Obtaining the percentages of ductility and brittleness of AISI/SAE 1020 and 304 steels, using digital image processing.

This research is the integration of a set of elements in a system of capturing, processing, and digital image analysis. It allows a better visual and numeric interpretation to determine the toughness, ductility percentage, and fragility of steel AISI/SAE 1020 and 30, getting better qualitative and quantitative observation of the results from the impact test (Charpy test). Patterns in the standard form, ASTM E23, were digitalized to evaluate the percentage of ductility/fragility of specimen testing. After, we calculated the area and the equivalent diameter of the material. using digital image processing and numerical comparison between the patterns specified in the standard form ASTM E23 and the testing in the impact test, and it allows to find the kind of pattern it is closest and determine which degree of ductility. Finally, the results were compared by three experts. The algorithm accuracy was 80%.


Introduction
Digital image processing has been used in multiple applications[1], [2], [3], [4].The applications refer to topics such as fruit selection [5], classification of medical images [6], [7], structural analysis [8], [9], [10] and materials classification [11], [12].Secondly, the Charpy test , has been used to determinate toughness [13], behavior analysis in thermic variation [14], and identification of different properties of materials [15].For the design of machine elements, it is important to evaluate characteristics such as ductility and brittleness in order to calculate strength [16], [17].One of the main drawbacks in determining brittleness and ductility is that it is associated with direct observation by experts through microscopic images.Human subjectivity is always a problem.In this work , we was estimated fragility and ductility using the ASTM E23 standard [18].

Impact Charpy Machine and Specimen test
The Charpy test is used in order to determine the impact energy on a test material.This is related to the stress perceived by a test specimen, and allows to make a correct selection of material according to end use[1], [18].Figure 1, shows the impact Charpy machine.Specimens specifications in Figure 2.  [18].Afterwards, both the master images and the fractures of the specimens were digitized using a kanon camera with 8 mega pixels of resolution.To improve the definition, a preprocessing of the fractures of the specimens was necessary, which consisted of adding graphite to the surfaces of the 3 fractures (see an example in figure 5).Finally, the application of the digital image processing algorithm (see figure 4). Figure 3 shows the flow diagram for methodology.Figure 4 shows the image processing algorithm.The process is based on morphological operations such as opening, closing and segmentation [3], [4], [19].The objective was to obtain binarized images in order to extract characteristics such as the equivalent area and diameter [20].Comparison was made using Euclidean distance (see equation 1) which is an tool to reasoning case based [21], [9], [22].To obtain good results, five repetitions were necessary, according to [23]. Figure 6 and 7 shows the results of the images processing.

Digitization of pattern images
Figure 6 shows the patterns of the standard ASTM E23. Figure 7 shows the capture process.3.4 Ductility / fragility Obtention, using Reasoning case based Figure 11 shows an example of the results.On the left side, the minor distance was to first image of the standard ASTM E23.This means, that the assessment Charpy test specimen is closely near at the first pattern (ductility 10% see figure 6).On the right side, the minor distance was the seventh (ductility 70%).Tables III and IV show that the percentage of ductility obtained by the algorithm is variable depending on the specimen.This indicates that being the same type of material, different results are observed.However, with the area (Tables I and II) as a descriptor the results are very uniform.In tables V and VI, the error observed was 20%.

Conclusion
The application of graphite in the ASI 304 specimens turned out to be an alternative determinant to carry out the image processing, allowing the software to differentiate the areas formed in the test and thus subsequently go on to make a numerical comparison with the areas formed and the Pattern image areas using Euclidean distance.Evaluation by expert in materials was the determining factor for determining the degree of reliability of the algorithm since it allowed to know with certainty which descriptor of form (Equivalent area or diameter) was ideal for carrying out this type of image processing.A very low error rate (20%) has been presented when area is used as a shape descriptor.However, it is valid clarify that the percentage of error is supported by the application of ranges discrete that the ASTM E23 standard has.Likewise, it was evidenced that the equivalent diameter is not the most adequate descriptor for this type of application, since the results were not in agreement with the results of the experts.External factors at the time of image capture is a critical factor since any anomaly that occurs at the time of capturing the image will affect the digital processing of images for this it is necessary to eliminate all the noises or factors outsiders involved in the scene An open problem is still precision, consistency and efficiency of the algorithm.Future work will be focused on the development of a sophisticated method of deriving an optimized high-precision matching result under the influence of noise and illumination sources.

Figure 4 .
Figure 4. Digital image processing algorithm

Figure 8 .
Figure 8. AISI 304 steel (left) and AISI/SAE 1020 steel (right) Figure 9 upper panel (first-left) corresponds to original image in RGB format.Second-center is the histogram of Image resized (and gray-scale).In bottom panel it can see the binarized image (area) and perimeter.The equivalent diameter is obtained as: perimeter/4.

Figure 9 .
Figure 9. Processing of a pattern image using the algorithm.(Area and Equivalent diameter)

Figure 10 .
Figure 10.Processing of a Charpy test image using the algorithm.(Area and Equivalent diameter)

Figure 11 .
Figure 11.Graphical representation of the pattern closest to the tested image using the Euclidean distance and the shape descriptors Area (Left Image) and equivalent diameter (Right Image)

Table I .
Ductility percentage in each of the replicas in the ASI 304 steel using the area as a shape descriptor.

Table II .
Ductility percentage in each of the test in the ASI/SAE 1020 steel using the area as a shape descriptor.

Table III .
Ductility percentage in each of the replicas in the ASI 304 steel using the equivalent diameter as a shape descriptor.

Table IV .
Ductility percentage in each of the test in the ASI/SAE 1020 steel using the equivalent diameter as a shape descriptor.

Table V .
Comparison of results between the Algorithm and the experts in the area of materials in AISI 304 steel at room temperature.

Table VI .
Comparison of results between the Algorithm and the experts in the area of materials in AISI/SAE 1020 steel at room temperature.