Page 280 - Aircraft Stuctures for Engineering Student
P. 280

8.7 Fatigue  261





















               Fig. 8.1 7  Goodman diagram.
               The probability that a specimen will endure more than N cycles is then 1 - P(N). The
               normal distribution  allows negative values  of N, which is clearly impossible in  a
               fatigue testing situation. Other distributions, extreme  value distributions, are more
               realistic and allow the existence of minimum fatigue endurances and fatigue limits.
                 The damaging portion of a fluctuating load cycle occurs when the stress is tensile;
               this causes cracks to open and grow. Therefore, if  a steady tensile stress is super-
               imposed on a cyclic stress the maximum tensile stress during the cycle will be increased
               and the number of cycles to failure will decrease. Conversely, if the steady stress is
               compressive the maximum tensile stress will decrease and the number of cycles to
               failure will  increase. An  approximate method  of  assessing the  effect  of  a  steady
               mean value  of  stress is provided by  a  Goodman diagram, as shown in Fig. 8.17.
               This shows the cyclic stress amplitudes which can be superimposed upon different
               mean stress levels to give a constant fatigue life. In Fig. 8.17, Sa is the allowable
               stress amplitude, Saq0 is the stress amplitude required to produce fatigue failure at N
               cycles with zero mean stress, S,  is the mean stress and S,, the ultimate tensile stress.
               If S,  = S,, any cyclic stress will cause failure, while if S,  = 0 the allowable stress
               amplitude is Sa.o. The equation of the straight line portion of the diagram is

                                                                                 (8.36)


               Experimental evidence suggests a  non-linear relationship for  particular materials.
               Equation (8.36) then becomes

                                                                                  (8.37)

              in which in lies between 0.6 and 2.
                 In practical situations, fatigue is not caused by a large number of identical stress
              cycles but by many different stress amplitude cycles. The prediction of the number
               of cycles to failure therefore becomes complex. Miner and Palmgren have proposed
               a linear cumulative damage law as follows. If N  cycles of stress amplitude Sa cause
               fatigue failure then 1 cycle produces 1/N of the total damage to cause failure. There-
               fore, if r different cycles are applied in which a stress amplitude Si (j = 1: 2, . . . , r)
   275   276   277   278   279   280   281   282   283   284   285