0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)
112 Ansichten6 Seiten
A force transducer is attached to the midpoint of a free-free beam and is used to measure the force on a rigid mass that vibrates with the beam. The ratio of force to mass times acceleration is measured for several different masses over a frequency range that includes the first four odd natural frequencies. A simple theoretical model is developed that explains the vertical axis shift in the calibration curve.
A force transducer is attached to the midpoint of a free-free beam and is used to measure the force on a rigid mass that vibrates with the beam. The ratio of force to mass times acceleration is measured for several different masses over a frequency range that includes the first four odd natural frequencies. A simple theoretical model is developed that explains the vertical axis shift in the calibration curve.
A force transducer is attached to the midpoint of a free-free beam and is used to measure the force on a rigid mass that vibrates with the beam. The ratio of force to mass times acceleration is measured for several different masses over a frequency range that includes the first four odd natural frequencies. A simple theoretical model is developed that explains the vertical axis shift in the calibration curve.
from the SAGE Social Science Collections. All Rights Reserved.
Base Strain Effects on Force Measurements
by P. Cappa and K. G. McConnell ABSTRACT -Force transducers directly interact with their environment. In this study, a force transducer is attached to the midpoint of a free-free beam and is used to measure the force on a rigid mass that vibrates with the beam. The ratio of force to mass times acceleration is measured for several different masses over a frequency range that in- cludes the first four odd natural frequencies of the beam. Then the force transducer is mounted so that the beam's strain is isolated from the transducer. The tests are repeated. The results dramatically illustrate the effects of base strain on the force measurements. A simple theoretical model is developed that explains the vertical axis shift in the cali- bration curve. KEY WORDS-force transducer, dynamic calibration, base strain, environmental effects Nomenclature a: acce1eration A: a constant [see eq. (9)] b: beam widtb B: a constant [see eq.(13)] BS: base strain test cases C: charge amplifier feedback capacitance d: deptb of beam Ea: acce1erometer output vo1tage Er. force transducer output vo1tage F: force El: beam's bending stiffness Kn: constant dependent on nth mode of vibrati an g: acce1eration due to gravity h: beam depth d/2 IS: iso1ated test cases (by using iso1ation a mass) 1: beam 1ength m: ca1ibration mass man: apparent seismic mass of nth mode mb: beam mass men: excess seismic mass for nth mode m5: transducer's seismic mass Nl: iso1ated test cases (by using iso1ation a mass and a nut) S: straight-line slope through data of voltage ratio vs calibration mass Sa: accelerometer's vo1tage sensitivity (V/g) Sr. force transducer's vo1tage sensitivity (V/unit force) Sq: force transducer's charge sensitivity (pC/unit force) Se: force transducer's base strain sensitivity Y o: v erti ca! axis intercept P. Cappa is Associate Professar (SEM member), Department of Meclumics and Aeronautics, University of Rome "La Sapienza," Via Eudossiana 18, 00184 Roma, Italy. K. G. McConnell is Professar (SEM Fellow), Depan- ment of Aerospace and Engineering Meclumics, Iowa State University, Ames, IA 50011, USA. Paper was presented at SEM 1994 Spring Conference and Exhibits. 284 Vol. 36, No. 3, September 1996 Yon: vertical intercept far nth mode 13n 1: nth eigenvalue mu1tip1ying factor E: base strain (!>n: second spatial derivative of nth mode shape at transducer's 1ocation <Pn: nth mode shape at transducer's 1ocation lntroduction A force transducer directly interacts with the environ- ment in which it is used. When a force transducer is mounted on a large steel block and impacted by another force transducer mounted on a hamrner, the ratio of the two signals in the frequency domain is a constant over a broad range of frequencies. This indicates that the calibrati an pro- cedure for impact hamrners is adequate only for operating frequencies well below the mounted natural frequency of either transducer. At higher operating frequencies, where stress-wave-propagation effects become important, the cali- bration data made with such procedure may exhibit con- siderable errar. McConnell 1 has shown three distinct types of environments that affect transducer performance. The three environments are (l) mounted on a rigid type foun- dation, (2) mounted on an impact hamrner, and (3) mounted to a structure that is driven by a vibration exciter. The transducer's metrological characteristics are influenced by these different environments. We can also calibrate a force transducer by mounting it onto a vibration exciter and measuring the input sinusoidal acceleration of the exciter head and the output sinusoidal force signal while using various masses, as outlined in Dally, Riley and McConnell. 2 A consistent calibration result is ob- tained over a broad range of frequencies using this method so long as we stay away from any transducer resonances. Such calibrations or tests as the two described briefly in the above paragraphs give a false sense of security re- garding the transducer's performance in other environ- ments. The objective of this paper is to create a test environment where the force transducer's base stran is a significant factor. This experiment is one of a series of ex- periments that are being conducted in order to identify the factors that affect overall dynamic testing. An earlier study by McConnell and Varoto 3 shows that reciproca! of the fre- quency response function (FRF) data cannot be obtained in the case of a free-free beam. The overall basic question is, What parameters and/or phenomena cause these problems? Theoretical Model of Test Environment The test environment is shown in Fig. l(a). The force transducer base is attached to an isolation mass, while the transducer's seismic end is attached to calibration mass. An (a) callbratlon mass A (b) Callbratlon mau m Fig. 1-Description of test setup along with typical calibra- tion results: (a) schematic of test set-up, (b) typical output as a function of calibration mass m accelerometer is attached to the free end of the calibration mass and its mass is included in the value of m. The iso- Iation mass is used to isolate the force transducer base from the beam's bending strain. The following is a brief description of the way the ex- periment is set up. The test method is similar to that de- scribed in Dally, Riley and McConnel1. 2 The force trans- ducer has a time-history output voltage given by: Be = Se F = Se (m + m.) a (l) where ms is the transducer's seismic mass and a is the ac- celeration of the calibration mass m. The force transducer's voltage sensitivity Se is related to its charge sensitivity Sq and the charge amplifier's feedback capacitance C by (see Dally, Riley and McConnell). 2 S- Sq c- C (2) The transducer's charge and voltage sensitivities should re- main constant for a wide range of frequencies. The accelerometer has a time-history output voltage that is given by E=S!. a a g (3) where s. is the voltage sensitivity. Solving for a in eq (3) and substituting into eq (l) we can sol ve for the output voltage ratio as TABLE 1-CALIBRATION MASSES Masses Used Name Symbol Mass (gms) MassA mA 245 MassB ma 243 Masse mc 244 Accelerometer maccel 20 Bolts mbon 1.5 ave Total Mass for Each Test Name Symbol Mass (gms) Mass 1 m, 268 Mass2 m2 512.5 Mass3 ma 758 (4) which gives a straight-line relationship between the vol- tage ratio and the calibration mass m. The slope of this line is S and is given by Se g S=-- s. (5) The value of S should be a constant value if we are sig- nificantly below the transducer's resonance. A plot of eq (4) is shown in Fig. l(b) where the abscissa is the cali- bration mass m. We see that the horizontal intercept of the plot is the negative of the seismic mass m.. The vertical intercept Y 0 is given by Y0 = S m. (6) Experimental Setup The experimental setup used a cold-rolled steel beam that was 93 in. (2362 mm) long with a rectangular cross section of 1.0 x 1.25 in. (25.4 x 31.75 mm). A Kistler model 912 force transducer was connected to a Kistler model 5040 dual mode charge amplifier that was set with a l 000 pF feedback capacitor and the standardization po- tentiometer position of unity so that eq (2) was valid. The manufacturer has no specific instructions for the appli- cation of this force transducer as far as environments are concemed. This force-transducer charge amplifier was con- nected to the A input channel of the Bruel and Kjaer dual- channel signal analyzer. A Kistler model 808A accelerometer was connected to a Kistler mode! 504 charge amplifier that was adjusted to bave a voltage sensitivity of 10.0 m V/g. This sensitivity was established by using back- to-back calibration with a Kistler mode! 808T acceleration calibration standard. The accelerometer's charge amplifier was connected to the B input channel of the Bruel and Kjaer 2032 dual-channel signal analyzer. Three different tests were conducted using this experi- mental setup. In the first test sequence, the force transducer was connected directly to the free-free beam so that beam bending was directly applied to the transducer base and is referred to as the bending strain (BS) configuration. Either Experimental Mechanics 285 force transducer N bea m E s Fig. 2-0rientation of force transducer one, two or three calibration masses were used. The indi- viduai masses and their sums are given in Table l for each test case. In the second test sequence, the force transducer was iso- lated from beam bending strains by an isolation body con- sisting of a steel disk that was 1.500 in. (38.1 mm) in diameter and 1.00 in. (25.4 mm) in height. This is referred to as the isolation case and is designated by the IS symbol. The three calibration masses were attached and tested in order as before. In the third test sequence, a nut was placed between the isolation mass and the beam in order to reduce the beam's bending moment effects even further. This is referred to as the nut isolation case and is designated by NI. The results showed no significant improvement over the second test se- quence data results. Consequently, little effort was ex- pended on this test. The force-transducer orientation relative to the beam is shown in Fig. 2. The beam was mounted horizontally in the east-west direction. The transducer-cable connector po- sitions are called north, south, east and west when aligned with these directions. The transducer is viewed head-on from the seismic side in order to define these directions. The seismic side is the end connected to the calibration masses in this experiment. Experimental Test Results The frequency analyzer was set in broadband measure- ment condition to cover a frequency range of O to 800 Hz with a resolution of 1.0 Hz while using rectangular a win- dow function for measuring transient vibrations. Since the force transducer, the accelerometer, the cali- bration mass and the isolation masses, in the IS and NI TABLE 2-FREE-FREE BEAM NATURAL FREQUENCIES (Hz) cases, were attached to the midpoint of the free-free beam, as shown in Fig. l(a), only the odd natural frequencies are present. The even natura} frequencies should not be pre- sent, because both excitation and instrument location are at an even frequency node point. However, due to some cross-axis sensitivity to either acceleration or bending mo- ments, the frequency spectra of each signal showed a weak resonant peak when compared to the odd natura} frequen- cies. Hence only odd frequency data are used as being rep- resentative of the phenomenon being measured. It should be observed that the peak acceleration responses at each frequency were nearly equal for each odd mode of vibra- tion, increasing slightly with frequency. The test procedure was straightforward for each calibra- tion mass used. First, the beam was impacted at its rnidpoint along the longitudinal symmetric axis of the transducers and the masses using a PCB model 086B09 electromag- netic impact hammer in order to produce a consistent im- pulsive input. The transient impulse response signals from the accelerometer and force transducers were recorded by the B & K analyzer, and their respective frequency spectra were averaged. This impact was repeated about five times in order to obtain a good estimate of the voltage ratio Er lE.. Second, the acceleration autospectral density was examined to determine the four odd harmonic frequen- cies f1, f3, etc. Third, the input-output frequency spectra ra- tio was read for each frequency f 1 , f 3 , etc., to obtain the voltage ratio EriE. as defined in eq (4). This experiment was repeated three or four times in order to obtain an av- erage set of data. Then either the calibration mass was changed, or the mounting method was changed (force transducer directly applied to the beam, case BS; force transducer isolated from the beam's bending strain by the isolation mass, case IS; force transducer isolated from the beam's bending strain by the isolation mass and a nut, case NI), or the direction of the force transducer mounting was changed. The experimental data was reduced using the statistica} and plotting capability of the MathCAD software. The force transducer was more sensitive to base strains when the connector was mounted in either the north or the south directions and less sensitive to base strains in the east and west directions. The north data is shown in Fig. 3 as plot of voltage ratio Er lE. vs the calibration mass m for the case where the transducer base was attached directly to the beam (case BS); the standard deviation values were always less than 0.5 percent. Figure 4 shows the same plot (north direction) for the case where the transducer base was iso- Mode Number Calcul. 1 Low Calcul. High Meas. 2 Low Meas. High ModeType 24.0 30.0 24.0 odd mode 2 66.3 82.9 even mode 3 130 163 125-128 163-168 odd mode 4 215 269 272-273 even mode 5 321 401 309-316
odd mode 6 449 562 even mode 7 597 747 575-585 738-747 odd mode 1. Natura! in bending about 1.0 in. (25.4 mm) dimension for low and 1.25 in. (31.75 mm) dimension for high. 2. Range measured smce dependent on masses attached. 286 Vol. 36, No. 3, September 1996
13 12 u iO g 8 7 6 5 4 3 2 0 o D -1st vlbretlon 111d1 3rd v lbretlon 111d1 -- 5th vlbretlon 111de 7th vlbretlon lOde 600 700 600 Mass (gli) Fig. 3-Voltage ratio vs calibration mass m for base strain BS case while orientated in north direction lated from the beam's bending strain by the isolation mass (case IS); the standard deviation values were always less than 0.1 percent. There is a significant difference between these two plots. According to eq (4}, these plots sbould be the same, but they are clearly different in that the vertical axis intercept Y 0 is significantly different. Now let us ex- amine the results more carefully. Table 3 sbows the measured slope in eacb case. The numbers must be multiplied by l x w- 2 and bave units of 1/gms. The slope varies from 1.040 to 1.205, a 15.9- percent variation. The average slope in tbe BS case is (1.085 0.0489) x w- 2 (1/gm). The average slope in the IS case is (1.055 0.0167) x 10- 2 (1/gm). The standard derivation in the BS case is about 4.5 percent while it is 1.58 percent for the IS cases. The slope in the BS case is 2.84 percent bigber than in the IS case. Thus it appears that base strains cause about three times more dispersion in the slope as well as approximately 3.0 percent bigber slope value on the average.
13 12 11 10 g 8 7 6 5 4 3 2 o o 100 200 300 - 1st vibretlon 111de 3rd vibretlon Mdi -- 5th vibretlon Ede 7th vibretlon Ede 400 500 600 700 800 Mass (gli) Fig. 4-Voltage ratio vs calibration mass m for isolated base strain IS case while orientated in north direction Table 4 sbows tbe effective cbarge sensitivity of the that is given with units of pC/kg. The cbarge sensttlVlty ranges from 47.2 to 54.6 pC/lb (104.19 to 120.53 pC/kg}, a variation of 15.6 percent. The average cbarge sensitivity in tbe BS case is 49.2 2.21 pC/lb (108.61 4.88 pC/kg), wbile it is 47.9 0.69 pC/Ib (105.74 1.52 pC/kg) for the IS case. The dispersion for the BS case is about 3.2 times that for the IS case. 1t is seen that the average cbarge sensitivity is about 2.7 percent larger than that obtained in the isolated case. This is a result that we would anticipate from eqs (2) and (4) and the re- sults in Table 3. The seismic mass can be estimated by using eq (6) with results tabulated in Table 5. These values are seen to bave a large range wben the anticipated value sbould be around 5 gms. However, a quick review of the average values for eacb test case sbows that the BS case is significantly af- fected compared to the IS case. These results suggest that base strain causes significant intercept cbanges. Now let us TABLE 3-MEASURED SLOPE (1/gms x 10- 2 ) VOLTAGE RATIO VS CALIBRATION MASS VIB North North South East East East W est W est Mode BS Case IS Case BSCase BS Case IS Case Nl Case BS Case ISCase 1st 1.205 1.098 1.094 1.157 1.081 1.069 1.152 1.052 3rd 1.101 1.058 1.070 1.088 1.055 1.057 1.019 1.055 5th 1.073 1.059 1.067 1.056 1.057 1.053 1.081 1.046 7th 1.041 1.044 1.050 1.045 1.044 1.044 1.060 1.040 Experimental Mechanics 287 TABLE 4-TRANSDUCER'S MEASURED CHARGE SENSITIVITY (pC/kg) VIB North North South East East East W est W est Mode BS Case ISCase BSCase BSCase IS Case Nl Case BS Case IS Case 1st 120.53 109.93 109.49 115.89 108.17 106.84 115.23 106.40 3d 110.15 105.96 107.06 108.83 105.52 105.74 101.99 105.52 5th 106.40 105.96 106.84 105.74 105.74 105.30 108.17 104.64 7th 104.19 104.42 105.08 104.64 104.42 104.42 105.96 104.19 see if we can develop a theoretical model to define the vari- ables that cause these changes. Theoretical Model to Explain Shifting lntercept Let us return to eq (l) and assume that the force trans- ducer voltage is given by (7) where Se is the strain sensitivity and E is the base strain. In addition, we can use the Euler model of a vibrating free- free beam, see Rao, 4 to show that the strain is given by w h ere (9) is a constant for a given beam and (lO) depends on the mode of vibration. In eqs (8}, (9) and (10}, h is half of the thickness in the direction of vibration, cp. is the second spatial derivative of the nth-mode shape where the transducer is attached, cp. is the nth mode shape deflection where the transducer is attached, IIlt. is the beam's mass, El is the beam's bending stiffness in the direction of vibration, and l) is the nth eigenvalue multi- plier. Equation (8) shows that the base strain is proportional to the beam's acceleration where the force transducer is attached. Now we can combine eq (8) with eq (7) to obtain the force signal that is directly proportional to the beam's ac- celeration. When this definition of Er is used in the voltage ratio in eq (4), we obtain TABLE 5---MEASURED SEISMIC MASS (gms) VIB North North South Ec (Scg) -= -- (m+m,) E. s. (Il) ( A Se g) + s. K. = S m + y On where the vertical intercept for the nth mode is given by (12} It is clear from eq (12) that if base strain sensitivity occurs in the force transducer while in this test environment, the vertical intercept will be different for each odd resonant frequency because K. is mode dependent. The values of cp,/cp K. and KlK. ratio are shown in Table 6 for the frrst four, odd, natural frequencies. The values of K. are calculated for the beam's midpoint. The ratio of cp,/cp. quickly takes on a unity value as we move up to third and higher odd modes. The ratio of KlK. shows that the base strain effects drop off rapidly with increasing odd modes. If we solve for the apparent seismic mass in eq (12) for each mode, we obtain [ A Se g] m .. = m, + S s. K. = m, + B K. (13) w h ere (14) so that the excess seismic mass for the nth mode becomes (15) We can estimate the ratio K 1 1K,. from eq (14) if we know the excess seismic mass for each mode of vibration. We know from the transducer's stiffness and bare natural-fre- quency specification that the seismic mass is about 5.0 gms. Then, if we average the north-south apparent seismic mass for each mode under the BS test condition and then subtract 5.0 gms, we obtain the north/south ratio given in Table 6. Repeating for the east/west data, we obtain the east/west ratio column in Table 6. East East East W est Mode BS Case IS Case BS Case BS Case IS Case Nl Case BSCase IS Case 1st 363 -o.031 471 238 -1.37 11.43 266 4.44 3rd 56.2 0.424 78.6 34.1 3.30 4.26 67.8 3.12 5th 30.4 0.0143 35.5 22.0 2.67 2.86 14.1 5.51 7th 24.9 0.0679 25.6 15.2 5.74 7.42 10.83 7.97 288 Vol. 36, No. 3, September 1996 TABLE 6-THE EFFECT OF STRAIN ON SEISMIC MASS MEASUREMENT VIB Mode Pnl <Pr/<Pn Kn x 10-3 K11Kn North East AVE South W est 1st 4.73 -0.763 34.1 1.0 1.0 1.0 1.0 3rd 11.00 -0.987 8.15 4.18 6.6 5.4 6.0 5th 17.28 -1.00 3.34 10.2 7th 23.57 -1.00 1.80 19.0 The north/south ratios are better than the east/west ratios. This is mostly due to the fact that the force transducer was significantly more sensitive to bending strains about the north/south axis compared to the east/west axis. Thus the measured apparent seismic masses were larger, and there is Iess relative error in the values reported in Table 5. In ad- dition, the accuracy of both the slope and intercept values would be increased if additional mass combinations (such as five) were used, compared to three that were utilized. Conclusions To show the dependance of force transducer output on base strain, the base of a mono-axial load cell was mounted at the midpoint of a free-free beam while the sensing (seismic) end was attached to one of several calibration masses; force transducer was impacted using an electro- magnetic impact hammer. The resulting data clearly show that significantly different results are obtained that depend on the mode of vibration measured. The measured charge sensitivity is high, while the measured seismic masses are excessive. Then the force transducer was mounted on a fairly rigid isolation mass and the tests were repeated. The resulting data show a highly consistent charge sensitivity and a seis- mic mass that is close to the anticipated value. A theoretical model is proposed that assumes the force transducer output voltage is proportional to the sum of the effective force applied and the load celi base strain. When this model is combined with beam-vibration theory, an equation results that shows that the apparent seismic mass depends on the mode of vibration. The theoretical decrease and measured decrease in apparent seismic mass are rea- sonably correlated. Thus it is demonstrated further that force transducers are dependent on their environment, in the case bere investi- 14.7 19.0 16.9 20.4 30.9 25.7 gated, base strain caused significant measurement errors. These measurement errors depend on the orientation of the transducer relative to the direction of bending. Disclaimer There is no intent to discredit the Kistler Model 912 force transducer in this paper. A PCB model 208A03 force transducer of similar design showed the same strain-sen- sitivity trends. The design of these devices should bave lit- tle sensitivity, but manufacturing imperfections cause these characteristics to exist, and as users, we need to be aware of their existence. Acknowledgement Professar Paolo Cappa wishes to thank the Italian Na- tional Research Council for financial support under NATO- CNR Senior Fellowship No. 217.26. Both authors appreciate the support of both the Department of Aerospace Engineering and Engineering Mechanics and the Engineer- ing Research Institute at Iowa State University for the ex- perimental facilities provided. References l. McConnell, K.G. "The lnteraction of Force Transducers with Their Test Environment," Modal Analysis: Intemat. J. Analytical and Experi- mental Modal Analysis, 8 (2), 137-150 (1993). 2. Dally, J. W., Riley, W.F. and K.G. Instrumentation for En- gineering Measurements, 2d ed., John Wiley &: Sons, New York, 323-325 (1993). 3. K.G. and lflroto, P., "Force Transducer Bending Moment Sensitivity Can Affect the Measured Frequency Response Functions, " Proc. 11th Intemat. Modal Analysis Conference, Kissimmee, FL (Feb. 1993). 4. Rao, S.S. Mechanical Vibrations, 2d ed., Reading, MA, 394-408 (1990). Experimental Mechanics 289