Analysis of defects in carbon-carbon composite materials using digital microscope
In turn, the use of carbon-carbon composite materials (CCCM) with unique thermal, mechanical and erosive properties opens up wide possibilities for the development of rocketry, aerospace and engineering machinery, metallurgy and shipbuilding. By reducing the weight of the construction, CCCM provide an opportunity to increase the speed and the flight range of the missile, space and aviation aircrafts, and also allow to increase the operating temperature and service life of bearings, to raise the specific strain and temperature during hot pressing of refractory metals and compounds. Industrial CCCM include materials on the basis of the 3D reinforced structures of carbon polyacrylonitrile, hydrocellulose and pitch fibers bound by pyrocarbon, coke and hybrid matrix.
The quality of the surface layer is characterized by not only roughness, but also by the physical-mechanical parameters, the presence of foreign inclusions, burns, cracks, chips, fractures and other defects. Procurement operations (the stage of manufacturing the workpiece) do not provide the required accuracy and surface quality, therefore various machining methods are used.
Such quality indicators as estimated roughness and surface profile do not give the exhaustive information about the nature of the defects, and do not fully meet the specifications reflecting customer requirements, including don’t allow to estimate adequately the morphology (shape and depth) of characteristic defects. Therefore, the urgent task is improving the system of indicators of surface quality of composite materials, explanation of methods of control and development of software and hardware non-contact methods of identification and assessment of geometric defects of the surface of products of CCCM.
Based on the analytical overview of the currently used contactless control methods, the digital microscopy is chosen. To analyze the adequacy of the developed methods of control of surface defects, the sequence of technological processes of manufacture of the product is partially simulated.
Using control samples, the following process steps of obtaining the product of Argolon – 4DL (Fig.1) were studied:
• mechanical CNC machining in different modes (from operating to forced);
• deposition of the anti-oxidation coating of pyrolytic silicon carbide from the gas phase of methylsilane. The same coating should be deposited on all samples . This process step is caused by the need to protect the product from oxidation during maximum time, which slows down the combustion due to air friction during the flight of the product;
• high temperature processing of the samples with coating in the same temperature regime to neutralize the effect of thermal shock and rapid delamination of the material and to further mathematical processing of experimental results.
After completing each of the above-described process steps, the quality control of the product surface is carried out. The author explains the set of indicators of the quality of the surface of the test material after each of the process steps in the sequence shown in Fig.1.
Workpieces are made by compacting the frame that is assembled from CFRP rods with a diameter of 0.7 mm by coal-tar pitch according to GOST 10200 with followed carbonization and high temperature treatment. In the plane perpendicular to the Z-axis, the rods have a layer-by-layer stacking (Fig.2). In each layer the rods are oriented along one of the axes X, X’, X", angles between them are 120°. Layers with different reinforcement directions follow each other alternately. The surface layer of silicon carbide or silicon can inherit the pattern of the surface of the initial billet of Argolon – 4DL caused by its core structure.
The existing control system uses mainly qualitative indicators. The proposed complex also includes quantitative indicators, including new ones that were not controlled in accordance with the requirements of used technical specifications (TS), which is its principal feature.
The analysis of TS have revealed that control of defects is implemented with use of universal means of measurement and partly is carried out visually.
It is proposed to control the following defects after mechanical treatment:
• chipping of material – violation of geometry of the workpiece (particularly on the edges) due to the material property;
• chipping of rods – removal from the surface of the workpiece of the individual rods during mechanical action;
• holes in the material that are larger than the structural cells;
• chip – a violation of geometry of the workpiece (particularly on the edges) due to mechanical action;
• surface crack – a discontinuity of material on the surface of the workpiece, which is not a structural feature of the material. Crack length tens of times exceeds width;
• impurities – elements that are not typical for the material and cannot be separated from the surface without the use of mechanical means.
According to the currently used specifications, the surface quality is monitored for the presence of cracks, pores and chips visually and using a vernier calliper according to GOST 166 with a measurement error not more than ±0.05 mm, and indicator depth gage according to GOST 7661.
By means of microscopic and mathematical analysis of the results of the given above process steps, it is advisable to determine the correlation or functional dependence between the modes of mechanical processing of Argolon – 4DL, quality of the anti-oxidation coating and the erosion resistance of the material at high temperature processing.
20 samples of the Argolon – 4DL were selected for the experiment.
Before experimental study, a preliminary analysis of surface defects on test samples using the photographs obtained with a digital microscope was performed. The defects were classified to define criteria of evaluation in the context of the development of metrological software. Image defects are represented in Fig.3.
For a better visual representation of defects, it is proposed to consider three-dimensional visualization of digital photographs of the sample surface. Using the developed software, the photograph was converted to monochrome, then was extracted brightness values of each pixel (0 to 255). Brightness values were deferred on the applicate axis, and the abscissa and ordinate axes correspond to the grid of digital photography.
Presented in Fig.4 data shows that the brightness of the rod elements of the frame that are on the surface, provides a significant contribution to the values on the applicate axis. Sections of the core frame have a greater lightness relative to the rest of the image. In reality, however, these elements correspond to smooth, flat areas (roughness Ra = 0.16 µm), obtained by mechanical machining. Thus, the identification of "true" surface defects may be difficult during interpreting and processing of such diagrams.
Given the above, the development of hardware and software algorithm, which would allow to obtain an image suitable for quantitative analysis of surface defects of CCCM is appropriate.
Fig.5a shows a photograph of the surface of a material specimen with single defect. Fig.5b shows a 2D representation of pictures of the surface where the Y axis is the brightness (form 0 to 255 units), the X axis is the sequence number of the pixel (counting of values starts from the top left corner of the picture and goes across rows). Fig.5c is a 3D representation of the surface where the X axis is the horizontal sequence number of the pixel in the photo, Y axis is the vertical sequence number of the pixel, the Z axis is the brightness (0 to 255). Fig.5d shows a 2D view in cross-section, where the Y axis is the brightness (0 to 255), the X axis is the horizontal sequence number of the pixel.
Fig.5b shows that the range of values with the lowest brightness corresponds to the existence of the defect, so in the picture you can count the number of defective areas across the controlled surface, and the nature of the graph allows to determine the characteristic averaged shape of the detected defects.
Fig.5d shows a cross-section in the center of the defect. It is seen that the lowest brightness (19 units), that is, the darkest area of the object on the digital image, corresponds to the deepest point on the sample surface. In addition, the presented graphical model allows to identify the boundaries of the chipping, which, in this experiment range from 40 to 70 pixels.
In addition to the digital microphotography, the contact method using FARO ARM device (Fig.6) that is certified in the state register of measuring devices was used. This method allowed a comparative assessment of the conformity of the actual brightness values to the depths and protrusions on the surface. The FARO ARM accuracy is ±0.035 mm, but in a small area of measurements it can be reduced to ±0.015 mm, which is sufficient for experiments. In the future, it is proposed to solve the problem of reliable determination of defects depth by entering the calibration operation, the method for which is being developed.
Using ARM FARO coordinate measuring manipulator, the primary data were obtained in Delcam PowerInspect software. In digital microscope, the primary data were obtained using Python and Octave programming systems. Mathematical processing of data was performed in Excel.
For a comparative assessment, measuring the surface topography was performed by the contact method (needle probe) on the same line from the edge of the object (Fig.7). The length of the measurement field is 8.39 mm.
Fig.8a shows the result of measuring the depth of the hollows by contact method, and Fig.8b – data obtained with the digital microscope. In both cases, the step of measurement was 0.01 mm. In these graphs it is seen that in general, the nature of the curves in the trend lines coincides. The scatter of the data obtained with the digital microscopy is probably caused by the noise of registered matrix, not optimal exposition and glares on the surface of the material (it is possible that the material should be matted by a special spray). To bring the contact measuring and photographic data in the same coordinate system, a scale measure was used upon receipt of data from the microscope, and then the weight per pixel was determined, which is taken for the sampling rate of the receive data, or a step, and is equal to the indicated above 0.01 mm. A graph obtained using the contact method of measurement, originally had uneven step (from 0.02 to 0.07 mm), so the data were fitted using the method of piecewise linear approximation in Excel.
The measurement data were mapped in such a way that the X axis represents the measured depths and ledges, and the Y axis – the brightness of pixels in photos, which can take values from 0 to 255.
Fig.9a shows some scatter of the data. To simplify the presentation, 839 obtained values were averaged to 42 values with intervals of averaging comprising 20 values. The formulas for the average values for the intervals are presented below:
where xi and yi are the values of depths (or heights of protrusions) and brightness with step of 0.01 mm at the same distance from the edge of the test material, respectively.
The coefficient of determination R2 is calculated by the formula:
where yi are real values of y in each observation, ŷi is the value predicted by the model, ȳ is the average of all actual values of yi.
R2 shows how the conditional variance of the model differs from the dispersion of the actual y values. If this ratio is close to 1, then conditional variance of model is fairly small and it is highly likely that the model well describes the data. If the R2 is much less than 1, the model likely does not reflect the real situation.
In our case, R2 = 0,89, which generally indicates the correctness of the assumptions about the functional relationship between increasу of depth and decreasу of brightness of the pixel and increasу of height and the increase of brightness of the pixel.
The conditions for obtaining images with optimal information about the defects can be summarized as follows:
The defect must fully get into the field of view of the microscope, starting from the borders to the deepest cavity.
The sensitivity needs to be configured to show threshold values of brightness at the boundaries and at the deepest point of the object (the same is true for projections).
Due to aberrations due to inaccuracies of manufacture of the lenses of the microscope, the defective area must be placed in a central area of the field of view of the microscope.
The whole analyzed range of values of the depths and heights must be within the specified depth-of-field.
For detailed measurement of defects, the designed measurement system should have the possibility to move the microscope relative to the sample surface. Accordingly, the amount of movement of the center of the CCD should be monitored by special automated measuring arrays.
Lighting should be set for elimination of glares and shadows, which are one of the main problems to obtain the correct input data for the analysis. It may be appropriate matting the surface of the material using the spray.
To simplify the measurement system and make it more efficient, it makes sense to use machine learning techniques, which will classify the objects. It is also advisable to program the search algorithm and counting of the defects and of their metric characteristics. In certain approaches (neural network), the classification, identification and quantification of surface defects can be conducted in an automated mode that will provide a significant advantage over conventional measuring systems.
Thus, the results of these studies are as follows:
• definition of the studied parameters of quality of surface of Argolon – 4DL: roughness, cracks, cavities, chipping, porosity, waviness;
• definition of the task of identifying the impact of these parameters on quality of the anti-oxidation coating;
• definition of the task of software development to get the correct data about the object and to quantify the most typical surface defects;
• definition of the task of assessing the internal structure of the material on digital images of faces of the workpiece;
• definition of the task of development of methods of metrological control based on the data of digital photos;
• definition of the task of the analysis of errors of the received data. ■