System evaluation from impulse response fragments
Unknown · 2013
Read System evaluation from impulse response fragments by Unknown (2013) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/nasa-tm-19720011008-system-evaluation-from-impulse-response.txt.
Introduction
J K Baumeister and E H Hopper of the Marshall Space Flight Center wrote this technical note in March 1972, as NASA TM X-2521. It solves a common problem in systems analysis. Engineers often know a system only by its sampled impulse response. That list of numbers is awkward for further design work. What they really need is the transfer function, a compact formula for the system's behavior.
The authors found a procedure based on rational approximation. It computes the transfer function from a limited number of response values. The system's order does not need to be known in advance. Moving the result into the other common domain is also easy. Short, incomplete response records can still yield a usable model.
Catalog record
| CREATOR | Unknown |
|---|---|
| YEAR | 2013 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| LICENSE | Public Domain |
| RIGHTS | PUBLIC DOMAIN |
| SOURCE | NASA NTRS |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
System evaluation from impulse response fragments NASA Technical Memorandum N A S A TECHNICAL NASA TM X-2521 MEMORANDUM to SYSTEM EVALUATION FROM IMPULSE RESPONSE FRAGMENTS by J. K. Baumeister and E. H. Hopper % George C. Marshall Space Flight Center Marshall Space Flight Center, Ala. 35812 NATIONAL AERONAUTICS AND SPACE ADMINISTRATION • WASHINGTON, D. C. • MARCH 1972 T E C H N I C A L REPORT S T A N D A R D T I T L E PAGE . REPORT NO. 2. GOVERNMENT ACCESSION NO. 3. RECIPIENT'S CATALOG NO. NASA TM X-2521 4. TITLE AND SUBTITLE 5. REPORT DATE Jystem Evaluation from Impulse Response Fragments March 1972 6. PERFORMING ORGANIZATION CODE 7. AUTHOR (S) 8. PERFORMING ORGANIZATION REPORT tt . K. Baumeister* and E. H. Hopper M-538 9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. WORK UNIT NO. eorge C. Marshall Space Flight Center 70-103-19-04 11. CONTRACT OR GRANT NO. Marshall Space Flight Center, Alabama 35812 13. TYPE OF REPORT & PERIOD COVERED 12. SPONSORING AGENCY N A M E AND ADDRESS National Aeronautics and Space Administration Technical Note Washington, D. C. 20546 14. SPONSORING AGENCY CODE TASK DESCRIPTION By use of some identification or estimation procedure based on statistical theory, up-to-then totally unknown systems have to be identified by its impulse response. The impulse response is represented by a time series equally sampled by a sampling interval T, which is known from the identification procedure (Fig. 1). As a shortcoming, the impulse response is known only over a limited time interval. The available data must be converted to a form which describes the vehicle dynamic completely in the sampling moments and which \ can be used for further data pro- \ cessing. THE RATIONAL 1 APPROXIMATION 0 T 2T 4T 6T 8T 10T t The z-transform of the time series x(kT) is defined to be Figure 1. Time series represen- tation of impulse response. G(z) = c +c .z~1 + ... + c, z -k +. 01 k (1) where c = x(kT) with sampling interval T. K. The desired equivalent rational expression in z is of the form -1 -m a + a.z + ... + am z 0 1 m F(z) = (2) b. + b z~ + ' . . . + b z~ 0 1 n -1 If (1) is any series in z ", there exists a unique rational "approxima- tion" of the form (2) for each pair of integers m and n that agrees when ex- panded term by term with the series for more terms than any other rational expression with smaller or equal m and n [2]. It follows that, if G ( z ) has a rational expression, as is the case with most engineering systems, the exact rational expression should be found when m and n are chosen large enough. The choice of m and n does create a problem which will be discussed later. Once m and n are chosen, the rational approximation is found in multi- plying G ( z ) by the denominator of F ( z ) and collecting the like powers of z~*: -1 -2 -k c +c z +c z + ... + c z + ... 1. -1 + , -2+ + vv v - - - v -n Vo+ Vi2"1+ boV~2 + • • • (3) ,b -1 +b, c z -2+ b, -3 , -(k+1) iV i i iV + - - - + V k z The first m+1 powers of z can be forced to agree with the numerator of F ( z ) , and the next n can be forced to vanish. The equations for this are \= L c b ii . k - m ; (4) i+j=k thus, a = o Vo (5) and b.c. = 0 1 L i+j=k , k = m+l, m+2, ... m+n (6) with n thus, b.c + b.c + ... + b ,cn = 0 0 m+1 1m m+1 0 b (? + b.c + ... + b c =0 0 m+2 1 m+1 m+2 0 be + b.c + ... + b C A = 0 (7) On 1 n-1 n0 be f + be + ... + b c, = 0 0 n+1 In n 1 bnc + b.c + ... b c =0 0 n+m 1 n+m-1 nm The last equations are solved for b., b,, ... b and then substituted in the 01 n first m+1 to obtain a , a , ... a . Clearly, one of the b. can be chosen arbitrarily (anything except 0), and the rest are unique if the system of equa- tions is not singular. If m and n are chosen too large, the system of equations will be singular. If the rank of the system is r, the denominator of F ( z ) should be of degree r in z .If the system is solved using n too large, theoretically the numerator and denominator of F ( z ) will have common factors which could be factored out. In practice, these factors may not be exactly the same be- cause of the roundoff errors. Some procedure for checking whether the value of n is too large should be used to prevent the introduction of zeros in the denominator and also to prevent the solution of a singular system of equations. SCALING Assuming that the time sequence x(nAt) has been received from the time function x ( t ) through an ideal sampler, x(nAt) is attenuated by a factor— [3]. For correct scaling, the numerator of f(z) must be multiplied Z\ v by At. RECONVERSION TO THE TIME DOMAIN For later use in the realization of the procedure described previously, it is presumed [4] that F(z) is reconverted to the time domain by y k = Vk + ai\-T + * •' + amXk-mT ' Vk-T ' '' 'Vk-mT (8) where x and y are the input and output time sequences of the digital filter, K K. and x - mT and y - mT are the m- T earlier values of the input and output K & time sequences. Applying an impulse as an input to equation (8), the impulse response of F ( z ) can be calculated as a time series representation. IMPLEMENTATION OF THE PROCEDURE IN A DIGITAL COMPUTER For implementation of the method just described, it has to be realized that, according to the task description, the vehicle dynamics are completely unknown; thus, information on the order of the rational expression in both numerator and denominator's not available. Consequently, a scheme must be established to evaluate m and n. A suitable scheme has been found by taking logical steps as follows: 1. The order of the numerator and denominator of equation (2) is assumed to be equal; thus m = n. As later shown, this assumption does not introduce a significant error. 2. An initial estimate on m is made assuring that m . x s m . ,. estimate actual 3. With this estimate in m, the procedure previously described will be executed. If the estimate was too high, the matrix which has to-be solved in the process