Application of Bayesian analysis to determining fatigue load on composite material structures
| Autor: | Ito, Seiichi, Sugimoto, Sunao, Aoki, Yuichiro, Okada, Takao |
|---|---|
| Jazyk: | japonština |
| Rok vydání: | 2013 |
| Předmět: | |
| Zdroj: | 宇宙航空研究開発機構研究開発資料 = JAXA Research and Development Memorandum. :1-27 |
| ISSN: | 1349-1121 |
| Popis: | 実機規模の複合材構造の疲労寿命実証試験では、荷重増分係数(LEF: Load Enhancement Factor)に基づく加速試験が適用されている。LEF は静強度ならびに疲労寿命の変動に対する双方のワイブル確率分布モデルの形状母数で定義され、複合材強度データベースを基本としてLEF の値が設定される(CMH-17: Composite Material Handbook-17)。LEF 評価の基本データベースは広範囲にわたる試験条件から構成されるため、静強度ならびに疲労寿命の形状母数の不確定性は大きい。これに対してCMH-17 のLEF 設定では形状母数のモード値が用いられている。しかし形状母数の変動が大きいことから、本報告では形状母数の確率分布を仮定してこれらの母数をベイズの方法で推定し、LEF のベイズ的期待値を求める。一方、形状母数の分布が複雑である場合を想定し、スプライン関数を適用したノンパラメトリック確率分布を仮定してLEF の期待値を得る手法も記す。これらの確率モデルを基本として、CMH-17 のLEF 推奨値と、本報告のパラメトリックならびにノンパラメトリック評価のLEF 期待値との比較検討を行う。 The load enhancement factor (LEF) approach is proposed as a fatigue acceleration test to demonstrate the fatigue strength of composite material structures. The objective of the LEF approach is to increase the applied loads in the fatigue spectrum so that the same level of reliability can be achieved with short test durations. Some fatigue tests on the full-scale composite structure are executed on the basis of this method. In the LEF approach, the probability distribution of residual strength and fatigue life is assumed to be a two-parameter Weibull distribution. LEF is defined as a function of shape parameters from the Weibull distributions. As a consequence, the uncertainty in residual strength and fatigue life greatly affects LEF assessment. For simplicity, modal values of the shape parameters are used in the Composite Materials Handbook (CMH-17). This study investigates the uncertainty in the shape parameters in more detail. First, the shape parameters are modeled as a probability distribution, by using Bayesian analysis to evaluate the scatter of the shape parameter. As the so-called parametric method, the unknown parameters of the distribution are estimated using the Bayesian method from the database. The Bayesian expected value of the LEFs is calculated from the estimated probability distribution of the shape parameter. When the uncertainty of the shape parameter is large and the assumption of the usual probability distribution is problematic, the non-parametric estimation method for the distribution is applied. The probability distribution approximation that uses the spline function is used here. Last, this paper presents two types of expected LEF values mentioned above and their comparison with already reported results in CMH-17. 形態: カラー図版あり Physical characteristics: Original contains color illustrations 資料番号: AA0062046000 レポート番号: JAXA-RM-13-011 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|