Page 473 - Marine Structural Design
P. 473

Chapter 25 Reliability of Ship Structures                             449


                       g(xlt) = CUMU 0) - c, VSM, (4 + M,(Ol                         (25.23)
                  where M,,(f) is the ultimate strength, M&) and Mdt) are the  stillwater and  wave-induced
                  moments respectively; C,, and C, represent the model errors in predicting the hull's  ultimate
                  strength and combined total bending moment the ship experiences. The failure probability at
                  the time Tis expressed by
                             'I g(J+)<O  1

                       pf(T) = 5  5.- If. (xlt)dx fT(t)dt                            (25.24)
                              0
                  where fx (xlt) is the joint probability density function; fT (t) is the probability density function
                  of  occurrence time  T,  which  is  assumed as a  uniform  distribution, fT (t) = I/T . Therefore,
                  Eq.(25.24) can be rewritten by

                                                                                     (25.25)


                  By defining P,(t)  as a conditional failure probability at time 2,

                      P, 0) =  J.. Jfx (xlw                                          (25.26)
                            g(xlf)<O
                  the failure probability has in a simple form


                                                                                     (25.27)

                  The response surface method (Bucher, 1990) is applied when the limit state function g(r1t)is
                  expressed  implicitly  and  has  a  nonlinear  form  in  order  to  overcome  the  expensive
                  computational effort integrating Eq. (25.25) in evaluating the failure probability.
                  The basic  concept of the  response surface method  is to approximate the  original complex
                  and/or implicit limit state function using a simple and explicit function. The accuracy of the
                  results depends highly on how  accurately the characteristics of  the original limit state are
                 represented by  approximate function. The suitability of the response surface obtained relies
                 mainly  on  the  proper  location of  so-called  sampling points.  Many  algorithms have  been
                 proposed to select appropriate sampling points, which promise to yield better response surface
                 fitting. In addition, the basic function shape is  also known to be another major factor that
                 influences both the accuracy of the response surface method and the selection of the reliability
                 evaluation method.
                 Many practical reliability evaluation techniques are available once the failure surface G(x)  is
                 defined in an explicit closed form.  Among those techniques, first order method is commonly
                 used due to the efficiency with acceptable accuracy. An equivalent linearized limit state at so-
                 called design point is taken and the safety margin of the structural system is determined as the
                 minimum distance from the origin to the original nonlinear limit surface in the independent
                 standard normal space.
   468   469   470   471   472   473   474   475   476   477   478