패턴인식 기반 초음파 신호 및 화상처리와 레이저 초음파법에 의한 신뢰성 평가

Abstract

The first purpose of this research is to study on stability estimation of plant structure through classification and recognition about welding flaw in SWP(Spiral Welding Pipe). And, In this research, we used nondestructive test based on ultrasonic test as inspection method, and made up inspection robot in order to control of ultrasonic probe on the SWP surface, and programmed to signal processing code and pattern classifying code by user made programming code. Inspection robot is simply constructed as 2-axes because of welding bead with fixed pitch. So, inspection of welding part can be possible as composition of inspection part for tacking on welding line. For evaluation of flaw signal is reflected on welding flaw, user-made program codes are composed of signal processing and Bayesian classifier and perceptron and back-propagation neural network. And then, it can be shown that superiority of neural network method compared with Bayesian classifier for classification and recognition rate was confirmed. According to this result, we selected back-propagation neural network as classification and recognition method about the system of SWP stability Estimation. Through this process, we proved efficiency on the system of SWP stability Estimation, and constructed on the base of the system of SWP stability Estimation for the application in industrial fields. The second purpose, field requirements is an evaluation process with objective and excellent performance so as to secure competitive power of a system in terms of processing speed, precision and reliability in the system for testing semiconductor parts. However, currently utilized equipment requires operator's skillfulness and a number of sampling tests, so that different results may be obtained depending on working conditions of the operator. Accordingly, availability of the defect test algorithm that recognizes exact and standardized defect information in order to fundamentally resolve generated defects in industrial sites by giving artificial intelligence to SAT(Scanning Acoustic Tomograph), which previously depended on operator’s decisions, to find various defect information in semiconductor package, to reduce personal errors and then to standardize the test process was verified in this study. In order to apply the algorithm to the lately emerging Neural Network theory, various weights were used to derive results for performance advancement plans of the defects test algorithm that promises excellent field applicability. The third purpose, Nd;YAG Laser (pulse type) was used to emit ultrasonic signals to a test material. In addition, a total ultrasonic investigation system was designed by adopting a Fabry-Perot interferometer which receives ultrasonic signals without any contact. For non-destructive test SM45C which contains some flaws was used as a test material. Because it is easy to align light beam in receiver, and the length of the light beam does not change much even if convex mirror leans towards one side, confocal Fabry-Perot interferometer, which has stable frequency, and PI control are used to correct interfered and unstable signals from temperature, fluctuation and time shift of laser frequency. Stable signals are always obtained by the feedback of PI circuit signals in the confocal Fabry-Perot interferometer. The type, size and position of flaws inside the test material were examined by achieving the stabilization of an interferometer. This study presented a useful method which could quantitatively investigate the fault of objects by using a Fabry-Perot interferometer.