Vous êtes sur la page 1sur 5

PROJECT ASSIGNMENT 2011 The report on the assignment/project must be submitted by Thursday the 1st of December 2011.

After this date you will be given a new assignment which must be submitted within one month. Solutions can be submitted electronically by e-mail to hansbod@kth.se. It is allowed for two persons to co-operate and submit a joint solution. In order to pass you will have to book an appointment for a short discussion of your report during week 49 (5-8 December). You will not get your corrected assignment back before the 20th of December but you will be notified about changes you would have to make to submit a corrected version. It is therefore better to submit a solution by the deadline even if it is not perfect than to not submit a solution at all. Your report must be written in English and should contain the following sections: Introduction (a few words about the problem studied) Test setup (describe the experiments made and the equipment used) Theory and methodology (describe the calculations made, both equations and Matlab implementation) Experimental results and discussion (present and discuss the results based on the questions) Conclusions (summarize your conclusions)

PROJECT ASSIGNMENT FOURIER ANALYSIS AND ESTIMATION OF FREQUENCY RESPONSE FUNCTIONS USING RANDOM SIGNALS, SYSTEM IDENTIFICATION

Introduction
Measurements were made of the force input to a Plexiglas plate excited by shaker and the acceleration response at six different positions on the plate. This plate is used to provide standardized measurements for experimental modal analysis.

Test setup
Measurement object: Plexiglas plate Instrumentation: (1) Force transducer: Dytran 1051V3 (2) Accelerometers (8): B&K 4507 B 005 (3) Preamplifier 8-chanel PCP854 (4) Data acquisition system VXI Measurement setup: (1) Chanel 1: Force excitation in corner, point 5 (2) Acceleration measurements in the points, Chanel 2: point 6 Chanel 3: point 22 Chanel 4: point 23 Chanel 5 (broken): point 13 Chanel 6: point 18 Chanel 7: point 1 Chanel 8 (unknown point) Excitation: White noise Data aquisition parameters: Sampling frequency 4000 Hz Calibration factor: From a separate calibration measurement it has been estimated that there is a calibreration factor 1.3 for the ratio Force/acceleration.

Practical information
In this problem a data file will be used. Each group will get a data file, in Matlab format which can be downloaded from Bilda, to work on. There will be a list available at the computer lab session on Monday the 14:th of November where your group will be assigned a certain data file. The file contains a matrix with three columns: time (column1), force (column2), acceleration (column 3).

Problem formulation
a) Estimate power spectral density (PSD), cross spectrum and frequencies for the force and acceleration signals using the pwelch and cpsd commands in Matlab. First estimate the power spectral densities of he signals using pwelch [Sxx,f]=pwelch(x,WINDOW, NOVERLAP, NFFT,fs); where x is the time series to be analysed, WINDOW should be a Hanning window (WINDOW = hanning(NFFT, periodic) applied in this case and should have the length NFFT, NOVERLAP is the number of overlap points. In this case it should be set to half of NFFT to get a 50% overlap appropriate for random noise excitation. Vary the number of averages by changing the block size (NFFT = 4096 and 65536). Then estimate the cross spectra S12 and S21 using CPSD. Plot the resulting spectral densities against frequency, for instance using a logarithmic amplitude scale. Answer the following questions: What is the effect of the averaging on the resulting power spectral densities? How many averages do we get when we choose NFFT as 4096 or 65536? b) Calculate the coherence function from the power spectral densities and the cross spectrum from a). Plot the result and comment on the coherence result when comparing it to the power spectral densities from a). Pay special attention to the location of minima in the two plots and try to relate minima in the coherence to minima in the power spectral densities. Also comment on the effect of averaging on the coherence function results.

c) Calculate the power spectral densities and cross spectra directly by using the Matlab FFT command and compare your result to the results obtained in a). You will have to write a loop in Matlab to perform the averaging according to Welchs technique. You should also in this case multiply the time signals with a Hanning window before performing the FFT. Use Parsevals relation to make sure that you get the correct scaling. d) Estimate the frequency response function from the power spectral densities and the cross spectrum from a). Calculate both the H1 and H2 estimates, see chapter 8, and plot the resulting magnitudes (absolute values) and phases and comment on any difference between the H1 and H2 estimates. Comment on the difference between the H1 and H2 estimates and also compare location of frequency ranges with differences between H1 and H2 with the location in coherence function minima from b). Hint: To get the magnitude of X use abs(X) and to get the phase use angle(X) or phase(X).

e) Estimate the transfer function H(s) using the technique described in chapter 6 Exercise 6.1. This involves using the Matlab command: [B,A] = invfreqs(H1,2*pi*f,nb,na); where nb and na are the length of B and A vectors:
H ( s) = s nb 1 B1 + s nb 2 B2 + ..... + sBnb 1 + Bnb s na 1 A1 + s na 2 A2 + ..... + sAnb 1 + Anb

Test with a number of different values for na and nb. Check the result by calculating the frequency response function for the model: H=freqs(B,A,2*pi*f); Suggested values to test are: na = 4, nb = 3 na = 8, nb = 6 na = 16, nb = 14 na = 32, nb = 26 Check the agreement by plotting the amplitude: plot(f,abs(H1),f,abs(H))

Which system size na+nb gives a sufficient agreement? Calculate the poles and zeros of one of your estimated models using the command: [z,p,k] = tf2zp(B,A) Plot the position of the zeros and poles in the complex plane using: plot(z,'o') plot(p,'x') Comment on the result. Can you say something about system stability?

Vous aimerez peut-être aussi