Page 438 - Cam Design Handbook
P. 438

THB13  9/19/03  7:56 PM  Page 426

          426                      CAM DESIGN HANDBOOK


                                T
                                   2 ˙˙
                            J =  W F +  W F 2 ˙˙˙
                               Ú 0  1  fe  2  fe
                                                                       (13.54)
                                             2
                                    Y() -
                              +  Y()[ t  Y ()] +  V () t  Y () t  dt
                                  t
                                           t
                                                    2
                                 i        d     2
          where weight functions V 1(t) and V 2(t) are normal distributions given by
                                          k  -  n
                                                  ,
                                    v t () =  e 2  i =12               (13.55)
                                    i
                                         S
                                          d
          with
                                          Ê t  - ˆ 1  2
                                       n = Á  ˜
                                          Ë  S ¯
                                             d
                                       k ª10  4
                                       S ª10  -5
                                        d
          and the function Y d (t) is a polynomial given by
                                                       1
                                    Y t () = 6 t - 5 t  6  0  £ t £ ¸
                                           5
                                    d
                                    Y t () = 6 S - 5 S 6  1  < t £  T Ô
                                           5
                                    d
                                                        Ô
                             where T >1                 ˝              (13.56)
                                                        Ô
                                        -
                                       T t
                                    S =                 Ô
                                       T -1             ˛
             The boundary conditions of the “interior point” at time t = 1.0 are now indirectly spec-
          ified through the introduction of weight functions V 1 (t) and V 2 (t) in the cost functional.
          To ensure convergence of the variation-of-extremals algorithm as well as reducing numer-
          ical errors in the integration, it is necessary to start the optimization with a relatively large
                                  -2
          standard  deviation  of  about  10 .  With  succeeding  iteration,  constant  S d is  gradually
          reduced to a point at which the weighting functions V i(t), i = 1,2 approximate a delta func-
          tion. These weighting functions force the boundary conditions at this interior point to be:
                                          Y 1 () = .
                                              1 0
                                          ˙
                                               0 0
                                     and  Y 1 () = ..                  (13.57)
             These boundary conditions are not expected to give a cam displacement, Y c(1), of unity
          at time t = 1 since the system equation (Eq. [13.30]) is of second order. If this is a concern,
                             2
          an additional term V 3(t)Y ¨ (t) may be introduced in the cost functional to specify the inte-
          rior boundary condition Y ¨ (1) = 0.0 indirectly. Weight function V 3 (t) is defined identically
          to V 1 (t) and V 2 (t).
             With  the  definition  of  T (see  symbols),  the  previously  used  normalizations  remain
          unchanged. This minimizes the amount of modification needed in problem formulation
          and computation.
             The formulation of the optimization problem for a tuned D-R-R-D cam can now be
          given as follows.
             Minimize
                                T
                                    2 ˙˙
                            J =  W F +  WF  2
                               Ú 0  1  fe  fe
                                                                       (13.58)
                                                     2 ˙
                                              2
                                            t
                               +  V ()[ t  Y ()] + V () t  Y () t  dt
                                  t
                                     Y() -
                                 1        d      2
   433   434   435   436   437   438   439   440   441   442   443