Page 121 - Mathematical Techniques of Fractional Order Systems
P. 121

Exact Solution of Linear Fractional Distributed Order Systems Chapter | 4  109


             4.3  MAIN RESULTS
             The cornerstone of the analytical solution of (4.17) is in fact the impulse
             response of the integrator with the same weight function as in the differen-
             tiator involved in this equation. This is similar to the case of fractional
             order and integer order systems. Recall that the analytical solution of the
                                                         c  α
             fractional order system of linear differential equations D xðtÞ 5 AxðtÞ 1 BuðtÞ
                                                         0  t
                                                            α
             is written by using Mittag Leffler functions E α;β tðÞ 5  P 1N  t αk  .
                                                                  k50 Γαk 1 βð  Þ
             Mittag Leffler functions are constructed on the basis of iterated self-
             convolutions of fractional power function, i.e., t  α21 =ΓαðÞ which is the
             impulse response of fractional integration operator (4.7). A general form of
             the analytical solution is obtained in a similar way for distributed order
             equations. This idea is in fact analogous to the analytic approach to integral
             equations using resolvent kernels. Using this concept here makes sense due
             to the fact that it is always possible to turn Eq. (4.17) into the following two
             Volterra integral equations of the second kind



                              x 1 2 Ax 1 i 5 x 0ðÞ
                                            ;  xtðÞ 5 x 1 tðÞ 1 x 2 tðÞ  ð4:19Þ
                             x 2 2 Ax 2 i 5 Bu i


             in which itðÞ is defined in (4.16). In order to show this, let us take the
             Laplace transform of (4.19) which yields

                                 X 1 ðsÞ 2 AX 1 ðsÞIðsÞ 5 x 0ðÞ=s
                                                                       ð4:20Þ
                                 X 2 ðsÞ 2 AX 2 ðsÞIðsÞ 5 BUðsÞIðsÞ
                Dividing each relation by IsðÞ and replacing W 2lnsÞ 5 1=IsðÞ afterwards
                                                      ð
             results

                                                       ð
                            X 1 ðsÞW 2lnsð  Þ 2 AX 1 ðsÞ 5 x 0ðÞW 2lnsÞ=s
                                                                       ð4:21Þ
                            X 2 ðsÞW 2lnsð  Þ 2 AX 2 ðsÞ 5 BUðsÞ
                Solving the resultant equations for X 1 ðsÞ and X 2 ðsÞ gives
                                               21
                             X 1 ðsÞ 5 W 2lnsðð  ÞI2AÞ x 0ðÞW 2lnsÞ=s
                                                      ð
                                               21                      ð4:22Þ
                             X 2 ðsÞ 5 W 2lnsðð  ÞI2AÞ BUðsÞ
                In which I is the identity matrix with appropriate dimensions. The solu-
             tion of (4.17) is then obtained by adding the homogeneous and particular
             solutions XsðÞ 5 X 1 sðÞ 1 X 2 sðÞ as
                                          21
                                ð
                              ð
                                                 ð
                        XsðÞ 5 W 2lnsÞI2AÞ  x 0ðÞW 2lnsÞ=s 1 BU sðÞ    ð4:23Þ
                Multiplying each side of (4.23) by W 2lnsÞI 2 A yields
                                              ð
                            ð
                         ð W 2lnsÞI 2 AÞXsðÞ 5 x 0ðÞW 2lnsÞ=s 1 BU sðÞ  ð4:24Þ
                                                 ð
   116   117   118   119   120   121   122   123   124   125   126