ABSTRACT

The fatigue design of tubular joints increasingly uses a single S-N curve based on the measured 'hot spot' stress range. Statistical analysis of the measured life delivers a corresponding design fatigue life with appropriately low probability of in-service failure. If a rigorous approach to probabilities is followed it is apparent that design should also address the reliability of predicting the 'hot spot' stress range for a given load and joint type. Finite element and acrylic model studies and many of the parametric formulae lack realism of the weld detail and cannot adequately cover the effects of weld profile variability and tubular tolerances.

This paper presents the previously unpublished results of tests to determine Stress Concentration Factors for axial, in-plane and out-of-plane moment loading on over 30 steel tubular joints of realistic size and construction. It is apparent from the steel model tests that there is scatter in measured stress Concentration Factors for nominally identical joints. These results are combined with the existing database of comparable tests and used to, develop the concept of a design SCF defined in this paper. A detailed justification of the design formulae for SCFs is given. This new approach to SCFs incorporates the inherent variability of tubular joints in a manner which is consistent with the rigorous approach now being adopted for fatigue life assessment. This leads to a clearer understanding of the apparent scatter in observed performance of tubular joints and reduces the risk of gross underestimation of SCFs which is possible using current techniques.

INTRODUCTION

The last five years has seen a significant research effort and corresponding increase in knowledge on the effects of fatigue at the joints of tubular steel jacket structures. The major change in design approach has been the introduction of S-N curves based on the 'hot spot' stress range concept in place of S-N curves based on the nominal brace or 'punching shear' stress.

The latest design S-N curves are based on measured 'hot spot' stress range as found from extrapolation to the weld toe of the maximum principle stresses obtained at some distance away from the weld toe. Statistical analyses of the measured lives are conducted to derive mean and design (either lower bound or characteristic) equations. From the overall reliability analysis point of view the response of the component should be viewed in two parts:-

  • The probability for a measured 'hot spot' stress range that the life will exceed the specified design value

  • The probability that the stress Concentration Factor (SCF) used in calculating the 'hot spot' stress range for a defined input loading will underestimate the actual 'hot spot' stress range.

The second part of this requirement is the subject of this paper. The proposed new approach to SCFs incorporates the inherent variability of tubular joints and the observed scatter in measured SCFs for nominally identical specimens.

This content is only available via PDF.
You can access this article if you purchase or spend a download.