Sensitivity Analysis of Inspection for Reliability-Based Composite Material Design Based on Human Factor

点击翻译按钮获取中文摘要

Recently, an evaluation method of detection probability has been proposed by the authors for the measurement and evaluation of the industrial products based on the human detectable ability. In the psychometric mesurement, distributed parameta of pychometric curve had been presumed using Probit method concerning the three kinds of atribute of size, flatness degree and grayscale. Moreover, in this thesis, we propose an efficient method for the experiment and evaluation by Staircase method for the practical uses, and the detection probability of Staircase method is compared with that of Probit method, taking an example of experiments in which the methods are applied to measure the perception stimulation of panels, or subjects, concerning the three atributes that are extracted from the defect images in the inspection of FRP (fiber reinforced plastics) poroduct. In conclusion, the proposed method will be overviewed its efficiency and proved for the practical estimation uses of the parameter … More of pychometric function. From these works, the following three points are conclusions.1.Staircase method, the proposed evaluation method, showed that the experiment efficiency could be improved in the comparison with Probit method. In the actual experiment, the proposed method could reduce the number of trials to 43% or less compared with Probit method. As a result, it has indicated that it is effective to reduce the burden of panel in defect detection work.2.As for the PSE (subjective, equivalent value) mean value by Staircase method, a close value to the result of parameter estimation by Probit method was obtained. As a result, the practical possibility as a threshold of automatic check has been proved to be as good as Probit method.3.When the proposal technique was applied to perception stimulation with the three attributes extracted from defect image of FRP product, the coefficient of variations of subjective, equivalent value (PSE) proved larger than the result by Probit method in any attributes. Therefore, this result has indicated that proper attention is to be paid when the presumption is made at the confidence interval using decentralization and standard deviation. Less