Integrity Classification Algorithm of Images obtained from Impact Damaged Composite Structures

Problem statement: Many NDT systems used for damage detection in composites are difficult to apply to complex geometric structures, also, they are time-consuming. As a solution to the problems associated with NDT applications, an intelligent analysis system that supports a portable testing environment, which allowed various types of inputs and provided sufficient data regarding level of damage in a tested structure was designed and tested. The developed technique was a novel approach that allowed locating defects with good accuracy. Approach: This research presented a novel approach to fast NDT using intelligent image analysis through a specifically developed algorithm that checks the integrity of composite structures. Such a novel approach allowed not only to determine the level of damage, but also, to correlate damage detected by one imaging technique using available instruments and methods to results that would be obtained using other instruments and techniques. Results: Using the developed ICA algorithm, accurate classification was achieved using C-Scan and Low Temperature Thermal imaging (LTT). Both techniques agreed on damage classification and structural integrity. Conclusion: This very successful approach to damage detection and classification is further supported by its ability to correlate different NDT technologies and predict others.


INTRODUCTION
Carbon fiber composites are now fairly widely used in civilian and military applications. Delaminations are common defects found in these materials. Their presence leads to structural weaknesses, which cause failure of used components. It is important to develop effective nondestructive testing procedures to identify these defects and to increase the safety early enough to avoid catastrophic failure. Hence, all methods for fast and reliable inspection deserve special attention (Wolf et al., 2004;Gupta and Breitenstein, 2007;Gralewicz and Owczarek, 2005;Bohm et al., 2006;Montanini and Aliquo, 2009;Shah et al., 2006;Colvin, 2005;Tohgo et al., 2009).
Composite materials depend on their structural arrangement to obtain their desired mechanical properties. The fibers are generally of little practical use but with a well designed combination of fibers and matrix a reliable component with optimal performance is produced. Integrity of a material is based on quality of fabrication as the designed specifications are not only determined by the best available structural properties, but also any other combination of properties for a particular application. So, it is important to be able to tailor and optimize the manufacturing process with tools able to inspect defects that most often can have a marked effect on the component performance and functionality. Two main areas of concern in inspection and damage classification: • During manufacturing • In-service A developed inspection technique and classification algorithm should cover two main areas: • Critical damage identification and detection • Analysis and monitoring of damage progression Impact damage is a key issue in the design of composite structures where the impact event and extent are of importance. Damage occurs progressively during an impact and is a function of the impact event and structure resistance that is affected by material properties. Local and global effects need to be considered which gives an indication regarding the structure dynamic response. Method of impacting is also a factor where supported frames respond differently to impact compared to unsupported ones, indicating that boundary conditions significantly affect structure response and extent of damage.
The application of an impact can result in a dynamic stress which when established can induce a damage that propagates at a number of sites within the material thickness. Composites with their low transverse tensile strength can be prone to this type of effect.
Under normal conditions, material constituents in a structure are bound to their respective potential levels with relative stability. As impact energy is applied, shock waves (impulses) may cause damage such as fiber breakage or cracks (that can propagate over time). When a defect is induced, the original energy distribution would be affected, hence, new energy levels and pockets of energy sub-levels will be formed. This energy re-mapping can be correlated to the applied force of impact and classified through the developed classification system (Mouritz et al., 2009;Li et al., 2009;Breitzman et al., 2009;Kim et al., 2009;Hayman, 2007;Stoika et al., 2009) Predictive and classification analysis is an important tool. In developing such system (Zangani et al., 2007;Williams et al., 2008;Zhang and Richardson, 2004;Goebel et al., 2006a;2006b;Hu et al., 2006;Eklund and Goebel, 2005;Verdegaya et al., 2008;Chinnam and Baruah, 2007;Jenab and Rashidi, 2009), certain things have to be considered to enable criteria for classification and ranking for tested structures: In this study a novel classification algorithm based on searching an image for pixel re-distribution is used for damage classification. The proposed technique is suitable for high volume monitoring and inspection of safety critical components non-destructively. Figure 1a represents a model of a damaged composite sample. The Fig. 1 illustrates directions of wave travels through the tested sample once exposed to a testing source. Such signals are captured and analyzed by the developed ICA algorithm below.

MATERIALS AND METHODS
Integrity Classification Algorithm (ICA): This novel approach to defect detection is based on wavelength and color intermixing as illustrated in the following steps: • Before format conversion, the captured image is allocated a map as shown in Fig. 1b • The primary colors for each obtained image are intermixed according to and the results are stored and then converted to data using a purposely developed search through algorithm • The conversion of a searched through image will result in a text file that holds the necessary information regarding the state of the tested component • All resulted data files per image are then divided into strings of data each as a sequence stored in a specific file using our developed Matrix-Column Algorithm (MCA) before being correlated to produce a decision regarding component health In the MCA algorithm, the converted image is filtered into sequences S 1 -S m containing vectors of individual column matrices extracted from the converted source image data file. The overall extracted matrix consists of discrete combination of all column sequences as in Eq. 1: Where: a ij = Original matrix elements r ij = Amplitude factor From (1) we obtain: From (2) we obtain: where, θ is a normalizing factor. As the original matrix is simplified in (3), the data classification column matrix is represented in Eq. 4: The designed classification algorithm takes into account reference, undamaged sample images in its operations to achieve a decision regarding structure integrity as a function of required application. Considering Eq. 4, the required decision function can be derived as follows.
For a reference sample image, (4) becomes: With a damaged sample image given by: For damaged/undamaged decision and using (5) and (6) we obtain: The resulting values are substituted in a predecision matrix P: For a perfectly undamaged composite structure, P = 0. For damaged components, each element in P contributes by its values to the overall classification of level of damage. Using Eq. 3 and 8, we obtain: For practical applications, each element in Eq. 9 should not exceed a certain value β for the component to be acceptable. Hence, the final decision per testing technique is based on matrix F given by: For an over all correlated decision, a hybrid matrix is used as in Eq.11:  Figure 2 and 3 show C-Scan images obtained for 5 mm Woven Glass before and after an impact at 28.6 J, while Figure 4 and 5 show Low Temperature Thermal imaging (LTT) for the same component. Table 1 and 2 show the MCA algorithm results for C-Scan and LTT images, while Table 3 and 4 show same results as in Table 1 and 2 but re-grouped to show the MCA results before impact and after impact for both C-Scan and LTT.    Table 1-4 and Fig. 6-9, the following is realized:

RESULTS
• The difference in pixel occupation between pre and post damaged images • The difference in pixel concentration between C-Scan and LTT images for pre and post impact damaged components • The ability to construct a full spectrum of sequences that each set [S 1 ,…S m ] represents a type of image resulted from a different NDT testing technique and as some techniques are better in detecting certain defects than others, Intelligent algorithms can be used to predict the response using a technique from the response of others • It is easily noticeable from applying the above expressions that C-Scan Complements LTT and can be used as a second confirmation to the state of the tested structure      0  7  26  37  0  0  47  8  0  0  4  21  16  14  0  0  7  1  0  0  0  0  0  2831  0  0  27  0  0  0  0  0  1  0  0  0  0  2  0  0  19  623  2  32  2  2  122  4  • Figure 6 show the extent of damage occurred by 28.6 J to the 5 mm thick woven glass component when tested by C-Scan and LTT. The LTT technique and image contained better information regarding the severity of damage compared to the C-Scan one • The developed classification algorithm provided clear and accurate decision regarding component usability performance. Also, the known difficulty in the interpretation of the damage data is solved through a specialized analysis and interpretation algorithm (MCA) specifically developed to indicate the severity of the damage and its effect on the general performance of the component. This very successful approach to damage detection and classification is further supported by its ability to correlate different NDT technologies and predict others.