Search

Computer aided Interpretation of Ultrasonic signals depicting flaws in weld pieces is depicted in this work. In this work, feasibility of Long Short Term Memory (LSTM) for flaw characterization is studied. Owing to the advantage of LSTM, the first technique involves training LSTM directly with the signals as inputs and testing its ability to characterize the flaws from the input signals. Due to wide variation in the length of input sequences, which introduced sparseness in other sequences, overall accuracy is affected. Hence in the second technique, LSTM are trained with features of the signals and it is found that the overall accuracy for test data is 67.64%. These features are statistical parameters obtained from the approximation co-efficient of the input signals. The input signals are decomposed with a novel wavelet template.

Industrial Radiography is the most widely used NDT technique for weld defect characterization. As it is an indirect technique, an expert radiologist is required at the interpretation end to identify and characterize the defect in radiographs. However radiographs suffer from poor contrast thereby making it difficult to isolate the defect region from the background. Hence it is necessary to develop computer aided analysis tool for automated interpretation of weld radiographs. It involves an efficient contrast enhancement and segmentation technique for isolating the weld defect to its true size. In this work, wavelet based contrast enhancement technique has been developed to highlight the defect region from background. Biorthogonal 6.8 wavelet and Daubechies wavelet of order 4 are used for enhancing root/slag and porosity respectively. Thresholding and region growing are used for segmenting the region of interest.

Radiographic testing is NDT technique is used for weld quality assessment in industries. Generation of x-rays, Number of x-rays sheet and amount of energy absorbed by specimen are all random processor and hence result in noise. Specially metal alides in x-ray sheet appear as grain when digitized also the size of grain increases while zooming in digitization hence grain noise is most predominant noise in radiograph. Also an artifacts degrades the quality of x-rays. Hence it is necessary to development an efficient de-noising technique for removing noise in radiograph. In this paper a novel de-noising technique by using spatial and spectral domain and optimization of wavelets is done. Daubechies wavelet 2 provides better de-noising than the other techniques.

Non-destructive testing (NDT) techniques are most commonly employed for detection and characterization of flaws in the component. Apart from flaw characteristics, another parameter which is equally important to assess the structural integrity of engineering components is the material property. Nondestructive testing techniques offer several advantages over the conventional coupon-based techniques. Ultrasonic Testing is one of the widely used NDT techniques for material characterization. In this paper flaw detection such as lack of fusion and slag of welding carbonized steel and stainless steel material is characterized using discrete wavelet transform (DWT).In order to categorize the type of flaws, the given ultrasonic signal is subjected into DWT decomposition. Then, energy is calculated from DWT approximation coefficient. Finally corresponding welding flaw is determined using computed energy.

Ultrasonic Testing is the widely used NDT technique for flaw detection in thick walled weldments. It is an indirect technique and the signals are to be analyzed in order to characterize the flaw. Manual interpretation of these signals is subjective in nature and is dependent on the expertise of the individual. Hence the paradigm has shifted to automated signal analysis. In this paper a successful attempt has been made to develop a pattern among the flaws of same type without using Artificial Neural Networks. Here, the signals are analyzed with Stockwell transform and the pattern is determined. Also quantitative characterization is done with mean, standard deviation, root mean square value, peak to rms ratio.

Epileptic seizure can be detected by many ways but EEG signal prove to be the most important marker. Since EEG signal requires a strenuous effort to go through pages of recorded signal. Automatic seizure detection can be done by extracting features from the EEG signals and then feeding them to the supervised learning algorithms for classification and prediction. In this paper the features that are chosen are mean, standard deviation, skewness, kurtosis, interquartile range and mean absolute deviation. A comparative study of SVM and GRNN are done in this work and GRNN proves to be accurate for seizure detection applications.