLBO optimization algorithm, and based on machine learning. Based on the time of flight terahertz thickness measurement method, it is necessary to prepare a standard specimen to calibrate the refractive index, and extract the flight time by locating the reflection peak to calculate the thickness of the ceramic layer. Due to the large number of pores randomly distributed inside the ceramic layer and the need to correct the influence of surface roughness on the reflection signal, the refractive index changes with the spatial position, which is different from the calibration result, reducing the measurement accuracy. Based on the standard TLBO optimization algorithm, there is a problem of poor individual diversity in the population during the initialization process, the optimal solution search process is slow and easy to fall into the local optimum, resulting in low solution efficiency, poor accuracy and stability. Based on the machine learning method, a large number of data samples are required to train weights and biases, and to establish the input features and thickness intervals, but the construction of data requires the use of destructive detection methods to obtain accurate thickness values, and the training cost is too high.

[0005] When using a terahertz probe to scan a sample and obtain a terahertz detection signal, a common scanning method is to directly scan the sample surface through a two-dimensional guide rail. However, the two-dimensional scanning guide rail can only move in the X and Y directions. Conventional terahertz time-domain spectroscopy systems can only perform non-destructive testing on flat samples and cannot adapt to curved surface testing. Summary of the invention

[0006] The problem to be solved by the present invention is to provide a terahertz-based non-destructive thickness measurement system and method for curved thermal barrier coatings, which is used to solve the problems of low thickness measurement accuracy, low efficiency and susceptibility to complex curved surfaces in turbine blade thermal barrier coatings in the prior art.

[0007] The present invention adopts the following technical solution: a terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings, comprising: a trinocular vision measurement system, a sample, a robot, and a base;

[0008] The trinocular vision measurement system is used to calibrate the robot's motion parameters and tool coordinate system. The robot includes a six-axis robotic arm, a terahertz probe, and a tooling. A flange is also provided at the end of the robot, and the flange is connected to the base through positioning bolts.

[0009] One end of the six-axis robotic arm is connected to the flange, and the other end is connected to the terahertz probe through a tooling. The robot controls the terahertz probe to scan the sample through the six-axis robotic arm, and the sample is fixed on the base through a fixture.

[0010] The terahertz probe emits terahertz signals perpendicular to the surface of the sample, receives terahertz signals reflected from the surface of the sample, and transmits them to the host computer for terahertz signal processing.

[0011] Preferably, the tooling length is 112 m, which is consistent with the focal length f of the lens of the terahertz probe, so that the terahertz signal reflection point on the surface of the sample is the point with maximum energy.

[0012] The technical solution of the present invention also includes: a terahertz-based non-destructive thickness measurement method for curved thermal barrier coatings, which is applied to any of the above systems to perform non-destructive thickness measurement of curved thermal barrier coatings, comprising the following steps:

[0013] S1. Obtain the point cloud information of the sample to establish a model, perform feature point matching, and obtain the matrix transformation relationship between the workpiece coordinate system and the robot base coordinate system;

[0014] S2. Calibrate the robot base coordinate system and tool coordinate system based on the lens forward detection method;

[0015] S3, establish a virtual space, extract the surface normal vector of the sample, and plan the path of the robot;

[0016] S4, the robot controls the terahertz probe to scan the sample according to the planned path through the mechanical arm to obtain the terahertz time domain spectrum signal of the sample;

[0017] S5. Establish a terahertz signal analysis model;

[0018] S6. Through the terahertz signal analysis model, the terahertz time-domain spectral signal of the sample is optimized by the adaptive teaching and learning optimization algorithm to obtain the thickness measurement value of the thermal barrier coating on the curved surface of the sample.

[0019] Preferably, in step S1, a point cloud model of the sample is established in the following manner:

[0020] S1.1. Arrange multiple reference points on the surface of the sample as the basis for point cloud stitching during scanning, and fix the sample on the base;

[0021] S1.2. Scan the sample using the ATOS optical scanner;

[0022] Under the ATOS optical scanner, the coordinates of 6 feature points are selected as P i (x i ,y i ,z i )(i=1~6), in the real space, the corresponding feature point coordinates are measured by the probe and are P j (x j ,y j , z j )(=1~6);

[0023] S1.3, feature point matching based on least squares method;

[0024] S1.4. Export the STL file obtained after processing by the ATOS optical scanner, perform reverse reconstruction in the Catia software, match the digital model obtained by reverse reconstruction to the actual space, obtain the matrix transformation relationship between the workpiece coordinate system and the robot base coordinate system, and obtain the CAD model of the sample.

[0025] Preferably, in step S2, the robot base coordinate system and the tool coordinate system are calibrated in the following manner:

[0026] S2.1. Calibrate the robot base coordinate system through the trinocular vision measurement system, place the robot at zero position, and the joint angles of the six-axis robot arm are 0°, -90°, 0°, -90°, 0°, 0° respectively;

[0027] S2.2. Rotate the first axis of the six-axis robot arm, measure the coordinates of the fixed point of the robot end flange at a certain angle, fit the circle C1 according to the obtained coordinates, and determine the center of C1. The straight line l1 passing through the center of the circle and perpendicular to the plane where the circle C1 is located is used as the Z axis of the robot base coordinates;

[0028] S2.3. Reset the robot to zero position, keep the angles of other axes unchanged, rotate the second axis of the six-axis robot arm, measure the coordinate values ​​of the flange fixed point at the end of the robot at certain angles, and fit the plane M according to the obtained coordinate values;

[0029] S2.4. Calibrate the robot base coordinate system, measure the plane B where the base is located, take the intersection of the Z axis and the plane where the base is located as the origin of the base coordinate system, make the normal of plane M through the origin of the coordinate system as the Y axis, and determine the X axis according to the coordinate system rule;

[0030] S2.5. Calibrate the joint rotation axis: Each time the robot is placed in zero position, each axis of the six-axis robot arm is rotated separately, and the coordinate value of the end flange fixed point is measured each time it is rotated a certain angle, and a circle C is fitted in the SA software. k (k = 1 to 6), and draw perpendicular lines through the center of each circle to the plane where each circle is located, and obtain the straight line l k (k=1~6), are the six calibrated joint rotation axes;

[0031] S2.6, calibrate the connecting rod torsion angle value: import the measured data points into the SA software, calibrate the robot base coordinate system and the joint rotation axis, and after calibration, obtain the connecting rod torsion angle value according to the angle formula between two straight lines;

[0032] S2.7. Calibrate the tool coordinate system. The tool coordinate system is determined by translating the lens center coordinate system of the terahertz probe forward by the focal length f along the vertical line of the lens, and is used as the tool coordinate system.

[0033] Preferably, in step S3, a virtual space is established and a path planning is performed for the robot, and the method is as follows:

[0034] S3.1, parameter space sampling and normal vector calculation: perform NURBS modeling, uniformly sample a series of points in the parameter space of the NURBS surface as the starting point and end point of the arcuate path segment, and calculate the normal vector N(u,v) for each sampling point;

[0035] S3.2, generate an arcuate path segment: based on the starting point and end point obtained in S3.1, generate an arcuate path segment with the normal vector of the sampling point as the direction, and adjust the length and curvature of the arcuate path segment according to actual needs;

[0036] S3.3, path smoothing: connect all the arched path segments to form a complete detection path, and perform smoothing to ensure the continuity and smoothness of the path;

[0037] S3.4. The mathematical expression of the arcuate path segment is expressed as:

[0038] P(t)=S(u i ,v i )+tN(u i ,v i ) (0≤t≤L)

[0039] Among them, S(u i ,v i) is the coordinate of sampling point i, N(u i ,v i ) is the normal vector of sampling point i, L is the length of the arcuate path segment, and t is the parameter variable;

[0040] S3.5. After the detection path is generated, the robot model is established according to the software: the simulation space is established according to the calibrated tool coordinate system and workpiece coordinate system, the detection trajectory of the robot is simulated in combination with the surface normal at the interpolation point, and the script file generated by the offline programming post-processing is copied to the robot.

[0041] Preferably, in step S5, a terahertz signal analysis model is established by the following method:

[0042] S5.1. Establish a terahertz wave multilayer sample thickness analysis model;

[0043] S5.2. Based on the influence of sample surface roughness on terahertz signals, the Kirchhoff roughness model is introduced to modify the terahertz wave multilayer sample thickness analytical model, and the frequency domain equivalent reflection signal model of multilayer coatings is obtained;

[0044] S5.3. Perform inverse Fourier transform on the frequency domain equivalent reflection signal model of the multilayer coating to obtain the time domain equivalent reflection signal of the multilayer coating based on the surface roughness of the sample.

[0045] Preferably, in step S6, obtaining the thickness measurement value of the thermal barrier coating on the curved surface of the sample includes the following sub-steps:

[0046] S6.1. Use total reflection of a silver mirror to obtain a reference signal, combine it with the parameters to be optimized initialized by the optimization algorithm, and input it into the terahertz signal analysis model;

[0047] S6.2, calculating the input parameters through the terahertz wave multilayer sample thickness analytical model established in S5.1 to obtain a simulation signal of the terahertz analytical model;

[0048] S6.3, processing the simulation signal by using an improved adaptive teaching and learning optimization algorithm;

[0049] S6.4. Compare the optimized model simulation signal with the real terahertz detection signal of the thermal barrier coating, and the fitting difference between the simulation signal and the detection signal is the objective function;

[0050] S6.5. If the objective function does not meet the convergence conditions, return to S6.2 and repeat the optimization process from S6.2 to S6.4, update the parameter value to be optimized, until the objective function meets the convergence conditions, output the optimal parameters, and obtain the measured value of the thermal barrier coating thickness on the curved surface.

[0051] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0052] 1. The terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings of the present invention adopts high-precision calibration technology based on point measurement and reverse digital-analog reconstruction technology to construct a sample point cloud model, and uses the NURBS surface method to extract the normal vector of the irregular surface of the blade; the surface detection path planning technology is used to realize the vertical scanning of the sample by the probe at the end of the robotic arm, ensuring the measurement accuracy and stability, so that the detection can adapt to complex surfaces.

[0053] 2. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings of the present invention establishes a terahertz signal analysis model for multilayer coatings, combines modern heuristic algorithms with terahertz mechanical models, performs data analysis and processing, and realizes accurate evaluation of thermal barrier coating thickness. The model takes into account factors such as the complex refractive index and surface roughness of the coating material, thereby improving the accuracy of the measurement.

[0054] 3. Aiming at the problem that the algorithm is prone to fall into local optimum and convergence speed is slow in the process of analytical model thickness measurement, the present invention proposes an adaptive teaching and learning optimization algorithm (ATLBO). By improving the population initialization, teaching stage and learning stage, the diversity and convergence speed of the algorithm are improved, thereby achieving more efficient and reliable non-destructive thickness measurement of coatings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a schematic diagram of the structure of the terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings of the present invention;

[0056] Figure 2 This is a schematic diagram of the detection path trajectory of the present invention;

[0057] Figure 3 This is a terahertz time-domain spectrum signal diagram of the present invention;

[0058] Figure 4 This is a schematic diagram of terahertz transmission of the present invention;

[0059] Figure 5 The flowchart of the algorithm optimization based on the analytical model of the present invention;

[0060] Explanation of the numbers in the figure: 1-trinocular vision measurement system; 2-six-axis robotic arm; 3-terahertz probe; 4-tooling; 5-sample; 6-robot, 7-base. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the application is further elaborated in detail below in conjunction with the accompanying drawings. The described embodiments are only a part of the embodiments involved in the present invention. All non-innovative embodiments of other researchers in the field on this embodiment belong to the protection scope of the present invention. At the same time, for the step numbering in the embodiment of the present invention, it is only set for the convenience of explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0062] In one embodiment of the present invention, a terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings is provided. Figure 1 As shown, it includes: a trinocular vision measurement system 1, a sample 5, a robot 6, and a base 7;

[0063] The trinocular vision measurement system 1 is used to calibrate the motion parameters and tool coordinate system of the robot 6. The robot 6 includes a six-axis robotic arm 2, a terahertz probe 3, and a tooling 4. A flange is also provided at the end of the robot 6. The flange is connected to the base through positioning bolts, and the sample 5 is fixed to the base 7 through a clamp.

[0064] One end of the six-axis robotic arm 2 is connected to a flange, and the other end fixes the terahertz probe 3 to the six-axis robotic arm 2 through a tooling 4. The robot 6 carries the terahertz probe 3 through the six-axis robotic arm 2 to scan the sample 5. The terahertz probe 3 can emit and receive the terahertz signal reflected from the surface of the sample 5, and transmit it to the host computer for signal processing.

[0065] By scanning the sample 5 with the terahertz probe carried by the six-axis robot 2, the terahertz signal can be detected perpendicular to the surface of the sample 5, meeting the detection requirements of curved surface samples.

[0066] Preferably, the tooling length adopted in this embodiment is 112 m, which is consistent with the focal length f of the lens of the terahertz probe 3 , so as to ensure that the signal reflection point is the point with maximum energy, making the signal more accurate.

[0067] Based on the above-mentioned non-destructive thickness measurement system for curved thermal barrier coating, non-destructive thickness measurement of curved thermal barrier coating of the sample is carried out, which specifically includes the following steps:

[0068] S1. Obtain the point cloud information of the sample for modeling

[0069] S1.1. First, arrange as many reference points as possible on the surface of the sample 5 as the basis for point cloud stitching during scanning, and fix the sample on the tooling.

[0070] S1.2, then scan the sample with ATOS optical scanner, and select 6 feature points with coordinates of P i (xi ,y i ,z i )(i=1~6), in the real space, the corresponding feature point coordinates measured by the probe are P j (x j ,y j ,z j )(j=1~6).

[0071] The least squares criterion based on feature point matching is as follows:

[0072]

[0073] Through feature point matching, the digital model obtained by inverse can be matched to the actual space to obtain the matrix transformation relationship between the workpiece coordinate system and the robot base coordinate system.

[0074] S1.3. After processing, export the STL file and perform reverse engineering in Catia software to obtain the CAD model of the sample.

[0075] In particular, the normal vector of the surface at a certain point is obtained through NURBS surface normal extraction, which is used for subsequent planning of the robot's posture.

[0076] S2. Calibrate the robot and tool coordinate systems

[0077] In this embodiment, considering the installation position of the robot and the test environment factors of the experiment, the solution of lens forward detection is adopted. Before calibration, the three-eye vision measurement system 1 and the six-axis robot arm 2 need to be aligned as follows: Figure 1 way to install.

[0078] S2.1. First, calibrate the robot base coordinate system through the trinocular vision measurement system 1:

[0079] Place the robot in zero position, with the joint angles at 0°, -90°, 0°, -90°, 0°, 0°;

[0080] S2.2, keep the angles of other axes unchanged, rotate the first axis, measure the coordinate values ​​of a fixed point on the robot end flange at regular intervals, and fit a circle C1 based on these points;

[0081] The center of the circle is determined according to the fitted circle, and the straight line l1 passing through the center of the circle and perpendicular to the plane where the circle C1 is located is used as the Z axis of the base coordinate system;

[0082] S2.3. Reset the robot to zero position, keep the angles of other axes unchanged, rotate the second axis, measure the coordinate values ​​of the fixed points of the robot end flange at certain angles, and fit a plane M based on these points;

[0083] S2.4, Measurement base plane B:

[0084] The intersection of the Z axis and the plane where the base is located is taken as the origin of the base coordinate system, and the normal of plane M through the origin is taken as the Y axis. Finally, the X axis is determined according to the coordinate system rule. In the process of establishing the base coordinate system, it is important to keep the direction of the coordinate axis consistent with the direction of the coordinate system specified in the model.

[0085] After the base coordinate system is established, the robot base coordinate system is used as the working coordinate system, and the point coordinate values ​​measured subsequently are all values ​​in the base coordinate system.

[0086] S2.5. When calibrating the connecting rod torsion angle value, the method is similar to the method of calibrating the base coordinate system:

[0087] Each time, the robot is placed in zero position, and each axis is rotated separately. Each time it rotates a certain angle, the coordinate value of a fixed point at the end flange is measured, and a circle C is fitted in the SA software. k (k = 1 to 6), and draw perpendicular lines through the center of each circle to the plane where each circle is located, and obtain the straight line l k (k=1~6), are the six calibrated rotation axes. In order to obtain more accurate fitting results, each rotation axis is rotated as far as possible within the robot's working range.

[0088] S2.6. Import the measured data points into the SA software and calibrate the robot base coordinate system and the joint rotation axis. After calibration, the connecting rod torsion angle value is obtained according to the angle formula between two straight lines.

[0089] In this embodiment, 20 sets of joint angle data are collected in the robot operation space in the above manner, and the first 15 sets of data are selected to solve the equation AΔX=b, and the last 5 sets of data are used to verify the accuracy of the calibrated kinematic parameters.

[0090] Among them, A represents a 3n×12 matrix, ΔX represents the robot arm coordinate parameters to be solved, and b represents the deviation between each set of actual values ​​and nominal values. The expression is:

[0091]

[0092] ΔX=[Δa0 … Δa5 Δd1 … Δd6] T

[0093] b=[ΔP 1x ΔP 1y ΔP 1z … ΔP nx ΔP ny ΔP nz ] T

[0094] Among them, a0…a5 is the length of the six-axis connecting rod, and d0…d5 is the six-axis connecting rod offset;

[0095] S2.7. The tool coordinate system can be determined by translating the lens center coordinate system forward along the vertical line of the lens by the focal length f, as the tool coordinate system.

[0096]

[0097] in, 6 T tool Represents the transformation matrix of the tool coordinate system, toolx, tooly, and toolz represent the center coordinates of the end flange respectively.

[0098] S3. Establish virtual space and plan the path of the robot

[0099] S3.1, parameter space sampling: uniformly sample a series of points in the parameter space of the NURBS surface as the start and end points of the arcuate path segment.

[0100] Calculate the normal vector: For each sampling point, calculate its normal vector N(u, v).

[0101] S3.2, generate an arcuate path segment: take the sampling point as the starting point and the end point, and take the normal vector of the point as the direction to generate an arcuate path segment. The length and curvature of the arcuate path segment can be adjusted according to actual needs.

[0102] S3.3, path smoothing: connect all the arched path segments to form a complete detection path, and perform smoothing to ensure the continuity and smoothness of the path, as shown below: Figure 2 The trajectory shown.

[0103] S3.4. The mathematical expression of the arcuate path segment can be expressed as:

[0104] P(t)=S(u i , v i )+tN(u i , v i ) (0≤t≤L)

[0105] Among them, S(u i , v i ) is the coordinate of the i-th sampling point, N(u i , v i ) is the normal vector of the point, L is the length of the arc path segment, and t is the parameter variable.

[0106] S3.5. After the detection path is generated, the robot model is established according to the written software.

[0107] The simulation space is established according to the calibrated tool coordinate system and workpiece coordinate system, and the detection trajectory of the robot is simulated in combination with the surface normal at the interpolation point. The script file generated by the offline programming post-processing is copied to the robot.

[0108] S4, the robot scans the sample to obtain a terahertz signal;

[0109] In this embodiment, after the robot finishes scanning the sample, the following is obtained: Figure 3 The terahertz time-domain spectroscopy signal is shown.

[0110] S5. Establish a terahertz signal analysis model;

[0111] Schematic diagram of terahertz wave propagation, such as Figure 4 As shown in (a) in the figure, a terahertz wave multilayer sample thickness analysis model is established:

[0112]

[0113] Among them, r 01 represents the projection coefficient of the coating interface I, w represents the frequency of the incident signal, e represents the natural logarithm, = represents the phase factor and is the frequency domain equivalent reflection coefficient of the interface between the lth layer and the l-1th layer, and E0(ω) represents the frequency domain signal of the incident signal.

[0114] Considering the influence of the surface roughness of the sample on the terahertz signal, the Kirchhoff roughness model is introduced, and the formula (7) is modified to obtain the frequency domain equivalent reflection signal model of the multilayer coating, which is expressed as:

[0115]

[0116] in, is the frequency domain equivalent reflection coefficient of the interface between the lth layer and the l-1th layer, t 01 , t 10 is the transmission coefficient, r 12 , r 10 is the reflection coefficient, λ is the wavelength of light, and δ is the root mean square roughness.

[0117] For E total (ω) can be transformed by inverse Fourier transform to obtain the time domain equivalent reflection signal E of the multilayer coating considering the surface roughness of the sample total The expression of (t) is:

[0118]

[0119] in, represents the inverse Fourier transform.

[0120] S6. Optimizing the time domain signal of the terahertz analytical model through algorithms

[0121] The terahertz detection signal is fitted by the analytical model optimization algorithm to obtain the corresponding parameters including the coating thickness, such as Figure 5 As shown, the specific steps are as follows:

[0122] S6.1. Use total reflection from a silver mirror to obtain a reference signal and the parameters to be optimized initialized by the optimization algorithm and input them into the terahertz analytical model.

[0123] S6.2. Perform corresponding calculations on the terahertz analytical model obtained above through formula (7) to obtain the simulation signal of the terahertz analytical model, such as Figure 4 As shown in (b) in .

[0124] S6.3, process the simulation signal through the adaptive teaching and learning optimization algorithm:

[0125] The traditional teaching and learning optimization algorithm has poor individual diversity in the initialization process, slow optimal sister search process and easy to fall into local optimum, resulting in low solution efficiency, poor accuracy and stability. Therefore, an adaptive teaching and learning optimization algorithm is proposed.

[0126] S6.3.1. Population initialization

[0127] This embodiment proposes an improved Kent chaotic mapping, which transfers chaotic variables in the most concentrated interval to fewer intervals to ensure uniformity and improve the diversity of the initial population:

[0128]

[0129] Among them, μ is a random number between [0, 1], z n is the nth chaotic variable, with a value range of [μ, 1], l min is the lower bound of the minimum interval, l max is the upper bound of the minimum interval, rand is a random number between [0, 1], The current chaotic variable is in the interval with the largest number.

[0130] The formula for population initialization based on the improved Kent chaotic map is:

[0131]

[0132] Among them, x i is the initial value of the ith particle, l m is the upper bound of the search space.

[0133] S6.3.2 Teaching stage

[0134] Compared with the traditional teaching stage, in which individual students update their fitness values ​​according to the teacher, which easily leads to the problem of falling into the local optimum, this embodiment improves the teaching stage. In the early stage of iterative calculation, for individual students whose fitness is lower than the average value, their learning step size is increased to increase the global optimization ability, thereby obtaining a larger search space. The formula is as follows:

[0135]

[0136]

[0137] Among them, γ is a random number between [2, 3], which is used to increase the learning step size.

[0138] S6.3.3. Learning phase

[0139] Compared with the traditional learning stage, which uses random mutual learning between learners to obtain progress, this embodiment improves the learning stage and introduces suboptimal individuals (student individuals x j ) and fitness update status supervised learning stage performance. If it is in the late stage of iterative calculation and the fitness has not been updated for m consecutive generations, it means that the algorithm is weak in approaching the optimal solution, and suboptimal individuals should be used to help the population improve. The formula is as follows:

[0140]

[0141] in, is the suboptimal individual in the tth iteration, r2 is the supervision factor, which is used to characterize the influence of the suboptimal individual on the performance of the learning stage, and its value range is [0, 1].

[0142] S6.4. Compare the model simulation signal obtained after algorithm optimization with the terahertz detection signal. The fitting difference between the simulation signal and the detection signal is the objective function. At this time, the corresponding optimization problem is:

[0143]

[0144] P = [p i，lower , p i,upper ]

[0145] Where: E expreiment (t) is the terahertz detection signal, E theory (t) is the simulation signal of the terahertz analytical model, P is the set of parameters to be optimized, and p i,lower is the lower limit of the i-th parameter to be optimized, p i,upper is the upper limit of the i-th parameter to be optimized.

[0146] S6.5. If the objective function does not meet the convergence condition, the value of the parameter to be optimized is updated through the optimization measurement in the algorithm, and the process returns to S6.2. Otherwise, the optimal parameter at this time is output, and the optimal parameter is used as the measurement value of thickness, etc. to obtain the thickness measurement value of the thermal barrier coating on the curved surface of the sample.

[0147] In summary, the terahertz-based nondestructive thickness measurement system and method for curved thermal barrier coatings proposed in the present invention can adapt to the rapid detection of complex curved surface structures through dual robotic arm collaboration and path planning; by establishing an analytical model and combining it with an optimization algorithm, the detection accuracy is improved, thereby achieving high-precision nondestructive detection that adapts to complex curved surface structures.

[0148] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings, characterized in that: include: Trinocular vision measurement system (1), sample (5), robot (6), base (7); The trinocular vision measurement system (1) is used to calibrate the motion parameters of a robot (6) and a tool coordinate system. The robot (6) comprises a six-axis robot arm (2), a terahertz probe (3), and a tool (4). A flange is also provided at the end of the robot (6), and the flange is connected to a base (7) via positioning bolts. One end of the six-axis mechanical arm (2) is connected to the flange, and the other end is connected to the terahertz probe (3) via a tool (4); the robot (6) controls the terahertz probe (3) via the six-axis mechanical arm (2) to scan the sample (5); and the sample (5) is fixed on the base (7) via a fixture; The terahertz probe (3) emits a terahertz signal perpendicular to the surface of the sample (5), receives a terahertz signal reflected from the surface of the sample (5), and transmits the terahertz signal to a host computer for terahertz signal processing.

2. A terahertz-based nondestructive thickness measurement system for curved thermal barrier coatings, characterized in that: The tooling (4) has a length of 112 m, which is consistent with the focal length f of the lens of the terahertz probe (3), so that the terahertz signal reflection point on the surface of the sample (5) is the point with maximum energy.

3. A terahertz-based non-destructive thickness measurement method for curved thermal barrier coatings, applied to the system described in any one of claims 1 to 2, for non-destructive thickness measurement of curved thermal barrier coatings, characterized in that: The steps include: S1. Obtain the point cloud information of the sample to establish a model, perform feature point matching, and obtain the matrix transformation relationship between the workpiece coordinate system and the robot base coordinate system; S2. Calibrate the robot base coordinate system and tool coordinate system based on the lens forward detection method; S3, establish a virtual space, extract the surface normal vector of the sample, and plan the path of the robot; S4, the robot controls the terahertz probe to scan the sample according to the planned path through the mechanical arm to obtain the terahertz time domain spectrum signal of the sample; S5. Establish a terahertz signal analysis model; S6. Through the terahertz signal analysis model, the terahertz time-domain spectral signal of the sample is optimized by the adaptive teaching and learning optimization algorithm to obtain the thickness measurement value of the thermal barrier coating on the curved surface of the sample.

4. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 3, characterized in that: In step S1, a point cloud model of the sample is established as follows: S1.

1. Arrange multiple reference points on the surface of the sample as the basis for point cloud stitching during scanning, and fix the sample on the base; S1.

2. Scan the sample using the ATOS optical scanner; Under the ATOS optical scanner, the coordinates of 6 feature points are selected as P i (x i ,y i , z i )(i=1~6), in the real space, the corresponding feature point coordinates are measured by the probe and are P j (x j ,y j , z j )(j=1~6); S1.3, feature point matching based on the least squares method, expressed as follows: S1.

4. Export the STL file obtained after processing by the ATOS optical scanner, perform reverse reconstruction in the Catia software, match the digital model obtained by reverse reconstruction to the actual space, obtain the matrix transformation relationship between the workpiece coordinate system and the robot base coordinate system, and obtain the CAD model of the sample.

5. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 3, characterized in that: In step S2, the robot base coordinate system and tool coordinate system are calibrated as follows: S2.

1. Calibrate the robot base coordinate system through the trinocular vision measurement system, place the robot at zero position, and the joint angles of the six-axis robot arm are 0°, -90°, 0°, -90°, 0°, 0° respectively; S2.

2. Rotate the first axis of the six-axis robot arm, measure the coordinates of the fixed point of the robot end flange at a certain angle, fit the circle C1 according to the obtained coordinates, and determine the center of C1. The straight line l1 passing through the center of the circle and perpendicular to the plane where the circle C1 is located is used as the Z axis of the robot base coordinates; S2.

3. Reset the robot to zero position, keep the angles of other axes unchanged, rotate the second axis of the six-axis robot arm, measure the coordinate values ​​of the flange fixed point at the end of the robot at certain angles, and fit the plane M according to the obtained coordinate values; S2.

4. Calibrate the robot base coordinate system, measure the plane B where the base is located, take the intersection of the Z axis and the plane where the base is located as the origin of the base coordinate system, make the normal of plane M through the origin of the coordinate system as the Y axis, and determine the X axis according to the coordinate system rule; S2.

5. Calibrate the joint rotation axis: Each time the robot is placed in zero position, each axis of the six-axis robot arm is rotated separately, and the coordinate value of the end flange fixed point is measured each time it is rotated a certain angle, and a circle C is fitted in the SA software. k (k = 1 to 6), and draw perpendicular lines through the center of each circle to the plane where each circle is located, and obtain the straight line l k (k=1~6), are the six calibrated joint rotation axes; S2.6, calibrate the connecting rod torsion angle value: import the measured data points into the SA software, calibrate the robot base coordinate system and the joint rotation axis, and after calibration, obtain the connecting rod torsion angle value according to the angle formula between two straight lines; S2.

7. Calibrate the tool coordinate system. The tool coordinate system is determined by translating the lens center coordinate system of the terahertz probe forward by the focal length f along the vertical line of the lens, and is used as the tool coordinate system.

6. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 3, characterized in that: In step S3, a virtual space is established and a path planning is performed for the robot, as follows: S3.1, parameter space sampling and normal vector calculation: perform NURBS modeling, uniformly sample a series of points in the parameter space of the NURBS surface as the starting point and end point of the arcuate path segment, and calculate the normal vector N(u, v) for each sampling point; S3.2, generate an arcuate path segment: based on the starting point and end point obtained in S3.1, generate an arcuate path segment with the normal vector of the sampling point as the direction, and adjust the length and curvature of the arcuate path segment according to actual needs; S3.3, path smoothing: connect all the arched path segments to form a complete detection path, and perform smoothing to ensure the continuity and smoothness of the path; S3.

4. The mathematical expression of the arcuate path segment is expressed as: mathbfP(t)=S(u i ,v i )+tN(u i ,v i )(0≤t≤L) Among them, S(u i , v i ) is the coordinate of the i-th sampling point, N(u i , v i ) is the normal vector of the i-th sampling point, L is the length of the arcuate path segment, and t is the parameter variable; S3.

5. After the detection path is generated, the robot model is established according to the software: the simulation space is established according to the calibrated tool coordinate system and workpiece coordinate system, the detection trajectory of the robot is simulated in combination with the surface normal at the interpolation point, and the script file generated by the offline programming post-processing is copied to the robot.

7. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 3, characterized in that: In step S5, a terahertz signal analysis model is established as follows: S5.

1. Establish a terahertz wave multilayer sample thickness analysis model: Among them, r 01 represents the projection coefficient of the coating interface I, w represents the frequency of the incident signal, e represents the natural logarithm, β represents the phase factor, is the frequency domain equivalent reflection coefficient of the interface between the lth layer and the l-1th layer, and E0(ω) represents the frequency domain signal of the incident signal; S5.

2. Based on the influence of sample surface roughness on terahertz signals, the Kirchhoff roughness model is introduced to modify the terahertz wave multilayer sample thickness analytical model, and the frequency domain equivalent reflection signal model E of the multilayer coating is obtained. total (ω): Among them, λ is the wavelength of light, δ1 is the root mean square roughness; S5.3, for E total (ω) is subjected to inverse Fourier transform to obtain the time domain equivalent reflection signal E of the multilayer coating based on the surface roughness of the sample. total (t), the expression is: in, represents the inverse Fourier transform.

8. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 7, characterized in that: In step S6, the terahertz time domain spectrum signal is optimized by using the terahertz signal analysis model, and the method is as follows: S6.

1. Use total reflection of a silver mirror to obtain a reference signal, combine it with the parameters to be optimized initialized by the optimization algorithm, and input it into the terahertz signal analysis model. The parameters to be optimized include: the refractive index and extinction coefficient of each layer of coating; S6.2, calculating the input parameters through the terahertz wave multilayer sample thickness analytical model established in S5.1 to obtain a simulation signal of the terahertz analytical model; S6.3, processing the simulation signal by using an improved adaptive teaching and learning optimization algorithm; S6.

4. Compare the optimized model simulation signal with the real terahertz detection signal of the thermal barrier coating, and the fitting difference between the simulation signal and the detection signal is the objective function; S6.

5. If the objective function does not meet the convergence conditions, return to S6.2 and repeat the optimization process from S6.2 to S6.4, update the parameter value to be optimized, until the objective function meets the convergence conditions, output the optimal parameters, and obtain the measured value of the thermal barrier coating thickness on the curved surface.

9. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 8, characterized in that: In step S6.3, the improved adaptive teaching and learning optimization algorithm is as follows: S6.3.

1. Population initialization: Through the improved Kent chaotic mapping, the chaotic variables in the most concentrated interval are transferred to fewer intervals to ensure uniformity and improve the diversity of the initial population: in, is the nth new chaotic variable, with a value range of [μ, 1], μ is a random number between [0, 1], l min is the lower bound of the minimum number interval, l max is the upper bound of the minimum interval, rand is a random number between [0, 1], The current chaotic variable is in the interval with the largest number; The population is initialized based on the improved Kent chaotic map, and the formula is: Among them, x i is the initial value of the ith particle, l m is the upper bound of the search space; S6.3.2, teaching stage: In the early stage of iterative calculation, for students whose fitness is lower than the average value, the learning step size is increased to increase the global optimization ability and obtain a larger search space. The formula is as follows: in, is the position of student individual i in the search space in the tth iteration, r i is the learning step size, the value range is [0, 1], TF is the teaching factor, the value is a random integer between [1, 2], is the fitness mean, γ is a random number between [2, 3], used to increase the learning step size; S6.3.3, Learning stage: The performance of the learning stage is supervised by introducing suboptimal individuals and fitness updates. If it is in the late stage of iterative calculation and the fitness has not been updated for m consecutive generations, the suboptimal individuals are used to help the population improve. The formula is as follows: Among them, the suboptimal individual is the student individual x j , is the suboptimal individual in the tth iteration, r2 is the supervision factor, which is used to characterize the influence of the suboptimal individual on the performance of the learning stage, and its value range is [0, 1].

10. The terahertz-based nondestructive thickness measurement method for curved thermal barrier coatings according to claim 8, characterized in that: In step S6.4, the objective function corresponds to the optimization problem: P=[p i，lower ,p i，upper ] Where: E expreiment (t) is the terahertz detection signal, E theory (t) is the simulation signal of the terahertz analytical model, P is the set of parameters to be optimized, and p i，lower is the lower limit of the i-th parameter to be optimized, p i，upper is the upper limit of the i-th parameter to be optimized.

Images