STATISTICAL ANALYSIS OF THERMAL NONDESTRUCTIVE TESTING DATA

Abstract

The features of processing of active thermal nondestructive testing results are considered. Proved the necessity of search and introduction of new informative parameters in evaluation of thermograms in order to improve the reliability of control. Task of detecting and estimating the relationships between defect parameters and optimal testing time and the maximum value of temperature signal is set. Computer simulation of active thermal testing of two samples with artificial defects with known characteristics was performed. Also obtained the sequences of thermograms and formed the sets of initial data during simulation for correlation, regression and dispersion analysis of testing results. The method of dynamic thermal tomography was used to determine the levels of maximum differential temperature signal and optimal testing time. The estimates of correlation coefficient for various informational parameters of thermal control obtained. There is a high level of relations between the optimal control time and depth of defects. A high correlation also observed between the maximum value of temperature signal and depth of defects. The nature of relationships between various informative parameters of active thermal control established by the regression analysis. A one-factor dispersion analysis of the influence of defect parameters on optimal testing time and maximum value of the temperature signal was performed. High degree of mutual influence of all informative parameters is established. The conclusion made on the necessity of developing new modern methods for analysis the data of thermal testing. Revealed patterns in relationships between data show low efficiency of traditional statistical methods in tasks of active thermal testing. Alternatively, proposed to use the artificial intelligence technologies, in particular, neural networks.