ULTRASONIC METHOD FOR NONINTRUSIVE LOW-LIQUID-LEVEL SENSING

Transcription

1 ULTRASONIC METHOD FOR NONINTRUSIVE LOW-LIQUID-LEVEL SENSING E. James Ce rn and B. Br Djrdjevic Martin Marietta Labratries Baltimre, Maryland David M. Barnett Martin Marietta Astrnautics Denver, Clrad INTRODUCTION The physics and fundamental principles f ultrasnic pulse ech technlgy are weil understd, and pulse ech techniques have been used widely fr measurement f materials prperties such as mdulus, stress, and thickness, and fr defect detectin and flaw characterizatin [1,2]. We have applied the pulse ech principle t a new area and develped a nnintrusive methd f lw-liquid-level sensing fr ptential applicatin in space systems. The ultrasnic apprach t lwliquid sensing prvides a viable alternative t cnventinal appraches, which require the penetratin f the cntainer wall and special designs t accmmdate the sensr and t access the liquid. Penetratin f the cntainer wall may result in high, lcalized cncentratins f stress, which can cause structural weakness, especially in pressurized envirnments and space applicatins. In this paper, we review the theretical basis fr the pulse ech methd f nnintrusive lw-liquid-level sensing, and present the results f an analysis and several labratry experiments we cnducted t study the sensitivity and functinal relatinship between the signal amplitude and the liquid level ac rss the face f the transducer. The purpse f this study was t verify the perating principle f the system experimentally. The results f the experiments have verified the applicability f this ultrasnic nnintrusive methd f " lw-liquid-level sensing. BACKGROUND AND ANALYSIS The basic structure f the lw-liquid-level sensing system is shwn in Fig. 1. In this case, a pulse is reflected thrugh tw layers [3], the epxy adhesive that bnds the sensr t the cntainer wall and the cntainer wall itself. The acustic reflectin cefficient fr the 1047

2 Sens r 3 Liquid/air Fig. 1. Basic structure f the ultrasnic lw-liquid sensing system. cntainer wall varies depending n whether the liquid backing (i.e., fuel) is present r absent (i.e., air), and thus is used as an indicatr f the level f the liquid backing. The functinal relatinship f time (Tmn ) and acustic amplitude respnse (Amn ), fr such a tw-iayered structure can be written as (1) where t and A mn 2 d/v (d is the thickness and v is the velcity f the media), m+n R 12 * min(m,n) E k=1 (-1) n-k ( ~ ) ( 2) where (:) = m! m and n are numbers f reverberatins in n! (m-n)! medium land 2, T IO is the transmissin cefficient t the transducer AO is the incident amplitude, and Ri. = (Z.-Z.)/(Zi+Z.) is the reflectin cefficient f medium i t3 medidm j [ Z = ~V (the prduct f acustic velcity and density) is the acustic impedancel. Fr ur applicatins, the nly variable is the amplitude f the acustic signal frm the cntainer wall t the fuel r air interface, which is the rati f the nth ech amplitude fr the cntainer wall with a liquid backing (i.e., fuel) and withut (i.e., air). This can be expressed as where R,3 is the reflectin cefficient fr the cntainer wall backed by a liquia, and R'23 is the reflectin cefficient fr the cntainer wall backed by air. As Eq. (3) indicates, the mst sensitive ech t use in such applicatins is t select the highest detectable ech with an apprpriate signal-t-nise rati. (3) 1048

3 x Sensr surface 2r Fig. 2. The uncvered surface area [A(x)] f the transducer as liquid level (x) decreases. We als investigated the sensitivity f the system. As Fig. 2 shws, the sensr face is uncvered as the liquid level drps. Thus, the amplitude f the signal frm Eq. (2) is a functin f the effective ref1ectin cefficient, Reff. Reff is re1ated t R23, R'23' and the liquid level, x, via the percentage f area uncvered, A(x). Fr the first-rder apprximatin, Reff may be written as a simple linear cmbinatin f R23 and R'23 as, Reff = R 23 A(x)+Rz3 (l-a(x)). (4) where A(x) = (ß - sin ß cs ß)/Tl and ß = cs-1 (1 - x/r) fr a circular sensr. CONFIGURATION AND EXPERIMENTS Cnfiguratin Figure 3 is a blck diagram f the u1trasnic nnintrusive 10wliquid-level sensing system. Itrasn ic pul se r/ receivcr Galcd pcak dclcclr Ad hcsi ve Lgic circuilry Tank QUlPUI Fig. 3. Blck diagram f the u1trasnic nnintrusive 10w-1iquid-1eve1 sensing system. 1049

4 The system cnsists f three majr cmpnents: a sensr, an adhesive bnd, and the electrnics. The sensr is a specially designed ultrasnic transducer that can withstand the envirnmental effects likely t be encuntered in space applicatins, such as g-lad, vibratin, and thermal cycles. The adhesive als must meet the adhesin requirements fr space applicatins and have acceptable acustic prperties. We tested adhesives experimentally and selected Versilk 202 t bnd the sensr t the tank. Finally, the electrnics in the sensing system includes a custmized ultrasnic pulser/receiver, accept/reject lgic circuitry, and the desired signal utput. Experimental Prcedures and Results T demnstrate the principle f peratin, we used cmmn cmmercial instruments (1/4-in.-diameter 5-MHz transducers frm varius vendrs and a Snatest UFDS unit as the pulser/receiver), and tw aluminum plates (0.09- and in. thick, respectively), with and withut water backing. The perfrmance f the system strngly depends n the cmbined perfrmances f the ultrasnic pulser/receiver and the transducer. In general, all the transducers we used displayed similar signal respnses but with different amplitudes. Typical ultrasnic respnses with and withut the liquid backing are shwn in Fig. 4. By selecting the mst suitable ech (the highest detectable ech with an apprpriate signalt-nise rati), we can use the ultrasnic signal as a lw-liquid mnitr. We selected the seventh ech. The sensitivity f such a system is related t the sensitivity f the gated peak detectr. T determine the sensitivity f this system, we mnitred the amplitudes f fur echs -- the furth, fifth, sixth, and seventh -- as a functin f liquid level. The nrmalized ultrasnic amplitude vs the nrmalized liquid level fr these echs is shwn in Fig. 5. The general shapes f these curves agree weil with thse determined using Eq. (4). The sensitivity f this experimental setup i8 estimated t be 0.1 in. Fig. 4. (a) Ultrasnic signal respnse (with air backing) (b) Ultrasnic signal respnse (with liquid backing) Typical scillscpe tracings shwing the respnse f the transducer while cvered and uncvered. 1050

5 Fig w c ::::I t: -...J c. 0.6 r- <C == C r- w!:::! 0.4 r-...j <C r- c:: == r- z 00 r- I 0 f x + II x + II II 6, II 4th ECHO + 5th ECHO x 6th ECHO 7th ECHO I L I ~ I I. I I NORMALIZED LIQUID LEVEL Nrmalized ultrasnic amplitude vs nrmalized liquid level fr echs Finally, we perfrmed independent experiments t determine the ultrasnic signal respnse f varius adhesives. The results fr tw f these adhesives, Versilk 202 and 204, are prvided in Fig. 6. As this shws, Versilk 204 has better acustic prperties than Versilk 202. Hwever, Versilk 204 requires 24 h t be fully cured, whereas Versilk 202 is cured after nly 3 min. Since the signal amplitude can easily be adjusted t meet the design requirements using the electrnic gain circuitry, Versilk 202 was selected fr its ease f bnding. 1.0 ~.8 Versilk 204 ~.6 :;;! Versilk 202 Z 9 CI).4 ~ j Z Fig. 6. TIME (min) Ultrasnic amplitude respnse as a functin f curing time fr tw adhesives, Versilk 202 and

6 DISCUSSION AND CONCLUSION The results frm these labratry experiments have clearly verified the perating principle fr lw-liquid-level sensing. The next step is t establish prcedures regarding: 1) sensr variatin, 2) adhesive cure mnitring, 3) signal nrmalizatin, 4) envirnmental effects, and 5) functinality checks. Each issue is discussed separately belw. Manufacturing sensrs is a very cmplicated prcess and difficult t cntrl, and thus the perfrmance f the sensrs may vary drastically. A test prcedure, such as measuring the reflected amplitude frm an aluminum plate thrugh a fixed-distance water path, shuld be established t standardize the perfrmance f the sensrs. A prcedure is needed t determine the prper acustic pattern f an acceptable bnd during the curing prcess. The sensr is wasted if the bnd is rejected after the adhesive is cured because the sensr is bnded and cannt be remved. Once the prper pattern is established, the bnding prcess can be abrted during curing if the signal respnse des nt meet the criteria, and the sensr can be salvaged. The perfrmance f the entire lw-liquid-level system depends n the cmbined perfrmances f its cmpnents, the sensr, the adhesive bnd, and the electrnic circuitry. A methd f gain adjustment shuld be incrprated int the system's electrnic circuitry t cmpensate fr variatins in the signal caused by variatins in the sensr and the adhesive bnd. Temperature, humidity, and adherend surface preparatin dictate the adhesive curing time and influence the level f acustic respnse attained. Acceptable envirnment al cnditins fr bnding the sensr must be defined t ensure prper installatin and thus perfrmance cmparability. Finally, the system must be peridically mnitred during installatin and while in service t ensure that it is functining prperly. An array f LEU indicatrs shuld be included n the electrnic panel fr mnitring signal utput. In cnclusin, we have reviewed and verified the theretical principles applied t an ultrasnic lw-liquid-level sensing system. The results frm the labratry experiments and a cmputer analysis have demnstrated the validity f the apprach and the applicability f the ultrasnic principle fr lw-liquid-level sensing. In future wrk, we shall investigate the practical cnsideratins listed abve t ptimize field installatin and in-service mnitring. REFERENCES 1. R.C. McMaster, Nndestructive Testing Handbk (Rnald Press, New Yrk, 1959). 2. J. Krautkramer and H. Krautkramer, Ultrasnic Testing f Materials (Springer-Verlag, New Yrk, 1983). 3. L.M. Brekhvskikh, Waves in Layered Media (Academic Press, New Yrk, 1960). 1052