Development of A Reliability-based Method for Remaining Life Evaluation Using Extremes Statistics Analysis of Short Fatigue Cracks

本研究旨在发展一种基于负债的剩余寿命评估方法。该方法利用极值统计和图像处理估计的最大裂纹长度。研究内容包括：构建高温疲劳试验系统（主要是微机控制系统和应变测量系统），对奥氏体不锈钢SUS316L在650℃空气中进行蠕变疲劳试验，引入图像处理技术，包括开发一些新算法，对大量短疲劳裂纹进行尺寸测量，并对其进行分析。基于恒应变循环间隔极值统计的最大裂纹长度估算及其与实测值的比较，以及基于可靠性的剩余寿命评估方法的发展。主要结果总结如下：首先，将图像处理成功地应用于大量短疲劳裂纹的尺寸测量，并应用极值统计。测量所消耗的时间和精力可以大大减少。在<正负bbb10s的相对误差范围内，可以测量出大于30 gm的裂纹长度，但发现了蠕变疲劳中最大裂纹长度估计的局限性。即在裂纹合并发生前，最大裂纹长度的估计值与实测值基本一致，而在裂纹合并发生后，最大裂纹长度的估计值与实测值存在偏差。这在基于最大裂纹长度的剩余寿命评估中是重要的。最后，提出了“修正回归期”的概念，并在此基础上提出了一种基于可靠性的剩余寿命评估方法。该方法可以预测低失效概率（或高可靠性）下的剩余寿命。

This research was conducted aiming at development of a liability-based method for remaining life evaluation. The method makes use of the maximum crack length estimated by means of statistics of extremes and image processing. The research involves construction of a high temperature fatigue testing system (mainly, a control system using a microcomputer and a strain measurement system), execution of creep-fatigue tests of the austenitic stainless steel SUS316L at 650 ﾟC in air, introduction of image processing technique including development of some new algorithms, size measurement of a large number of short fatigue cracks, estimation of the maximum crack length based on statistics of extremes at a constant interval of strain cycles and comparison with the measured value, and development of a reliability-based method for remaining life evaluation. The principal results are summarized as follows.First, image processing was successfully applied to size measurement of a large number of short fatigue cracks to which statistics of extremes was applied. The time and efforts consumed in measurement could be greatly reduced. It was possible to measure the crack length larger than 30 gm within the relative error of <plus-minus>10Second, a limitation of estimating the maximum crack length in creep-fatigue was found. That is, it was found that up to occurrence of crack coalescence, the estimated maximum crack length was in agreement with the measured one, whereas after occurrence of coalescence the former deviated from the latter. This is important in remaining life evaluation based on the maximum crack length.Finally, the new concept of "modified return period" was proposed , and a reliability-based method for remaining life evaluation was developed using this concept. This method enables to predict the remaining life corresponding to a low failure probability (or high reliability).