METHOD FOR EVALUATING THE RESULT OF A SURFACE TREATMENT CARRIED OUT ON A WORKPIECE BY PARTICLE BLASTING
Patent Information
- Application Number
- DE502022001955
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-08-23
- Publication Date
- 2024-11-14
- Estimated Expiration
- 2042-08-23
AI Technical Summary
Current methods for assessing the coverage of particle-blasted surfaces in the aviation and automotive industries are unsatisfactory due to low accuracy, flexibility, and repeatability, with manual methods relying on human experience and automated methods struggling with precise classification near complete coverage.
A method using high-resolution optical measurements with white light interferometry to record the surface in a spatially resolved manner, determining the surface level and local gradient, and identifying areas not hit by particles based on predetermined thresholds, thereby calculating the uncovered area for accurate coverage assessment.
This approach provides a robust and accurate measurement of surface coverage, reducing errors and enabling reliable assessment of complete coverage, with an error of 1% in area determination leading to a falsification of only 0.02% when considering 98% coverage as completely blasted.
Description
[0001] The present invention relates to methods for assessing the results of a surface treatment performed on a workpiece by particle blasting. This finds application, for example, in the aviation and / or automotive industries. In particular, the invention relates to a method for automatically testing the coverage of particle-blasted, e.g., shot-peened, surfaces.
[0002] The aviation industry in particular places high demands on the reliability and safety of components while maintaining low weight. For this reason, a high proportion of rotating engine components, for example, are particle peened, particularly shot peened, to increase their fatigue strength. This involves bombarding the surface of the workpiece with abrasive material, usually steel, ceramic, or glass shot. The impact of the blasting material causes superficial deformation, which leads to the formation of a residual compressive stress layer on the workpiece surface. This residual compressive stress layer inhibits the formation and growth of surface defects. The resulting increase in fatigue strength depends largely on the degree of surface coverage. The degree of coverage refers to the percentage of the treated surface covered by impact spheroids.Accordingly, testing particle blasted surfaces for complete coverage, in practice this usually means coverage levels greater than or equal to 98%, is of particular importance.
[0003] The established methods to date are unsatisfactory due to their low accuracy, flexibility, robustness and lack of repeatability.
[0004] There are already various approaches to determining the quality of particle beam treatment of workpiece surfaces. One of the most widely used methods remains visual inspection using a magnifying glass or pocket microscope. Typical magnifications range from 10x to 30x. Here, a worker estimates the degree of coverage based on their experience and by comparing it to reference samples.
[0005] In a process known from US Pat. No. 3,950,642 A, a fluorescent paint is applied to the surface of a workpiece to be treated prior to blasting. The paint flakes off at the directly impacted areas. After blasting, the appearance of the workpiece is compared under UV light with prepared samples of known coverage, thus determining the coverage level.
[0006] US 5293320 A specifies the creation of a surface map by line-by-line tactile measurement of the surface. Contours, comparable to contour lines on maps, are searched for and classified in elevation sections of this surface map. The degree of coverage is determined based on the number and type of these contours.
[0007] US 5235517 A proposes a different approach. According to this approach, a line is measured tactilely on both the sample to be tested and a reference sample. The waveform and magnitude of the gradient are determined. The results from both samples are then compared, and a coverage ratio is derived.
[0008] In a method disclosed in US Pat. No. 5,581,483 A, a correlation analysis is performed on two ideally parallel line tracings of the surface to be tested. High correlation values are attributable to regular surface geometries such as turning marks. Low correlation values indicate random surface geometries, such as those caused by shot peening, and thus indicate high coverage levels.
[0009] DE 102010001286 A1 proposes taking a 3D surface image using a white-light interferometer or confocal microscope. Subsequently, the local heights caused by blasting are determined by subtracting the determined component shape from a known surface geometry of the untreated workpiece. Heights below the zero line are identified as impacts. The degree of coverage is then determined based on their surface area.
[0010] In a procedure proposed in FR 3100336 A1, a 3D depth map is created from photographic images of the blasted surface under different illumination directions. Deep areas whose variation is greater than or equal to 10% of the global variation in the image are identified as impact craters.
[0011] JP 2010017807 A proposes determining the degree of coverage by analyzing the height distribution of the individual points, although it is not specified which measuring instrument should be used for this determination. According to this document, the Gaussian distributions of the heights of blasted and unblasted surfaces have different widths and mean values; the measured height distribution is broken down into these two components, and the surface area portions are determined using the integrals under the distributions.
[0012] In a method according to JP 5511407 B2 or JP 2015160282 A, images of the surface are first taken using a telecentric lens with a spacer. Coaxial illumination is achieved via a beam splitter with an LED mounted orthogonally to the optical axis. In the method according to JP 5511407 B2, the impacts are identified using a simple brightness threshold, and the area is calculated by pixel counting. In the method according to JP 2015160282 A, the degree of coverage is determined using the average luminance (as the arithmetic mean of all pixel brightnesses) and corresponding calibration functions for the respective surface to be illuminated.
[0013] A procedure according to WO 2018185234 A1 exploits the correlation between the coverage and a parameter L* in the CIELAB color space. A corresponding correlation function is created using samples with different coverage levels. This document provides precise instructions for creating the images required for the procedure.
[0014] In a method according to WO 2019159424 A1, an image acquired by a camera is scaled to a standardized size, both in terms of resolution and the average diameter of the impacts. Color correction, filtering, and sharpening can then be performed if necessary. The evaluation is performed using a neural network. Coverage is classified in categories from 0% to 98%.
[0015] DE 102020108956 A1 proposes the detection of at least one impact geometry in order to calculate at least one quality parameter. This is to be used to control the blasting process. The detection system is described as an optical one, preferably based on the principle of white light interferometry. The detection system should be able to communicate with the blasting system control system, possibly with the assistance of a digital computer. No further details on the structure, algorithms, or method are provided. The known methods have the following disadvantages:
[0016] Manual testing methods for coverage testing, such as using a magnifying glass / pocket microscope, or even fluorescent dye, depend on individual factors such as experience, the worker's condition on the day, and their attentiveness. Precise and repeatable measurements are not possible. Furthermore, satisfactory documentation of the surface and the testing process itself is not possible. The use of fluorescent dye also leaves residues on the surface.
[0017] Previous automated testing methods typically rely on tactile line-by-line measurements or photographic recordings of the surface. Tactile surface measurement is time-consuming and is typically performed on measuring machines with limited measuring space and inflexible positioning of the probe tip. This makes testing large or irregularly shaped parts difficult or even impossible. Furthermore, the radius of the probe tip usually makes error-free scanning of the highly irregular blasted surface impossible.
[0018] Photographic recordings usually require strict and complex control of lighting conditions. Deviations from these conditions, as well as irregularities and contamination on the components being tested, lead to unreliable results.
[0019] Common to the known automatic methods is that they generally directly determine the covered portion of the surface. Based on the commonly used definition of 98% coverage as "fully blasted," a 1% error in determining the corresponding area leads to a 0.98% falsification of the measurement result.
[0020] Known automated approaches, especially those based on machine learning, have problems with sufficiently fine and accurate classification near complete coverage; their strengths lie instead in the range of low coverage levels.
[0021] This invention is intended to remedy this situation.
[0022] For this purpose, the invention proposes a method having the features of claim 1. Advantageous further developments are shown in the dependent claims 2 to 4.
[0023] Accordingly, in a method proposed here for assessing the result of a surface treatment performed on a workpiece using particle blasting, the surface treated by particle blasting is recorded with spatial resolution using an optical measuring method. This is ideally done with a high resolution. In particular, the resolution should be selected so that an average impact accounts for at least ten pixels in width and / or length, preferably at least twenty. For the smallest particles used in practice, this would be 7 µ m lateral minimum resolution, for the largest particles 0.13 mm. The height resolution should ideally be at least five times, especially at least ten times, as good as the above-mentioned minimum lateral resolution, especially the lateral resolution used. In particular, a lateral resolution of 5 µm or better and / or a height resolution of < 1 µ m in one or two lateral directions. The height, in particular, is recorded with spatial resolution. For each of the spatially resolved measurement points, a surface level and a value for a local gradient relative to neighboring measurement points are then determined. A surface level refers to the spatial position of a surface point relative to a reference point, e.g., a predefined origin point or an origin plane in the workpiece itself. The surface level thus provides a value for the elevation of the surface at the location of the respective measurement point.
[0024] Furthermore, according to the invention, the detected surface level is compared with a target level expected for the untreated workpiece at the measurement location of the measuring point, and a deviation is determined. The expected target level can be derived, for example, from a design model of the workpiece, but also from a measurement of the workpiece before treatment by particle blasting.
[0025] In a further step, those measuring points for which both the surface level detected according to the invention is above a level lower than the target level by a predetermined height threshold and the detected gradient is below a predetermined gradient threshold are then identified as measuring points not hit by a particle of the particle beam.
[0026] By dividing the number of measurement points identified as not being hit by a particle of the particle beam by the total number of recorded measurement points, a portion of the surface not hit by the particle beam is determined.
[0027] In the method according to the invention, unlike the known prior art methods, it is not the area covered by the particle beam that is detected and determined, but rather the uncovered area. Thus, an error of 1% in the area determination for a workpiece identified as fully irradiated with an assumed 98% coverage, i.e., 2% non-coverage, only leads to a falsification of the measurement result by 0.02%.
[0028] White-light interferometry can be advantageously used as an optical measurement method. This measurement technique has proven to be particularly robust and resistant to interference, especially from possible stray light. It can also produce high-resolution images of the treated surface with good spatial and height resolution.
[0029] Even if the surface images can generally be taken using a freehand measuring device, it has proven particularly useful if the spatially resolved recording of the surface treated by particle blasting is carried out using an optical measuring method with the aid of an optical measuring device arranged on an automated moving unit, in particular a robot, industrial robot and / or cobot (collaborative robot).
[0030] The method is advantageously carried out for a plurality of sections of the same surface, which are in particular disjoint. In particular, the sections are each recorded with spatial resolution, i.e. in particular the height in the sections is recorded with spatial resolution. The sections are in particular rectangular or square. In particular, the sections are arranged on a surface or along a line. In particular, the number of sections is between 10 and 500. In particular, the sections each have a size in the range from 10x10mm to 100x100mm. In particular, the sections are each spaced apart in the range from 1mm to 100mm and / or span a distance in the range from 5mm to 300mm and / or an area in the range from 10x10mm to 50x100mm. In particular, the sections are arranged equidistantly and / or regularly on the line and / or surface.This allows a particularly representative coverage to be determined with the least possible effort.
[0031] The evaluation described above is typically computer-aided, using software that performs the evaluation. This software can then archive the evaluation results, possibly together with the raw data, e.g., surface images, associated with the workpiece, as a kind of quality pass.
[0032] Since the deformation of the surface depends on both the material of the workpiece and the parameters of the particle blasting process, such as the geometric shape and size / size distribution of the particles, their material (e.g., glass, ceramic, steel, etc.), as well as the speed of the particle beam, which determines the momentum and energy of the particles upon impact with the surface of the workpiece, very different and individual impact patterns and geometries result for the blasted workpiece. Therefore, the values for the height threshold and the gradient threshold must be determined for a given combination of workpiece and beam parameters or approximately derived from a large database. As long as the diameter-to-depth ratio of the impact calottes remains similar, height thresholds and gradient thresholds obtained with other parameters can be used without problems.Advantageously, a set of height thresholds and gradient thresholds for different diameter-to-depth ratios of the impact spherical surfaces can therefore be obtained and / or stored. In particular, to implement the method according to the invention, the applicable height threshold and gradient threshold with the most similar diameter-to-depth ratio of the impact spherical surfaces, or an interpolation or a weighted average of the height thresholds and gradient thresholds with the most similar diameter-to-depth ratios of the impact spherical surfaces, is used.
[0033] For more precise determination, the values for the height threshold and / or the gradient threshold can preferably be determined based on an analysis of a reference pattern processed with a specific particle beam with predetermined particles and beam parameters and / or a simulation model of a processing pattern created for such a processing process and / or based on analyses of a plurality of reference patterns each processed with different particle beams (each with predetermined particles and beam parameters) and / or simulation models of processing patterns created for such processing processes.
[0034] The threshold values can be determined using statistical methods based on the height and gradient distribution in images of blasted surfaces on reference workpieces, ideally but not necessarily with minimal coverage. However, they can also be derived from a simulation based on considerations of the deformations expected due to impacts and induced residual compressive stresses.
[0035] The threshold values (height threshold and also the gradient threshold) can be understood, for example, as coordinates in a two-dimensional parameter space consisting of local height and local gradient, the measured degree of coverage CR G ( Z , Z ') forms the response variable.
[0036] To determine the threshold values to be applied, the mean value of the measured coverage can be calculated for several actual coverage levels (e.g. 60%, 80% and 100%) using several samples. CR Ĝ ,i ( Z,Z') including standard deviation σ i (Z,Z') are determined over the parameter space. The differences between the mean values of the measured coverage between two differently covered sample groups (over the parameter space) are then determined and then normalized with the sum of the respective local standard deviations. This results in a representation of the discrimination ability achievable with different threshold combinations. Sign ( Z,Z' ) over the parameter space. Sig Z , Z ′ = CR G ^ ,1 Z , Z ′ − CR G ^ ,2 Z , Z ′ σ 1 Z , Z ′ + σ 2 Z , Z ′
[0037] The result is given in standard deviations at the achievable distance between the measured values, which makes it easy to select optimal threshold combinations. The relationship between the measured and actual coverage can be established, for example, using a correlation function or look-up table. Alternatively, a separate, preferably predetermined, threshold can be used to distinguish between good and bad parts—that is, those with sufficient and those with insufficient coverage.
[0038] If, for example, a distinction is desired between 97% and 98% coverage, it is recommended to use samples with corresponding coverage in the range of 97% to 98% to determine the threshold values. The present invention can generally be assigned to the fields of testing technology, in particular automatic testing, and optical measurement technology for quality assurance. It preferably uses the optical measurement method of white light interferometry for high-resolution, planar imaging of particle-blasted surfaces. The preferred use of a collaborative robot (cobot) to guide the optical measuring instrument, such as a white light interferometer, enables both automatic positioning and measurement as well as manual guidance. The acquired surface data is then digitally processed.
[0039] With the method according to the invention, a high-resolution 3D point cloud of the surface to be tested can be generated, for example by means of a measuring instrument guided by a collaborative robot (cobot). As mentioned, a white light interferometer can preferably be used for this purpose. However, in principle, other methods for 3D surface measurement are also suitable. To facilitate the fine positioning of the measuring area on the component or workpiece to be tested, a laser aiming device consisting of a cross laser and a point laser can be used. The lasers can be positioned and aligned on the measuring device such that the center point of the cross and the point coincide at a tool center point of the measuring device. This makes it particularly easy for an operator to handle the measuring device and align it for the measurement.
[0040] For each point of the captured image, its surface level, i.e., its height, relative to the idealized workpiece surface is then determined, and the local gradient of the surface contour at that point is determined relative to neighboring points. According to the invention, points that lie above a height threshold and below a gradient threshold are identified as parts of the original surface not hit by the blasting material. The degree of coverage is then determined based on the uncovered surface portion.
[0041] In a method according to the invention, the images of the workpiece surface to be inspected can be taken, for example, using a measuring head of a white light interferometer mounted on a cobot, e.g. the CRX10IAL-30P-V-PN / 0 model from FANUC. The cobot can be arranged, for example, on a movable table. The necessary robot control and a computer for evaluating the images can be arranged on the table. The top side of the table can, in particular, also serve as a storage area for the components to be inspected. In locations with increased vibrations, it may be necessary to mount such a table on dampers. The measuring head with a white light interferometer, e.g. the H6 model from Heliotis, can be attached to a flange of the robot. The interferometer can be aligned either parallel or orthogonal to the axis of rotation of the flange.The measuring head can also contain a cross and a point laser, which form a targeting device. The lasers can be aligned so that the point and the center of the cross align at the interferometer's tool center point. The measuring head can be positioned automatically by the robot or using the hand-guided function. Once a recording is triggered, the data is transferred to the computer and evaluated.
[0042] To minimize errors, the interferometer must not exceed a maximum tilt relative to the surface normal of the workpiece during acquisition. The maximum permissible tilt depends on both the precise surface geometry and the required accuracy.
[0043] In a preprocessing step, several individual images can be combined to form a larger area if required. Implausible measurement points and noise can then be removed using a statistical outlier filter. Here, it is possible to replace missing points with plausible values using a modified Gaussian filter. The theoretical surface of the component can then be determined from the point clouds prepared in this way using a fitting algorithm. Alternatively, the theoretical surface can be taken from a design project for the production of the component. The local heights of the points can now be determined either by calculating the geometric distance to the theoretical surface or by subtracting its Z value at the same X and Y coordinates. The local gradients can then be calculated from the heights determined in this way, e.g. using a Sobel operator.All points that are above the height threshold and simultaneously below the slope threshold are identified as part of the uncovered surface.
[0044] The following two approaches can be used to determine threshold values: The threshold values can be read from the height and gradient histograms of a test measurement. Alternatively, the threshold values can be derived from theoretical considerations regarding the expected surface deformations for the specific method used and the specific workpiece. When using read threshold values, it is recommended to perform the test measurement on a sample with a low degree of coverage and then adjust the threshold values if necessary to correctly identify known fully and incompletely covered test pieces. It is important to determine separate threshold values for each combination of material and beam parameters.
[0045] Possible embodiments of the method according to the invention will be explained below purely by way of example using the following purely schematic figures. They show: Fig. 1 is a flow chart, Fig. 2 is a flow chart, Fig. 3 is a flow chart, Fig. 4 is an illustration of the determination of the height deviation and Fig. 5 is an illustration of the identification of measuring points.
[0046] The Figures 1 to 3 show flow diagrams of exemplary embodiments. In order to understand the core of the method shown, which, starting from the local height determination based on the height threshold value and the gradient determination and the determination of the gradient threshold value, determines the degree of coverage, i.e. the surface proportion of the original surface, as it is generally Figure 3 shown, the Figures 1 and 2 possible further details: This is how Fig. 3with white light interferometry, followed by data processing using SOR (Statistical Outlier Removal), which removes outliers based on their excessive distance from neighboring points.
[0047] This is followed by fitting with the known target contour (contour fitting) to determine the local elevations. For gradient determination, the data set is then completed using imputation, for example, by determining new measured values for the measurement points removed by SOR, for example, through interpolation. Fig. 2 shows a simpler procedure without SOR and imputation.
[0048] Fig. 4This diagram schematically illustrates contour fitting. The measured fine contour is shown in dashed lines. The coarse geometry, i.e., the planned contour, is shown in solid lines. The illustration is limited to a two-dimensional section. By determining the deviation between the fine and coarse contours, the local height can be determined (indicated by an arrow) at which the coarse contour was bent into a straight line / plane.
[0049] Fig. 5 illustrates the determination of the part not hit by the particle beam. Shown above is the local height profile along a two-dimensional section line. The arrow to the left indicates the determination of the part of the detected surface level that lies above a level lower than the target level by a specified height threshold, and the arrow to the right indicates the determination of the part that falls below a gradient threshold.
[0050] To determine the portion of the recorded surface level that lies above a level lower than the target level by a specified height threshold, the level is initially drawn to the left in dotted lines. The portion that exceeds the level is shown in dashed lines, and the portion that falls below it is shown in solid lines.
[0051] To determine the portion that falls below a gradient threshold, the gradient is determined. The portion that falls below the gradient threshold is shown as a dashed line on the right.
[0052] Below, the part is shown in dashed lines, which, based on the two previous analyses, is considered not to have been hit by the particle beam.
[0053] It is important to ensure that the upper and lower images are arranged laterally and without offset, while the middle images are offset to the left and right. It can be seen that, for example, the areas at the beginning of the craters are considered treated, even though they still exceed the level, because there is a steep gradient at the beginning of the crater. On the other hand, the areas of lower gradient at the bottom of the craters are also considered treated because they (significantly) fall below the level. This demonstrates that the method enables a particularly reliable assessment.
Claims
1. A method for assessing the result of a surface treatment carried out on a workpiece by particle jets, wherein (1) The surface processed by particle jets is recorded in a spatially resolved manner by means of an optical measuring method and for the spatially resolved measuring points in each case a. a surface level and b. a value of a local gradient in relation to neighbouring measuring points is determined, (2) wherein the detected surface level is compared with a target level expected for the untreated workpiece at the measuring location of the measuring point and a deviation is determined, and wherein the following steps are also carried out: (3.1) Identify those measuring points for which a. the surface level detected according to (1) a. is above a level that is lower than the target level by a specified height threshold value, and also b. the gradient detected in accordance with (1) b. falls below a specified gradient threshold, as measuring points not hit by a particle of the particle beam, (3.2) Dividing the number of measurement points identified in (3.1) as not hit by a particle of the particle beam by the total number of measurement points detected to determine a portion of the surface not hit by the particle beam.
2. A method according to claim 1, characterised in that the optical measuring method used is white light interferometry, laser light section sensors, a stereo chamber system and / or a shape-by-shading method.
3. A method according to one of claims 1 or 2, characterised in that the spatially resolved recording of the surface processed by particle jets is carried out by means of an optical measuring method using an optical measuring device arranged on a cobot.
4. A method according to one of the preceding claims, characterised in that the values for the height threshold value and / or the gradient threshold value are determined on the basis of an analysis of a reference pattern processed with a specific particle beam with predetermined particles and beam parameters and / or a simulation model of a processing pattern created for such a processing operation.
