TETM U3C WS [mm M3ST CP RE

MMCZS--AE-S-59

" ( ~'cS) c? TM.: UOwm.T Sa •T wcmT -_______________Tst__

S16-3] May, 1990 21005-5059 SITP-FES 27 Dec 1989UAEC PRWJECT NO: 8-EG-175-018-035 rM2LMTJ2 A==: US Army

TYPE CF TEST: Ft. Belvoir RDUE C44] -1

Technical Road Testing of the 18,000 BIU Air CMIT M NO: D:| Craxiticners (Vibraticdn Profile Not Applicabl( d I' 3 f

ITIC FILE COPY o. ND:330-63012-40

AD'A226 900 BV . .T

- y-ciobjective of this test was to reco~rd the frequencies and inagnitxies o,~viand shtck force Jixrts at various specified locations of the air conditioner during thenormally expected life cycle transportation enviramwaet.

/

TEST rM (S)

Two 18,000 BTU (horizcatal, om-pact) air conditionrs (A/Cs) were mounted in astardard S-280 shelter in a manner consistent with normal (historical) location andhadware. Upon arrival at U.S.Army Canbat Systems Test Activity, Aberdeen Proving Ground,MD. (USAC=TA), the shelter assembly was loaded on a 2 1/2 ton truck ard secured usingstardard tiedowns and procedures. To produce a worst case ( maxi==• shock and vibration)profile for truck mounted equipmnt, the shelter was eTpty except for the installed aircr-iditicrers. Te track was not weighted with any additional ballast.)

-- testing was conducted on road corses at the Munson And Perryman test areas ofAberdeen Proving Ground. The vibration and shock envircnmnts for the test w-re derivedfrom discussiors between Belvoir R D & E Crt and USSTA. The various courses representa wide selecticn of the severety of possible transport envirormnnts that the A/Cs may besubjected to during a normal life cycle. The limited mileage of the coducted testingwas not interded to represent the total equivalent wear of a life cycle of normal use. ,

The tri-axial acceleraneters were installed on the roadside A/C at the locationsmarked by the test sporsor. The instrumnted testing consisted of the mini•n durationperiod reqaired to mcunitor a representative sample of each test course at each selectedgrxn speed. Cperaticnal speeds of the test vehicle were controlled by a pace vehicleequimped with a calibrated fifth wheel driven speedanetr. Instrumented testing wascrducted on the Mzso area izproved gravel road, Belgian block, two-inch washboard,three-ix•i spaced burp, and six-inc washboard test crses, and the Perryman P1I levelcross-codrty road course. Detailed course descripticns are contained in TEC=4 TOP1-1-011.

The accelerometer data cutputs were relayed to high speed magnetic tape at an onsitemobile cmputenized data collection van. AIl data was validated at the time of the test." Data reduction to the format requeqted by the sponor was acr-r~pli-,Ied with the use ofspecial czputer progrwmnir developed at t•;-sqTA. Details of the instrmcntation

tlation, mcnitong, and data reducticn are contained in the enguineering report(SAFl 1).

STEAP,-4T Farm 180, 1 Jul 74 Edition of 10 Jul 70, may be us, U,

S *

~MT MM NO: SITX=-AE-S-59 (Cantimnd Use additional sheets, if rqyi)x-d)

Prior to any testing the cperation of the A/Cs was checked. The A/Cs were con tedto a'portable eletrical generator pcx-•-r supply. Cperation was onfirmd by listening tothe fan unit operate, and confirming cool air was circulated inside the shelter and heatedexhaust air was discharged fran the A/C unit. Mihe checks were performed by the spxisorrepresentative. No instnmented mearemnts ware required for the operaticral unit. Thisbrief cperational capability confirmation check was performed after the completion of theirstrumented data collecticn for each different test cxurse.

. In addition to the instrumnted mileage, some ncranstrumnted mileage was aoaimlatedto complete the sponsor's requested testing. The operaticn of the A/C units was confirmedby WXSIA personnel at the cxpletion of all testing in the same manner as above. At therequest of the test sponsor, an additional 150 miles of noninstrumented transportabilitytest operaticon was cndted on the Perryman Prl cross-country test ourse.

1he total mileage accumulated during the testing at LMAAM is not considered asignificant portion of the total life cycle mileage for t1e shelter mounted A/Cs and isnot conidered ccrxlusive of any findings of operational capability or endurarne.

The recorded data of the frequencies and magnitudes of vibration and shock forceinputs at varicols locations of tha air conditicner was reduced to a series of plots ofPower Spectral Density (g2/Hz) for each displacmnt axis at each monitored location.Each chart is identified by test corse axi ground speed information. The data isprovided to the sponsor in the requested formats. Mhe data my be used to produce ararxcm-cn-rarda vibration table schedule for an i.-ut more realistic of actualconditicos than the outdated sine sweep schedule currently used for laboratory testscreening and develcpmTnt of A/C units. The data will be ccapared to a.irrentspecificaticrns to determine if changes in future proction requirements shold be made.

A detailed presentation of the test setup, instrumentation and data interpretation ispresented in the Engineerin Report (Dncl 1).

This is the final report for this project.

Nictolas SchultzLeigh Sanders tM&AýTxbkert MKinnn A

IEtI( A. At, T DflECItR cI L. MMWOK, DnfnM JASE DIR, SACS-h

This test record signifies thatCdr, TDXt, AMITN: A=-TA-T (2 COPIMS) the nxysted tes-ting has bexn01r, ETIEC, ATIN: SITFE-FES (2 CPIES) ciaTleted. It cdes not constitute01r, LWV A, ATN: ST7-7-AE-S?4 (2 COPIES) apprval or izsanproval of theDir, USABIL, AM': SLC3R-CT-T (SrFl) (2 COPIES) test item by U.S.APMY ccmT

.4dm, Ter- Info Ctr, ATI: FLAC (2 0PIE2S) SYSTID IEST ACtITY, Abc-xdwen'rYvinM Qr , MD

STAP-T Form 180, 1 Jul 74 Edition of 10 Jul 70, may be used

ENGINEERING TEST DIVISION

VIBRATION TEST BRANCH

USATECOM PROJECT NO. 8-EG-175-018-035

DATE: 27 June 1990 REPORT NO. 90-LR(V)-62

TECHNICAL ROAD TEST

OF

18,000 BTU AIR CONDITIONERS

(ROAD VIBRATION TEST)

DATES OF TEST: 16 through 31 May 1990

PREPARED BY: R. A. McKINNON

APPROVED:

X AAChiefEngineering Test Division

U.S. ARMY COMBAT SYSTEMS TEST ACTIVITYABERDEEN PROVING GROUND, MARYLAND

90 08 10

TABLE OF CONTENTS

SECTION 1. EXECUTIVE DIGEST

1.1 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1-1

1.2 TEST OBJECTIVES ..................... ......................... 1.1-1

1.3 TESTING AUTHORITY ..................... ........................ 1.1-1

1.4 TEST CONCEPT . ................................................. 1.1-1

1.5 SYSTEM DESCRIPTION .................... ....................... 1.1-1

SECTION 2. SUBTESTS

2.1 ROAD VIBRATION TEST ............. ....................... .. 2.1-1

SECTION 3. APPENDIXES

B TEST DATA ................... ........................... B-I

G PULSE CODE MODULATION DATA ACQUISITION SYSTEM. ...... .......... G-1

H DATA VERIFICATION ............... ........................ .. H-i

I VIBRATION DATA ANALYSIS ........... ...................... I-1

J DEVELOPMENT OF LABORATORY VIBRATION TEST SCHEDULES .... ....... J-1

Acces-ofi F,:rNTIS Cl-,,e.1 #

QUIC TAB LEUojnnooj:icCj L2Justfifc'to I-,

STATEMNT "A" per Nichal Schulterze 8y ___q_

Belvoir Research, Development andEngineering Center/STRBE-FES, Ft. BelvoirVA 22060-5606 Av~iIl:bIty C<desTELECON 8/14/90 VG A

Dim

(Page iI Blank)

SECTION 1. EXECUTIVE DIGEST

1.1 SUMMARY

The shock isolators under the air conditioners did rot attenuate thevibration energy to the air conditioner while the isolators under the compressordid reduce the energy over the upper parts of the 0 to 500 Hz frequency band.

The air conditioner compressor was subjected to shocks caused by itsimpacting other components.

The vibration test schedules developed and listed in Appendix B aredependent upon the transport scenario given in ITOP 1-2-601, 1 March 1988. Thiswas used because the sponsor was unable to provide an item specific transpcrtscenario. Changes in the total transport distance and/or the percentage offatigue inducing road surfaces will affect the laboratory test time.

1.2 TEST OBJECTIVES

The objectives of this test were:

a. To determine The vibration environment during road operations.

b. To develop laboratory vibration test schedules to simulate the fieldenvironment.

c. To determine the ability of the 18,000 BTU air conditioners to operate

after exposure to the field vibration environment.

1.3 TESTING AUTHORITY

None.

1.4 TEST CONCEPT

The 18,000 BTU air conditioners were subjected to road vibration testing.The testing was conducted from 16 through 31 May 1990. The Vibration TestBranch of the U.S. Army Combat Systems Test Activity (USACSTA), Aberdeen ProvingGround (APG), MD was responsible for instrumenting the test item, conducting thetest, and acquiring and processing all test data.

1.5 SYSTEM DESCRIPTION

Two 50/60 Hz, 3-2hase, 208-volt, 18,000 BTU/hr compact horizontal airconditioners, KECO Model No. FI8H-3SB, were installed in the front of an emptyS280(B)/G shelter (SN 138) as shown in Figure B-1. The air conditioners weighapproximately 77 kg (170 lb) tach. A rubber grommet was located above and belowthe mounting bracket at each of the four air conditioner mounting bolts. Fourrubber shock isolators were installed at the base of the compressor.

(.2-1(Page 1.1-2 Blank)

SECTION 2. SUBTESTS

2.1 ROAD VIBRATION TEST

2.1.1 Objectives

The objectives of this test were:

a. To determine the vibration environment during road operations.

b. To develop laboratory vibration test schedules to simulate the fieldenvironment.

c. To determine the ability of the 18,000 BTU air conditioners to operateafter exposure to the field vibration environment.

2.1.2 Criteria

None.

2.1.3 Ijst ProceCzre

The empty S-280 shelter, housing the two air conditioners, was loaded andsecured in the bed of an M35A2C 2-1/2-ton 6x6 dropside cargo truck(VIN: NK080J 0640-10781) (fig. B-l). The truck was equipped with size 9.00-20tires, Load Range D, inflated to 345 kPa (50 psi). The roadside air conditioner(SN 873466) and mounting hardware were instrumented with nine uniaxialpiezoresistive accelerometers arranged in three orthogonal axes as specified bythe test sponsor. The curbside air conditioner (SN 873480) was notinstrumented. Photographs of the accelerometer locations are included inFigures B-2 through B-4. Table B-1 contains the locations and specifications ofthe accelerometers used during this test.

An on-board pulse code modulation (PCM) data acquisition system was used toacquire data during the various test configurations. A detailed description ofthe acquisition system is presented in Appendix G.

Prior to testing, an electrical calibration was performed on all channels.The accelerometers were calibrated by shunting the transducer with a calibratedresistor which produced a differential electrical output equivalent to acalculated acceleration. This calibration was performed in both the positiveand negative directions.

A linear least squares curve fit was performed on the calibration data todetermine system linearity and to provide a quality check on the calibrationdata. Any chtnnels which demonstrated high noise levels, DC offsets, and/ornonlinearities could, thereby, be detected and corrected before test initiation.

2.1-1

2.1.3 (Cont'd)Vibration data were acquired while the air conditioners were transported

over the test courses at the speeds listed in Table 2.1-1. The speeds weredetermined from a fifth wheel and indicating head mounted on a pace vehicle. Anattempt was made to input the vehicle speed from the prime mover's speedometerinto the data stream. This was not done because the vehicle's speedometer cablewas not functioning properly, therefore requiring the pace vehicle. Data wererecorded for approximately 31 seconds during each different speed run on thetest courses. The short length of two of the test courses (Radial Washboard andThree-Inch Spaced Bump) mandated that less data be acquired.

TABLE 2.1-1. TEST COURSES AND SPEEDS

SpeedTest Course km/hr mph

Belgian Block 8, 16, 24, 32 5, 10, 15, 20Two-Inch Washboard 16 10

Radial Washboard 24 15Three-Inch Spaced Bump 32 20Six-Inch Washboard 4, 8 2.5, 5Perryman Paved 32, 48, 64, 80 20, 30, 40, 50Perryman Cross-Country No. 1 8, 16, 24, 32 5, 10, 15, 20

The instrumented air conditioner was operationally checked periodically andat the conclusion of the road vibration test.

A limited amount of data processing was performed at the test site for thepurpose of data verification immediately following the data runs. A morecomplete form of verification and all data analysis were performed on a separatecomputer system after the data acquisition process was completed. Dataverification and analysis procedures are presented in more detail inAppendixes H and I.

The final data (after verification) were processed qy performing therequired Fourier transformations and creating power spectral density (PSD) filesfor the individual test runs. A summary of the parameters used to compute thePSD functions are presented in Table 2.1-2.

TABLE 2.1-2. PSD COMPUTATION PARAMETERS

-Parameter Value

Sample rate, Hz 2083.333Block size 2048Frequency bandwidth, Hz 1.017Windowing HarmingNumber of linear spectral averages 32

2.1-2

2.1.3 (Cont'd)Amplitude distributions (rms, +peak, -peak, +99.9 per.-ent, -99.9 percent,

+99 percent, and -99 parcent) were performed on all channels for each run. Theamplitude distribution data were compiled (tables B-2 through B-7) byhistogramming the data into a 512-bin field and calculating cumularivedistributions. The average value for each channel was removed to account for DCoffsets in the instrumentation. The peak values presented in this report arethe true peaks, unlike previous Vibration Test Branch reports which reported the99.9-percent value as the peak value in order to remove extraneous noise fromthe data. The current data verification procedure is able to identify noisydata so it can be eliminated from the analysis.

The laboratory vibration test schedule development process is started bygrouping (overlaying) PSD data of the mounting bracket for critical speed runison four of the Munson t-.st courses. The courses and speeds selected were theones traditionally used in the development procedure and are listed inTable 2.1-3. These PSDs are overlayed to provide a composite spectrum.

TABLE 2.1-3. VIBRATION SCHEDULE TEST COURSES AND SPEEDS

Speed

Test Course k_/hr M~h

Belgian Block 32 20Two-Inch Washboard 16 10Radial Washboard 4 24 15Three-Inch Spaced Bump 32 20

The overlay process was accomplished by using the individual average plusone standard deviation spectra for the grouped conditions. The use of theaverage plus one standard deviation is intended to add some level ofconservatism to the data. For the individual spectra, the addition of astandard deviat'on has the effect of slightly more than doubling the averagevalue while the addition of the standard deviation during the overlay processwith a sample size as small as -wo runs has the effect of enveloping the data.

The overlayed data were then exaggerated to reduce the laboratory test timeto a more reasonable value. A real world exposure time for cross-cotuntryoperation was calculated to be 330 hours, based on the average speed of thefour test courses utilized and the expected total transport distance listed inITOP 1-2-601 dated 1 March 1988 40,000 km (25,000 mi). For a wheeled vehicle,it is customary practice to ignore the vibration levels from smooth terrainssince those levels are normally lower than those measured while the vehicleoperates on cross-country surfaces. Sixty-five percent of the total distance isassumed to be on cross-country surfaces. One-third of the cross-countrydistance is assumed to be of the severity of the Munson Test courses. Thisreduced the distance to 8,700 km (5,400 mi). In order to reduce the laboratorytest time an exaggeration factor of 2.0 was applied to the-data. Thisamplification reduced the test time to 24.5 hours for each of the three axes.

2.1-3

This is still a relatively long time for a laboratory vibration test; however,the exaggeration factor was limited to 2.0 because it is unreasonable to amplifythe data to such an extent that the test level will exceed the yield or ultimatestrength of the material. The schedules are included in Figures B-5through B-7. Table B-8 contains the breakpoints for these laboratory vibrationschedules. A more detailed description of the air conditioner environmentschedule development is presented in Appendix J.

Frequency response (transfer) functions were performed by dividing responsePSD by thair corresponding input PSD and taking the square root of the result.This is discussed further in Appendix 1.

2.1.4 Test Findines

There was no visible damage to the air conditioners as a result of the roadvibration test. The air conditioner was operational throughout and at theconclusion of the road vibration test.

The acceleration data were analyzed in the form of amplitude distributions(tables B-2 through B-7), time history plots, and power spectral density plots(fig. B-8 through B-1-1).

The acceleration levels increased on each course as the vehicle speedincreased. All of the test courses with the exception of paved road had acritical speed in which the accelerometers on the compressor measured shockloading. A typical time history plot showing these shock pulses is included inFigure B-17. These shock pulses were present at all of the measured speedsabove the critical speed.

The frequency response functions indicate that the vibration energy wasamplified through the shock mounts located at the bottom of the air conditioner(fig. B-18 through B-20). The frequency response functions between thecompressor and the compressor mounting bracket show attenuation in the verticaldirection throughout almost the entire 0 to 500 Hz band (fig. B-21). Thetransverse and longitudinal transfer function plots show attenuation above200 and 325 Hz, respectively (fig. B-22 and B-23).

2.1.5 Technical Assessment

The shock isolators under the air conditioners did not attenuate thevibration energy to the air conditioner while the isolators under the compressordid reduce the energy over the upper parts of the 0 to 500 Hz frequency band.

The air conditioner compressor was subjected to shocks caused by itsimpacting other components.

2.1-4

2.1.5 (Cont'd)The vibration test schedules developed and listed in Appendix B are

dependent on the transport scenario given in ITOP 1-2-601, 1 March 1988. Thiswas used because the sponsor was unable to provide an item specific transportscenario. Changes in the total transport distance and/or the percencage offatigue inducing road surfaces will affect the laboratory test time. Thecontrol accelerometer for the laboratory vibration testing should be placed inthe same location as where the data were obtained (fig. B-4). The airconditioners undergoing the laboratory test should also be mounted to thevibration exciter in an identical manner as those during this road t:st.

2.1-5(Page 2 1-6 Blank)

SECTION 3. APPENDIXES

APPENDIX B - TEST DATA

Table No. Description

B-1 Transducer locations and specifications.

B-2 Amplitude distribution data.through

B-7

B-8 Laboratory vibration test schedule breakpoints.

o. Description

B-i Photograph of 18,000 BTU uir conditioners, S280 shelter, andM35A2C truck.

B-2 Photographs of accelerometer locations.through

B-4

B-5 Laboratory vibration test schedule plots.through

B-7

B-8 Power spectral density plots.through

B-16

B-17 Time history plot showing shock loading.

3-18 Sample frequency response function plots.through

B-23

3-1

b- ai C' ,- 0, .4 , * I

0

O. 0' '0 0 -C 6M %f0 11v -4 Ift lOW'

6 bo 0 000000 On -14 s-sw r kn V n kn6)L n kn V e-inJ - wL0s

z 14. 4 1 . I I IO. I

rAS P-) -C .- 44 2: Z- t

C1 en ~ 0 -

co 04 P0C40%

at 0 0 1 -en e

JJ am .2. z to4 .4o~((N~4~C

0 0 0 0 0 0 A'-~-

Of- 2-!%C4L

�j (N � �- 0 ?� O� 0" u�

�

�

-� -4-4 �4 , ,

0 00" 0% dPI 0'.0.-.4�Lr�4

0� �-40�4i�04� -� f .- 4Ou� CN � .-4 � '.0

�22222 �' �

%t' �4 0�O4(Nu�4:I- 0'l .- d4: 44 . 0%1 (N .-4 .-4 � f. t. r-. (N -�4:

� o�i� .-4 .-4 -� 4 r-. (N � e'� 410 0

-� -4 -, C" �U�(�4.-4p p -�1-4 -'4 01 �3I'�'O!"�0 I-' ..-4*0 I o-4��0�-c'jr'-� ...� �0.. 0'I '*'*'*.' 0.. Cci .- 4-4 (N I'-. �. '.0 � (N '.00 0+1 -4-4

U(N '0Z 4.d Zo i.� �-4'04�Lf� C $� �- 41 � .� 0%4�'CCNC�.-4U4f'� 1-4 41

41 �4�f�'04�

C .� C .�I (N0�!'-(N�4(N �

4: 4: 4*� (N � � � 00 '�- (N� '-4 '.0(N'00.r�0��-

* a4 - (N�(Ne-lf-�* 0 � 0.-40r--r�N4r'-,-� 0 U,

1 �

-4 �fC"4.r'.0'.o (N �Ifz� M ca� '� -4-4-4

U .- �0 04: -4 414141 s-I 4)4141.0 � E- .0

U(JUC

�.0.0.0 .0.0.04)0 4)0

- 4)04)04)0 See41 CCC 000 41 CCC 000

� 0 � � 0 �.4-44J��jCCCOOOOOO JJ CCC000000

000000

000000 000000941 CCCUU)�I'A4)�Ifl 41 CC�UU,4�Q�

000414141414)41000414141414141

0.0. 0.0.'� a a E

-4'-�.-4000000 ."4.'-40000004:4:�OQr.

B- 3

Ln ~ ~ ~ ~ ' I? %D CN 17 ON2 4o 0t vl 4% n ýe

en en -4 . . . .

ý4 C I N %D ~ CF k 00 1 n

rl r V-4 * In cc %0 r- r- i

In4 .1 ..( ( . .- .' .e . . . .

w 0c% yýWS%w0l% %C ,a r 1 Qw e

-4 tN44J %

z0 CP% vi w 2I)e4 ot1

04 ON 'T P, f.1 ON4 ý% .

00'u) eo rrso r-- X 17% 01 LM 01 -4S( ~

U5- t-4 - 4 4 4 ( V-1 E4 - 4- 44-

+61

0 4) 4)

00 b0b o a

0 r. 0 00 o4J "AJ0.006.0.JAJ 4J

o ~o0 000 o00

0 00 0 00 0 00 0 00V in U2 to U) 0)U I~ ( U U) U U) S) q

-4 4 -4 0 0 0 0 0-4

M- ý r- C4 ~~

.-4 47 1ý Er0 N

0o M~ m 0'4 m3-4~

'C~- -% 4 0

1; 1ýC 1

-. 4 cy - sýAd

4:,

-4 1e -)r N

Go

0C 0 0C. 4..4 1,J

0 -4-400 v

Io 0 .- ,- 0-- '-

TABLE B-8. LABORATORY VIBRATION TEST SCHEDULE BREAKPOINTS

Vertical Transverse Longitudina.Freau£•y Amplitude Frequency Lkmplitude Freguency Amolitude

5.00 0.5555 5.00 0.0555 5.00 0.10146.00 0.1526 6.00 0.0455 6.00 0.08118.00 0.2911 7.00 0.0479 8.00 0.27399.00 0.2161 9.00 0.1171 9.00 0.2672

11.00 0.0399 10.00 0.1200 10.00 0.310113.00 0.0409 13.00 0.0277 14.00 0.055814.00 0.0537 14.00 0.0298 16.00 0.058715.00 0.0419 16.00 0.0233 17.00 0.077116.00 0.0475 19.00 0.0402 24.00 0.039421.00 0.0243 22.00 0.0291 27.00 0.042522.00 0.0311 22.00 0.0206 28.00 0.031526.00 0,0268 23.00 0.0245 20.00 0.048128.00 0.0163 25.00 0.0142 32.00 0.035730.00 0.0249 27.00 0.0182 35.00 0.068134.00 0.0319 28.00 0.0153 41.00 0.069835.00 0.0419 29.00 0.0191 47.00 0.042537.00 0.0379 34.00 0.0142 49.00 0.064839.00 0.0463 41.00 0.0245 56.00 0.023440.00 0.0379 45.00 0.0149 57.00 0.036642.00 0.0475 49.00 0.0338 69.00 0.030046.00 0.0335 54.00 0.0110 70.00 0.038549.00 0.0475 59.00 0.0264 74.00 0.026551.00 0.0268 63.00 0.0186 81.00 0.044754.00 0.0327 68.00 0.0444 122.00 0.005856.00 0.0163 69.00 0.0277 130.00 0.009658.00 0.0268 81.00 0.0613 150.00 0.010162.00 0.0176 85.00 0.0284 161.00 0.006164.00 0.0261 91.00 0.0383 178.00 0.002367.00 0.0163 149.00 0.0047 199.00 0.008970.00 0.0282 155.00 0.0074 215.00 0.001471.00 0.0167 195.00 0.0022 226.00 0.004781.00 0.0255 217.00 0,0024 270.00 0.0024

102.00 0.0189 224.00 0.0069 287.00 0.0032109.00 0.0104 240.00 0.0052 310.00 0.0012128.00 0,02961 265.00 0.0093 368.00 0.0135135.00 0.0148 314.00 0.0037 457.00 0.0005149.00 0.0275 320.00 0.0069 481.00 0.0038168.00 0.0107 346.00 0 0034 500.00 0.0018182.00 0.0204 371.00 0.!61193.00 0.0199 386.00 0.0018226.00 0.0013 402.00 0.0058260.00 0.0159 448.00 0.0006500.00 0.0026 500.00 0.0012

B-6

C14

'41

.4'

N a

* U

Ak

6- . 7

a

4

/

tsp� S

4,

U,'U

.0

Lio02'14 411* + 0)43

¾ Li0.

0U

0Ar's'"

0 � o4

t1� "--'v 41 4 -�

U

C -

''Ii ;''1 N-

43

LiI 0

4' A 80t

I I'1 I

I' I

111

I r

941'I I

j.1. I - II I

I I44 I A

V 1I I I

r- 41, I

B- 8

4-4

A

ca

.34,4

-,0

1.4- B-9

,� j,'.�'. rb* u-s 4

44 � J Pr �V��1 05

4',", F tg 4"';' '4 '4- 4

try 9 Li4 4)M

U�0

.0

toA'' 4

1, -.4- 0

1.4

.4 4)C0-4

-t '4 'Ia''A ' ' '�

'4,''"''

4" 0

4/ 443�r 4' . 2

4 0N

' '"A'

'*4%,A. -� $.,4,,4§,.1 ".

- 4.4 ( ' :'t

t�'r�"� ''N*� -4

"'A'' ' 4 �- "''cI4 ' .4

[tA '� 2 'nt.�4>c2�w <'�'� 7 'j� ' '> ' "' '4''

"s' ��%�4' I',

NJ� 4 'i Vs.U ' '4

4-4- 44"'' '"ji

4" 'r4'.

4rn• a ' 44rh,.ss. 'iAo.�cet Mt ka4� aw& 44 � '.ajk �

3-10

-d)

ii)

4.)

0-.4

LaiJ'

C)4

cc4

cc

oG

'-4D

-H U3, S 3

B-11

c

600

B-12)

Cie

Id-

- 0q

w' *

cc~

U! -4

24 V3ad CS 0 - 05.

-B-- 13

61:

3r304

3C0

PL4

rx~

%S CD .

do-.- CD IW

31

100

a01

B-154

SO 5

B 1

~~0

p-4

S• •o

-•-46

, , , I I II I I I I0

02

0

PI

400

B-17

Chi,

to4

LA

to4f~p 4w

w MAI

ZH 44 os5 C0

02

- .-

2 0

* .qg

ZN N3*,d • " - (~

B-19

C4:0 0

Iin(.1 -4

B-20

L3.

c2;0

* C

I.-4

40.

UH U3 20

B-21.

3r.

-4

* 0

-4

0 0

-72r

i-i2

LO0

r-N

CDN

cc_ 0 cc

- 0

otce-C- 0.

0

h4

-4

r4-

-- 2 3

LO

w.

- 4w

m~ -4

-4

0

(~~ 0

NL

0*

010

10 CU

0D

CD

10

OR 000

0 0 ;0 0 0 0

CD00 0 0 0o

*w *

B24-

aa

'2' 0

I, U7

3c2

o* 4-)

W 0

aL

M'J C

0 *

09 0

44 do

CD 42

U) -4

I I I ICDccdevODqCDO wO

CD U) do Id

B-25

0-4

l..

'4,I

CD

00

z8

co C to

toIn Iv l

B-2C

II

lo

op

S.

Cc*

L 2CY 0

0

CD4

I-.

CD0

U anQ)

ot

10

to zqs i 0CD lw c

0l In S.aI

B-27

-4 ' -4

CU

M o

0

CPCD

-40

* -o

(n.4

41)CD1

- °

I--C

oQim

B-2

cm4. C1C. dJ

do co

B-284

b.

0

'II

- 4

141

B-294

an

000

(00

cm .o op

2-

-ýPage ~ B-0 lak

APPENDIX C. PULSE CODE MODULATION DATA ACQUISITION SYSTEM

Figure No. Descriprion

G-I Block Diagram of PCM Dar Acquisition System

G-1

APPENDIX G. PULSE CODE MODULATION DATA ACQUISITION SYSTEM

Each of the transducers were connected to a USACSTA signal conditionerwhich provided excitation, amplification, calibration, and low-pass filtering.The accelerometers were low-pass filtered at 500 Hz. Interchangeable six-poleButterworth filter pads were installed in the signal conditioner with the filterfrequency being the nominal frequency at which the transducer response isattenuated by 3 dB. A block diagram of the PCM data acquisition system isincluded In Figure G-1.

The output from each channel of the signal conditioner was connected tc anEMR Model 372-03 Pulse Code Modulation (PCM) Encoder. The encoder multiplexedthe incoming analog signal which was then sampled at 2083.333 samples persecond per channel by a sample-and-hold amplifier, digitized by a 10-bitsuccessive approximation analog-to-digital converter, and converted tononreturn-to-zero-level (NRZ-L) code for transmission. A programmable read-onlymemory was used for the enc~der format control. The encoded PCM data were theinputs to a Conic Model CTL 510 230.9 MHz Traasmitter for transmission to therewote data handling facility. The signal conditioner, encoder and transmitterwere mounted in the cab of the prime mover. The transmitted NRZ-L code wasreceived by a Scientific Atlanta Series 420S Receiver and recorded on aHoneywell Model 1-1 PCM Tape Recorder along with voice annotation of theindividual test runs and IRIG-B time code. Simultaneously, the PCM pulse trainwas passed to a Loral Instrumentation ADS-100 PCM Decommutator.

The ADS-100 system consists of sevtral modules. The input buffer and bitsynchronizer module3 recovered the serial pulse train from the data link noiseand disturbances. The frame synchronizer and data distributor modulesdemultiplexed the serial pulse train into 16-bit words. A parallel input modulewas used to input digital IRIG-B time code into the ADS-100. The compressor anddirect memory access (DMA) modules performed data sorting and transfer to thehost cJmputer, an hP2? -E series minicomputer. The data were temporarilystored on the HP systew uisk and later transferred to digital tape to provide apermanent storage medium. The ADS-100 system was independent of host control;however, software residing on the minicomputer controlled the hand shakingbetween the ADS-100 and the HP21.MX-E during data acquisition.

G-2

iii

> CL,w 2,

"oc

!0 o

F--

S0

0 OF

C-)CDe C-&

V

000

a~ c'w>

C-

(Pag 0G-4 Bak

APPENDIX H. DATA VERIFICATION

The data verification that was performed consisted of searching theindividual channels for frame errors, wild points and DC shifts. A frame erroroccurs whenever the PCX stream is interrupted, which induces artificialhigh-level spikes into the data stream. Data approaching full scale values, inaddition to frame errors, are flagged by the frame error check program, becausethe program defines a frame error as being any data value which is greater than98 percent of full scale. The same program checks the data for wild points andDC shifts. A wild point is just like a frame error except its amplitude issmaller and does not approach full scale. The program defines a wild point as apositive or negative change of more than 100 resolution steps from the previouspoint. A DC shift occurs whenever the data average changes rather abruptly andthen remains constant after the change. The program defines a DC shift as ashift of more than 25 resolution steps from the previous average, with theaverage being based on approximately 1/6 second. These errors are obvious whenplotted in the form of a time history, but the quantity of data collected issuch that viewing it all as a time history is impossible. For this reason,the frame error check program is utilized, and the locations of all such errorson the system disc are printed out for a permanent record.

As a further step in the data ver~fication process, acceleration amplitudedistribution data were compiled by histogramming the data into a 512-bin fieldand calculating cumulative distributions. Table H-1 shows sample accelerationamplitude distribution data. The average value for each channel was removed toaccount for DC offsets in the instrumentation. The percentile columns representthe percentage of time the data was in that particular range. For example,99.9 percent of the time, the data value for channel 1 was less than 1.02 butgreater than -0.71. The units for the accelerometer amplitude distribution dataare g's.

TABLE H-1. SAMPLE AMPLITUDE DISTRIBUTION DATA

Run Number, Test Course, Speed, etc.

Channel 1 0.28 1.25 -1.05 1.02 -0.71 0.68 -0.55 0.37 -0.36Channel 2 0.31 1.32 -1.02 1.08 -0.78 0.71 -0.62 0.39 -0.42Channel 3 0.23 1.34 -0.83 0.96 -0.65 0.66 -3.50 0.29 -0.27Channel 4 0.22 1.22 -0.82 0.94 -0.58 0.62 -0.42 0.26 -0.26Channel 5 0.24 1.04 -1.02 0.85 -0.64 0.59 -0.49 0.33 -0.31Channel 6 0.36 1.50 -1.57 1.19 -0.92 0.81 -0.73 0.46 -0.46Channel 7 0.39 1.66 -1.37 1.30 -0.97 0.86 -0.81 0.50 -0.49Channel 8 0.21 1.27 -0.83 0.95 -0.60 0.64 -0.44 0.24 -0.24

H-1

The program which performs the amplitude distribution analysis and createstables such as the one above performs a number of data validity tests for eachchannel and provides messages such as:

a. Channel inactive.

b. Data one-sided.

c. Data noisy.

d. Data clipped.

e. Large rms value.

f. Large DC offset.

g. No data spread.

h. Shock present in data.

If a large number of samples are processed and if the dynamic range of thedata i-i small, a bin overflow can occur. This happens when more than32,767 samples occur in any one bin and usually appears as an inactive channel(rms set to 0.0). Although the amplitude distribution program is not foolproof,and the rules which determine data validity are somewhat arbitrary, the programprovides a very useful tool for *quick look" verification and analysis.

The test data were then checked for stationarity or time invariance. Dataare collected and analyzed for short periods of time wherein seasonaldifferences in performance and long-term vehicle degradation have no influence.Hence, a single time history record should adequately define the distribution ofthe data. Multiple sample records are used in all steady-state analyses underthe assumption that the data are stationary. This provides a broader, moredependable sample population and statistical base for analyses such asacceleration amplitude distributions and acceleration power spectral densities.The test for stationarity is based on the following assumptions:

a. The data to be analyzed are not of transient nature (e.g., singlemechanical shock).

b. The sample record taken will reveal the non-stationary character of therandom process in question.

c. The sample record is very long relative to the lowest frequencycomponent in the data.

d. Any nonstationarity of significance will be revealed by changes in therms value of the data with time.

H-2

-* Because the objective is to measure short term rapidly changing phenomenarather than seasonal differences or vehicle degradation, it is assumed that a,b, and c, for this purpose, are true. A nonparametric statistical test, calleda run test, is used to check assumption d. The run test consists of computingthe rms value for a series of subrecords from the time history (e.g., 30 ruisvalues over 1-second intervals for a 30-second record), finding the median valueof the computed rms values, and counting the number of runs (crossings above andbelow the median value) in the set of rms values. The number of runs is thencompared at the 5-percent level of significance to that for theoretical Gaussiandistribution, and the hypothesis that the data are stationary is either acceptedor rejected. Printouts list the average value, maximum absolute value, and rmsvalue for the data in the same blocks in which the PSDs will be computed soinconsistent blocks of data can also be' deleted from the analyses.

The statistical moments, or weighted excursions about the mean, were alsocomputed as part of the verification process. Only the first four moments aregenerally of interest and are described below:

M. - 1/n xi - x

M2 - 1/n!7(xi-x)2

M3 - l/n"Z(xi-x)3

M4 - 1/n'_(xi-x)4

The first two moments are commonly referred to as the mean and variance,while the third and fourth moments are sometimes called the skewness andkurtosis of the distribution. More often, a normalized version of the skewnessand kurtosis are computed as shown below:

Skewness - M3

(M2 )3 / 2

Kurtosis - M42I

(K 2)2

A normal distribution has a skewness of zero (as do all symmetricdistributions) and a kurtosis of 3. If the kurtosis of the data set is higherthan this, it is an indication that there is considerable datd out at highamplitude values. Because the kurtosis value is based on the fourth power ofthe distance from the mean, it is extremely sensitive to wild points, even insmall quantities. This value has proven to be an effective screen againstabnormal data.

H-3(Paga H-4 Blank)

APPENDIX I. VIBRATION DATA ANALYSIS

The acceleration power spectral density was computed by dividing the timedomain data into successive blocks of 2048 points, converting to the frequencydomain using the Fast Fourier Transform (FFT), multiplying this result by itscomplex conjugate, and linearly accumulating (averaging) these results over theentire data run. The number of linear averages applied was dependent upon theamount of valid data available for a particular data run. For a linearlyaveraged process, the number of statistical degrees-of-freedom is equal to twicethe number of averages used. The amount of averaging (degrees-of-freedom)determines the degree of confidence that the value measured is a truerepresentation of the actual physical phenomena. An error band, based on thenumber of averages, can be computed for various confidence levels from thechi squared/degrees-of-freedom distribution. When the FFT algorithm is used, anassumption is made that the time record being transformed (data block) isrepeated throughout time. If the time record contains an integer number ofcycles within the data block, the assumption is valid and the waveform is saidto be periodic within the time record. In most cases, the data are not periodicwithin the record which causes a truncation of the signal at the end of the datablock. Since the assumption is one of a repeated waveform, the analysis processassumes that the truncation is repeated throughout the entire data record. Theeffect in the time domain is an apparent discontinuity in the representation ofthe data signal. In the frequency domain, the discontinuity appears as sidelobes or additional frequency components and is known as leakage. A time domaintruncation technique known as windowing is employed to reduce the leakage. AHanning window was employed for this analysis.

The dynamic characteristics of a physical system can be described by afrequency response function, H(f), which relates system response to a giveninput by:

Y(f) - H(f) X(f)

where

Y(f) - Fourier transform of system responseH(f) - frequency response functionx(f) - Fourier transform of system input.

Since both Y(f). and X(f) can be complex values (real and imaginarycomponents), H(f) can be a complex value. In practical terms, the relationshipbetween the real and imaginary components at any frequency are used to find thephase relationship at that frequency.

I-I

The frequency response function is computed as the ratio of the cross powerspectrum, C , divided by the input auto power spectrum, G . The cross powerspectrum .isnefined as the Fourier transform of Y(t), Y(f), multiplied by thecomplex conjugate of the Fourier transform of x(t), X*(f).

H(f) - G yx(f)/G xx(f)

where

H(f) - frequency response functionG (f) - cross power spectrum between input and responseGyx(f) - auto power spectrum (PSD) of input.xx

Since the frequency response function contains real and imaginary values,it can be difficult to use in normal form. For this analysis, the polar form(magnitude and phase) was calculated and presented (magnitude only). The polarform was calculated from:

H(f) - (H r(f) 2 + H i(f) 2 ) 1 /2

where

H(f) - magnitude of frequency response functionH (f) - real part of frequency functionH i (f) - imaginary part of frequency response function.

When calculating the frequency response function, extraneous inputs andsystem nonlinearities can cause errors in the computation. The coherencefunction is a procedure used to check the validity of the frequency responsefunction and thus can be used to check causality (input a causes response b)between two signals. The value of the coherence function is a dimensionlessnumber between zero and one. A zero value indicates that the system may haveextraneous inputs or may be highly nonlinear. A value of 1 indicates completecoherence between input and output (linear system, only one input). Thecoherence function was calculated as follows:

C2xy(f) - GyX(f) 2 /(Gxx(f) G yy(f))

where

C2 (f) - coherence functionG _xf) - magnitude of cross power spectrumGyx(f) - input auto power spectrumGxx (f) - response auto power spectrum.yy

The computational techniques used to produce the cross power spectrum,input auto power spectrum and response power spectrum (block size, averaging,windowing, etc.) were the same as for the POD computations discussed previously.

1-2

These functions were not computed directly; however, the program dvpsd wasused to compute the transmissibility function described here as:

T (f) - transmissibility functionGY(f) - input auto power spectrumG (f) - response auto power spectrum.yy

It can be shown that the transmissibility function is equal to the squareof the magnitude of the frequency response function if the coherence functionhas a value of one and is larger than that in all other cases. Thus, apseudo-transfer function was calculated by determining the transmissibility andtaking the square root of the result. This pseudo-transfer function representsan upper bound of the actual frequency response function.

1-3(Page 1-4 Blank)

//

APPENDIX J. DEVELOPMENT OF LABORATORY VIBRATION TEST SCHEDULES

A series of computer programs have been written for the purpose of creatinglaboratory vibration schedules from field data. The programs are designed toautomate the process as much as possible and require operator intervention onlywhen judgements are required (such as shaping the schedules). The followingdescription is a "walk through" of a typical development project once the field(raw) data have been collected and verified.

The program PSDFL computes PSDs from raw data and stores them in a file.The program assumes that the data are in normal PCM digital multiplexed form andare located on the data disc. The program makes use of a control file, $PSDCL,which contains:

a. The name of the calibration file (a file which contains the valuesrequired to convert the integer data to engineering units).

b. The name of the amplitude distribution file (a file which c=tainsinformation on the location of the desired channels in the data stream and abrief location description of the desired channels).

c. The data sample rate.

d. The FFT block size (currently limited to 2048).

e. The word length (number of bits) of the data (typically 10).

f. The lower and upper frequencies to be analyzed.

g. The use (or not) of a Hanning window.

When the program is run, the operator enters the run number, the startingdisc track (location of data on the disc) and the number of tracks to be used(the amount of data analyzed). The program creates a data storage file with aname based on the run number. Each segment of time domain data is windowed witha Hanning window to reduce spectral leakage if this option is selected. Theprogram then computes three PSDs for each channel listed in the amplitudedistribution file. The first PSD computed is a linear average of PSDs computedfor segments of data of a length defined in the control file. During thisanalysis period, the. standard deviation and the peak value for each spectralline are computed. One standard deviation is then added to the average value tocreate a PSD which represents the average Nalue plus one standard deviation ateach spectral line (limited by the peak value in each spectral line).Experience has shown that the variation in the data over a test run issufficient to cause the standard deviation to be slightly larger than theaverage value, thus adding one standard deviation to the average has the effectof doubling the average value. Both this PSD (average plus one standarddeviation) and the peak value PSD (largest value at each spectral line) are alsosaved in the file.

J-l

Each data file corresponds to a test run (specific test course, speed, andtest condition) and consists of a major title (a line that specifies theamplitude distribution file used) and information on each channel analyzed,organized as follow:

a. An ASCII title line which contains the run number and a brief

description of the channel location.

b. An ASCII information line which contains:

(1) The FFT block size.

(2) The data sample rate.

(3) The analysis bandwidth.

(4) The peak value contained within the PSD.

(5) The number of linear averages used.

(6) The use (or not) of a Hanning window.

c. A binary record block which contains formatted integer data. (Thedata are formatted by dividing the entire spectrum by the peak value, thenmu:.tiplying these values by 10,000 to save file storage space; i.e., integernumbers use 1/2 the storage space of real numbers.)

The program PSDFL is run on the data for N runs with data from X channelsbeing analyzed from each run (note that X need not be the same for each run).The program will produce N PSD files (one for each run).

At this point, an overlay process is started. A program called PKPSD isused to create the overlayed spectrum. The operator enters (via a control file)the name of the master storage file, the names of the data files created byPSDFL (essentially, the data runs to be analyzed), the data elements (channellocations) within those files to be analyzed (those locations for which a commonspectrum is desired), which PSD values are to be used in the overlay process(average, average +1 standard deviation or peak) and the number of standarddeviations (due to the overlay process) to be added to the average valuecomputed during the overlay process. The program does not require that the dataelements to be analyzed Ie the samc for each file analyzed; however, all data tobe compared must have the same analysis bandwidth (a check is made by theprogram and the data are rejected if this is not the case).

J-2

A '

I- 7

tI

* /

The data normally used for the overlay process for each particular channelto be included in the process are the average plus one standard deviation PSDdata. Some of the rationale for the use of these data (rather than the averageor the maximum) are as follow:

a. The paak value at each spectral point is a value that occurs rarely(once) and is susceptible to contamination by noise (false data from anysource). To create a schedule based on peak data is to assume that the testitem is subjected to such an environment for a reasonable length of time, whichis false. The result is an overtest.

b. Some conservatism is built in by adding a standard deviation to theaverage to account for other terrain and vehicles for which no data wereobtained, effects of tire pressure, vehicle wear, track tension, etc.

The program computes the average and the standard deviation of all PSD's tobe overlayed (at each spectral line) and again saves the average plus X(selected by the operator, but normally 1) standard deviations (limited by thelargest value at each spectral line encountered during the overlay process).Conservatism is thus added twice during the overlay process. The use of theaverage plus standard deviation curve from the individual location PSD'saccounts for variations in a single location during a data run while theaddition of a standard deviation during the overlay process accounts forvariations among locations or test runs (courses). The standard deviations arecomputed during separate processes and do not have the same values.

It is the responsibility of the operator to determine the data locationsand data runs to be overlayed and to insure that the data for each of the runsand locations chosen are valid. The program PKPSD prints the channeldescription of each channel used in the overlay process so that the operator mayverify that the requested data have been analyzed. The rms value of each ofthe individual spectra is computed and printed with a recommendation to discardany data with an rms value less than 1/3 the average (all elements overlayed) ormore than 3 times the average rms value. Locations which fit this description(if any) are identified, but no action to delete these locations is taken by theprogram.

A program called CRVIB is next used on the master (overlayed) spectrum todefine the break points (frequency and amplitude points which define thespectrum in the digital vibration control system) and allow the operator tosmooth the spectrum for laboratory simulation. The master spectrum is plottedon a graphics terminal in a log-log format, and the cursor is turned on. Theoperator may position the cursor at any location on the plot and enter thelocation of the point (frequency and amplitude) by hitting any key (except E).The location of the cursor, in screen co~rdinates, is converted to engineeringunits of frequency (Hz) and amplitude (g /Hz), and these values are stored in afile. The program utilizes the "rubber band line" technique when the cursor ismoved, thus giving the operator a preview of the smoothed spectrum. The use ofa log-lcg domain plot allows better resolution in the low frequency portion ofthe spectrum which contains most of the spectral energy.

J-3

Laboratory test time can be reduced over actual exposure time in the fieldby applying an exaggeration factor to the amplitude (PSD) values of theschedule. The rationale is based on an equivalent damage theory assuming thatthe failure mechanism is fatigue. The techniques for computing an exaggerationfactor are based on a detailed knowledge of the test vehicle real-world scenario(i.e., the real-world exposure time), the desired test time and the materialfatigue characteristics. The accepted equation for test time compression is:

(Wl/W2)b/n - T2/T1

where

W- " field PSD amplitudeW2 - laboratory PSD amplitudeT1 - real world exposure timeT2 - laboratory test timeb - 9 (endurance curve con._ant)n - 2.4 (stress-damping constant)..

The breakpoints (and thus the spectrum) are now exaggerated using theprogram CRPSD. It is left to the operator to compute the value of theexaggeration factor based on the desired test time to real time ratio. Theoperator enters the exaggeration factor (by which the PSD levels are amplified)and the name of a new storage file into which the exaggerated spectral data willbe stored. The program takes the breakpoint lata from the applicable file,multiplies the amplitudes by the exaggeration factor and stores the values intothe new file. The program then takes the new breakpoints (exaggerated) andgenerates a complete spectrum (with the same analysis bandwidth) by calculatingpoints between breakpoints using a straight line fit in the log-log domain.This created spectrum (the actual laboratory schedule) is stored in a separatefile also named by the operator. A program called SHAKR is then run using thejust developed laboratory spectrum as data to insure that shaker displacementlimits are not exceeded by the test schedule. The program compares the spectralvalue at each frequency from 5 to 15 Hz with the maximum value known to functionproperly on the USACSTA system and alerts the operator to any changes requiredin the spectrum.

J-4

a

REPRESENTATIONS OF THE (COMPUTER COMBINED) VIBRATION AVERAGES

FOR EACH MONITORED DATA CHANNEL

DURING THE TRANSPORTATION PROFILE TEST

OF THE 18K BTU HORIZONTAL AIR CONDITIONER

A°

C33

ze3

ODIco3

- D 4

ZH U34OS a3

qI

I..

ml .

H WN O

3.I

qI

o q*

U.N

3,

I*

0

ZN � - cZSd

IDI*x

I _ _ _ _ _ _ _ _ _ ---

Q!

ZH ýM3 OS 5 - UZ3d

qI

I-a

'I

qIq

II '1a-

*

I I*

*rJ -4* ** a a

* Ia

Ix7 m

00

ZH M3 OS t- S

I..

--.

4..

at

z

.J -

10dD

ZH M34 OS t - aS

-

CDso

NN

U7

ZH 3d OS e -a34

ama

1C4

wIDZM a~ 03Z

-1

(-I '1

•--1Iz

4r4

S a

CDl 04

ak

10

CD0

(%J -3 OS!- ~

t I

I,

-""

09

3c0

- 39

do

ZH 43d OS !n-s

* 0

0 0!cc

C3

co, *a iiido -W,=o

c cc 02

zm v4 osv as

CL.

ZH -34S S

3.I

-4

,J3,

F14l'd • q

I

*

*

q

2'

1 :1In 2-

ZN WJd � -

3r3.

0 q

Im

zmv4e I

A

04

Lo

ZMW4O S

3S

I3

S

3I

- 3go

I-,

3

3

3;

I-J

3*z

3

3

*S U

- US.d

CD comI

ZH b3a OS !5 -S

,I

cm

w G

ZH -O

i I

('J -

q

a

cc

I VC

'O V

ZM •3d OS 0 - •

qr

44I

r4

'AM

LnqZH M4 OSZ aS

3c.

SN

IH

aci

cc

cc,

C-4

ZH M34 OS!

,/i

/l

q

CID

P r4

CD CD a

ZHWdC Zq3.A

K .

/1 _ _ _

t t , •, ...... •---"•'

/I,/q

!q:: • • ii i i i i• i ii J

IcI

ac

t U"cc v C

doq

co D-O

aHW403e a-

3c3

444

cm LO co

qI

a

q

q

a,

ZN � OZ -

q

qI

-i q*

-

Q. -

ZN W3d OS -

qIq

qI

a

=

qI

3'fri

I-j

q

LO- ('Snul -a a a a

Z14 �Jnd O�

I 3

atI

v04

ZH- U34 OS 0 - SJd

i3

- 3

t~i

cc.

-I

NI

out

tzI--,

do L Z- A

V)0CIcot G!,

qIq

- 3.9.

3.I3

3,

II

.0 3-

I

2 3.('J -

* * 0*S S S0

OS 0 -

00*

S.

fri- zLiJ �

3,

I-

3w

3

2

00 00 0000 00 a a

OS � -

LC

C2

ZH 834d OS - tSh

cici

3

m 04

-. 5 .

c W

w

('J

ZH ~ -as~i V

33.IxI

Psq

ab3N)

14D

-M U3 C

I

/

J

-4-

II

t�1

I..

- 3'

fri

-J

0I.0

0

*1� C., -0 0 0 00 0 0 00 0 0 0

0

ZN � 0 -

,1

.4-p

p -ck/

cc/

z10

CD CDTH m3d asn aI

I

NV

I ~'1

- q-I

m

I V

ZH •1d 0=•- U:

NowOKN

Vs

cc.Pq

ON

q

uv

I-

Vi

1-. ..*

,. a

• • i -

S.. .. . .-- . • • • m • mm m m mm ml mm

Nd

cc

1H 0' s s

q

NQ

IZN •3d S • - •

qi

o3

cc

ap

~l r

C') mn u v a

'V.

q.

'4

C.,

C

-

I

- L,...__ to

ZN �iJ�d OS � -

Iq

um

a's

l-,-,

cWa

co

'4

m 4n

,aNtvc

ZH Mi~d OS n -s

CA44

f, L -

'C

*4

It

�

*

___ -- -I

(3

'0

ZN �3id OS -

pJi

1-1 - Ile

ZHM4 S5 3

wz9

00

o,

3cQK

EC~

'Vt.

f: zz mC C%

ZH~~ 4n s s

C4.

ei'i

U9doIuqH

0to

_________________ ~-.-~------.---------do

ZH ZiJ.~ US - s

1 3.A

Oil

.4%

:50

Zvi Nn -aOS V ; 3

*-4

3'

.t�.

U.'-4

*

- I-� I

-1-t-�--�L I

'0 ('*) -4 2** *

ZN � -

cc

t;

cc

IuS..n

q•q

-4

Im

'.4

-4

I IIIIIII I ii iiiiiiiii IIIlllllllll IIIIIII IIIIIIIIIIII II II 1 [

cc)

4mW4Z. Sa o

, : el M

NN

ZH UI 0-

7 NIN

/q

cc

*8

"4•

II-

Z H 4

IZ

Io

* i

/N

6r6

Lt~l u

'1.3

q

w

'4

*1

11

4

I-

-A

�1-,�

:7

vi

ei

ZN -

IIN

l ;i

Z&;14 0ý ) a-

If

bo

CO-

('4 -

q

r'j0

I

-

Li'A

I �LL2rL

C, U,r\J

a a a

�3d �S � - as4i

___________________________________________ n,� nt.,tr �

'I.

-- [

- I- I•

4

ZH N~e o: --

1.4.

-

I I

U) UP,Le U

W3

007J

kk.

Z3

zm v~d s n C3

.d

as

H-i '

"" • "' • , |i i i i = I

"a.

N

.3

cc

IIT(-II

Zaio: 0

q

2

N

II

I*-

-

Z InS S

0

-4

ii

3'La

�i�i

a- L)

-4

-i

-4

� �

'0

S S S

ZH �I1.i OS 0 -

U,

II

S o

214~I~ t~ q

cm~

00

2HU4 3Z T

3.

3.

*-J

-4

r�4 (4)Id, -.

4 4 4 4 40

4D4

S S o

"Z-" d •o;~

q

-4L

ZH U3 OZ t-Z

Ns

C4,

q.

- 3

" S.'.. 8.~

-1

m~r~t~ *4

:• • • • i i i i -jIII

41

Iu

2L.

to

4H V3 s

U

• 7•.. • ' • i

III

L -

N

tall

II-

LnL

ali

08.

, -

-4l

Ile

ZHH- Zn s

It

uq

-4-

ZH4c S I4

~U9

qlpl

2-

'4 Q1

C-i

U, W34 e asA

Ll

Cii

ZH U3 OS CSI

Ell

NL

............

j W"

--U7s

ZH d~d n a as

II

~1 61

II'

Z-d4O 2

0&Il

c�J

�Y)

II

iS

-It, .4

-3

'3

I.

K

- L I -'�----.� �

'A,

'V C,, .4 VZN �iJd 0.. - rl:4d V

LI

'IJ

ZH Snv f: n

o

I9ZH U3 es

ww

I.I

IAd

cEO

-J?

I I Y A'I Ii

09.

I

W4•

o9 __3o 0- Vo d24

AlA

CIC'qoI

COF

ICI

000

UU)

NN

q

LoJ.

-4 43-S !

q

as4

In

I I

ZH 43di OS a,-,

C4

I,

'II"

m

SCA

m V

ZH ui3d os n

1.1.

* ,p

-j

• l • -

3cC')

213

IHS4OS0C4

w.

C-,

-I

-

4c�o-t � (\J -

a,a

ZN �

q

I.e

cc

U L/

Oj~

.4 ~'C7

C4 s4

LA

ZH t13di OS 5 - S

IN

34 O

ZH NJd OS • -

'LJ

C3.

N8

09

C3

444

Ei

43D

cim

(M CO

H Z d OS !9 I I I II I SII

U7

IxICD aI V

0HV4O S

ul

09

W~

CO)

ZH u~j 03 -n a

p

w.

CMD

m l

I==

7 1

-w

w'-

II

0 '3 ' '• •.f '.

F;9?

H=

C,,

* 7

LIC,

* 0 4e

Z •

'4

34

C19~

co2

-ps

(JD

0P4~~

cc

ww

V44

z -

U')3

CD c'N W'

wm

In e

Iff

C-D

oucu

co q.0 t

ZHNe O s

Loil

CD

C.44

m m

zH aad cso s

C4.

Lai-m

1 4 J cv

CD

0z

f4

=HW4C s

Cr%

Fa

th

I-x

ol

cm

cc

a'

OD0

ZIA a3d OS t) - s

cz

m C

ZH 3

OS 0 - •--•.Sd

U-,

If

(e-

LAJ

Qc

Li

ICID

z

Oo

La

H M3

Is

q"

OVCD d-

ZHN4 Se s

III

cc

oww

C4

L~Ln

-\ -

• o

S S S S

p,.I00•I

ZN i .(

iS

2

_'y

r44

E f

*s

- E

Cie

64:

ao'

ZH vd -! LS 6

I

2s.

04

I IQ) fr c

coi

-WC Zr

-A

cx.

Ul

-J-

09.

ci Ira

ZH 434~ OS~ t asA

en

00.

IIv

'-7U

2H W3 OS 0-

0%

B

LaJ �

0.0

a'

____ -w

I,, -

ZH �Jd O� � -

IIk

, w w I I

:1.

C-,

- -

ZH 43-d OS - asd

7

r4rLi~

cmx

oV

r C D

° C9

I-O

-J,•...i••'' '-'-

z,

tod

.i4

ca

ZH M!J34 OS (T -

eo M

clcD

zH Nzd o-- 5~ -

oo

-1

0410I

I --- r,

- M

0 f-

Mok.

ZH 43dJ -s a

IC-

0F

ZH Wd o I K.

v.3

IN

INI

zm A--j s !5-

3cFdo

'I V!z H C3 1 0 Sllý s I

. / :

O-01,'a-

C 4

04

cmJ

a- C

zH v~vie -a

&r4

C4

L)J

t

WH

I I _ _...-:••, ,: ! . ••L

2 j

t C:

/D

ZN?/O

/L

ox~I

Sid

Z" H , : ,

I..

--

.*• '

-4)

• ", j -

3a

cx

Lai

I-Ize

zH ýiidns as

3.

axa

c3.

d.

o t

Lcu

ZH ,

0l

8-7fZIM

-l

Oltl

cc C.

LOD

co ~ l -~~

I 2'

UQ

ct

CDt

ZH 4i3d -21 s

zzrr�c2wT �

N

SH.

4-,

I.4

-4

'4

.4. ('3 4

ZN �JJ 1i O� 0 -

*

Axq

00q

am

H N3 C t

U,

I.

I-4

z

U)

V

I��

cr�'.0 (NI

ZN ZdJd O� � -