Fault Detection of Aircraft Cable via Spread Spectrum Time Domain Reflectometry

Transcription

1 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp Sensors & Transducers 4 by IFSA Publshng, S. L. Detecton of Arcraft Cable va Spread Spectrum Tme Doman Reflectometry, Xudong SHI, 3 Xaohu LI, 3 Zhangang YANG, 3 Chengzong LIU, 3 Hongguang LI Tanjn Key Laboratory for Cvl Arcraft Arworthness and Mantenance of Cvl Avaton Unversty of Chna, Tanjn, 33, Chna Ground Support Equpments Research Base of Cvl Avaton Unversty of Chna, Tanjn, 33, Chna 3 Aeronautcal Automaton College of Cvl Avaton Unversty of Chna, Tanjn, 33, Chna E-mal: stone_3@sna.com, cauclxh@qq.com Receved: 4 March 4 Accepted: 9 March 4 Publshed: 3 March 4 Abstract: As the arplane cable fault detecton based on TDR (tme doman reflectometry) s affected easly by varous nose sgnals, whch maes the reflected sgnal attenuate and dstort heavly, falng to locate the fault. In order to solve these problems, a method of spread spectrum tme doman reflectometry (SSTDR) s ntroduced n ths paper, tang the advantage of the sharp pea of correlaton functon. The test sgnal s generated from ML sequence (MLS) modulated by sne wave n the same frequency. Theoretcally, the test sgnal has the very hgh mmunty of nose, whch can be appled wth excellent precson to fault locaton on the arcraft cable. In ths paper, the method of SSTDR was normally smulated n MATLAB. Then, an expermental setup, based on LabVIEW, was organzed to detect and locate the fault on the arcraft cable. It has been demonstrated that SSTDR has the hgh mmunty of nose, reducng some detecton errors effectvely. Copyrght 4 IFSA Publshng, S. L. Keywords: SSTDR, Correlaton functon, Immunty of nose, detecton, LabVIEW.. Introducton As s nown, arcraft cable s playng a more and more mportant role n flght safety. Wth the long-term flght, the arcraft may suffer the problems of cable agng, and the arcraft cable problems have been consdered to be the man cause of many dsasters. What s more, varous of faults may occur to the arcraft cable because of the contnuous shae and humd envronment. Whle the plane s n flght, the cable falure s lely to result n catastrophc destructon to the arcraft. Thus, the detecton and localzaton of faults wth hgh accuracy s requred strongly for dagnoss and mantenance of wrng systems. Although the applcaton of tme doman Reflectometry (TDR) s utlzed wdely, there are stll some problems, such as all nds of nose sgnals, whch wll result n the superposton of a varety of sgnals ncludng some nose sgnals, brngng the dffculty for the extracton of the reflected sgnal. Meanwhle, the degree of attenuaton of the reference sgnal ncreases gradually wth the ncrease of the propagaton dstance. Therefore, t s dffcult to dentfy the reflected sgnal, resultng n great errors to the measurement results []. Artcle number P_RP_89 35

2 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp For the above reasons, an expermental setup, based on LabVIEW, was bult up for the arcraft cable fault detecton. The test sgnal, wth hgh nose mmunty of nose, has the sharp pea of correlaton functon, whch maes t well suted to detect and locate some fault on the cable, reducng the detecton errors greatly [-4]. SSTDR can help mantenance personnel detect fault type and locate the fault on the cable qucly, and mprove the detecton effcency for arcraft cable fault.. Prncple Analyss.. The Prncple of Detecton SSTDR s a method of fault detecton based on spread spectrum technology [5]. The dagram of SSTDR cable fault detecton s shown n Fg.. In ths applcaton, the MLS s modulated by sne wave n the same frequency, generatng the test sgnal st () for SSTDR. The correlaton between the ncdent and reflected sgnals dentfes the type of fault and ts locaton. In the begnnng, the test system wll send the sgnal st () onto one end of the cable, whch wll be reflected by some arbtrary number of mpedance dscontnutes on the cable. The reflected sgnal rt () = ast ( wll return to the test system after some transmsson delay. Usually, along wth the reflected sgnal wll be some nose sgnals. Cross-correlate the total sgnals wth the reference sgnal delayed by τ, and by changng the delay tme τ, to mae cross-correlator output maxmum value. Accordng to the feature of crosscorrelaton, at ths pont, there s τ = t t s the, reflecton tme correspondng to the cable fault pont. If the velocty of the sgnal s nown, multply t wth the tme t, the fault pont can be obtaned. What s more, the system can tell the dfference between the open-crcut fault and short-crcut fault by analyzng the pea of cross-correlaton. When the second pea s rght-sde-up, the fault wll be dagnosed wth open-crcut fault. On the contrary, the second pea s nverted, then short-crcut fault wll be confrmed [6]. sn( ω Fg.. Dagram of SSTDR cable fault detecton. Assume the velocty of the sgnal s constant n the cable,.7 tmes the speed of lght. If the propagaton delay τ of the reflected sgnal s got, the fault locaton L can be obtaned from the followng formula [7]: where c s the speed of lght. L=.7 c τ, () MLS s shortened from the largest lnear feedbac shft regster sequence. N-lnear shft regster s shown n Fg., and the correspondng characterstc polynomal equaton s expressed n formula (), where x has no practcal sgnfcance, whose coeffcent s or. x x 3 x x n.. Desgn of Reference Sgnal c c c c 3 c n The optmal test sgnal depends on the nature of the applcaton. For sngle cable, we select MLS, modulated by snusodal sgnal n the same frequency, as the test sgnal for SSTDR. The autocorrelaton of MLS has obvous pea value and has the lowest sde lobes, so the reflected sgnal can be dentfed easly. It s, therefore, optmal for detectng the fault of sngle cable. Fg.. N-lnear Shft Regster. n n n n = f( x) = c + c x+ c x +Λ+ c x = c x ( c = c = ), () 36

3 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp N-lnear shft regster conssts of n states, n and the length of the sequence s N =. The shft regster s controlled by an external cloc, n accordance wth the cloc shft output. As shown n Fg. 3(a) s the 3 bts of MLS smulaton waveform. In ths paper, the test sgnal s generated from MLS modulated by sne wave n the same frequency. That s to say, sne wave remans prmtve phase when the MLS appeared hgh, whle nverted when low phase [8]. The smulaton waveform of MLS, 3 bts, modulated by sne wave s shown n Fg. 4(a). The modulated MLS help to mprove the SNR (Sgnal Nose Rato) of correlaton, mae the system dentfy the useful sgnal easer, and have the pea of correlaton be more sgnfcant, mprovng the ablty to resst the nterference of nose. The 3 bt of MLS and ts modulated waveform are separately smulated n Fg. 3(a) and Fg. 4(a), whose normalzed autocorrelaton waveform are shown n Fg. 3(b) and Fg. 4(b)..5 Ampltude(V) Tme(ns) x 4 (a) MLS smulaton waveform (a) Desgn of program for test sgnal n LabVIEW Normalzed Auto-correlaton(V) Tme(V) x 4 (b) Normalzed auto-correlaton of MLS Ampltude(V) Normalzed Auto-correlaton Fg. 3. MLS and ts Normalzed auto-correlaton smulaton waveform Tme(ns) x (a) Modulated MLS smulaton waveform Tme(ns) x 4 (b) Normalzed auto-correlaton of modulated MLS. Fg. 4. Modulated MLS and ts Normalzed auto-correlaton smulaton waveform. (b) The test sgnal waveform n LabVIEW Fg. 5. The test sgnal waveform and ts Desgn of program n LabVIEW..3. Correlaton Algorthm Correlaton functon algorthm s the core of the arcraft cable fault detecton system, whch reflects the dependency of the test sgnal and reflected sgnal. When the test sgnal and the reflected sgnal are dentcal n phase, a unque maxmum wll be output. Usng correlaton functon, the system can dentfy the tme dfference between the test sgnal and the reflected sgnal, then the cable fault locaton can be obtaned [9]. Snce the reference sgnal s so low, the test sgnal, because of attenuaton, becomes qute wea when t s reflected bac agan from the fault pont.what s worse, the reflected sgnal, usually nterfered by some nose sgnals, wll suffer serous deformaton. Through the correlaton algorthm, the system can mae the reflected sgnal, even f submerged n the 37

4 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp nose, be dentfed easer, mprovng the antnterference ablty of SSTDR. Let rt () be the receved sgnal from the cable fault pont, defned as follows: rt () = ast ( τ ) + nt (), (3) where nt () s the nose sgnal, a s the ampltude of reflected sgnal a st ( τ ) relatve to st () and τ s the tme delay before recevng reflecton. By shftng the reference sgnal st () wth τ, cross-correlate t wth the reflected sgnal, and the cross-correlaton can be expressed as: R = = sr ( τ ) = r( t t ) dt t t ) a t τ ) dt + a t τ ) + n( t t ) dt t t ) n( dt, (4) where t s the estmated tme delay of the test sgnal. The mathematc expectaton value E()of the cross-correlaton R sr ( τ ) s consdered as: = { ( τ) } E R sr = E t t ) T S t t ) a t τ ) dt+ a t τ ) dt t t ) n( dt, (5) Snce the nose sgnal nt () s duraton Tn >> Ts, the reference sgnal st () and nt () are asynchronous, that s, nt () s statstcally uncorrelated to st (). So the correlaton of st () and nt () s nearly equal to zero and the formula (5) can be obtaned. It can be concluded that although the reflected sgnal s the superposton of the njected sgnal and nose sgnal, however, st () and nt () are not synchronous, so that the cross-correlaton of them s near zero. That s to say, the nose sgnal nt () has nearly no effect on the cross-correlaton of sgnal. It s also nown that the cross-correlaton of the reflected sgnal and reference sgnal, just when they are dentcal n phase, has only one pea. The local pea tme of the cross-correlaton functon, τ = t,wll be utlzed to accurately measure the propagaton delay of the reflected sgnal, whch s then to be converted nto the fault locaton wth nowledge of the velocty of propagaton []. In the mathematcal theory, t has been proved that the correlaton method has the strong ablty to flter out the nose. The SSTDR test sgnal, nterfered by the whte nose, and ts normalzed auto-correlaton functon smulaton waveform are shown n Fg. 6. Ampltude(V) Normalzed Auto-correlaton(V) Tme(ns) x (a) SSTDR test sgnal nterfered by whte nose Tme(ns) x 4 (b) Normalzed Auto-correlaton of SSTDR sgnal nterfered by whte nose Fg. 6. SSTDR sgnal nterfered by whte nose and ts normalzed auto-correlaton smulaton waveform. Because of the nose, the SSTDR test sgnal results n serous dstorton, but ts auto-correlaton has stll obvously sharp pea. That s to say, the nose has no effect on the result of auto-correlaton. Theoretcally, t proves the feasblty of the test sgnal wth hgh mmunty of nose sgnals n the detecton of arcraft cable fault..4. Techncal Analyss Accordng to the correlaton functon theory, the test sgnal st (), whose auto-correlaton has a certan perodcty, when more than one cycle, there wll be more than one auto-correlaton peas, whch may be confused wth the cross-correlaton of the cable fault pont, resultng n msjudgment. So the correlaton must be completed n a cycle tme. However, the fault locaton cannot be acheved f the fault pont s too close to the test end, wth the superposton the st () auto-correlaton and the cross-correlaton. So the range of the cable length, to be tested, must meet the followng expresson: T c V L N Tc V, (6) where V s the velocty of sgnal on the cable, about.7 tmes the speed of lght; N s the length of MLS. The followng formulas can be got from (6): The mnmum detectable dstance: L mn = T c V, (7) 38

5 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp And the maxmum detectable dstance: Lmax = N Tc V, (8) When the velocty of propagaton V s constant, the range resoluton can be obtaned: Δ L = τ t V, (9) So, the resoluton s consdered to be closely related to the tme delayτ. reference sgnal and reflected sgnals, and obtans the round-trp propagaton tme to calculate the locaton of fault. 3.. Algorthm Process In the organzaton and management of the PXI bus controller, the test sgnal wll be generated by the arbtrary waveform generator, then the computer complete the sgnal processng, along wth the termnal dsplay functon. The man algorthm process chart s shown n Fg Expermental Setup 3.. Equpment Components To demonstrate the ablty of SSTDR to detect and locate short-crcut and open-crcut types of fault on arcraft cables, an expermental SSTDR system was organzed as shown n Fg. 7. The system conssts manly of the PXI bus controller (computer for expermental SSTDR), arbtrary waveform generator, dgtal osclloscope, and a T connector. Fg. 8. Dagram for the man algorthm process. 4. Result and Analyss Fg. 7. Expermental setup for SSTDR. To automatcally control the nstruments, the arbtrary waveform generator (utlzed to generate some certan frequency of SSTDR test sgnal) and dgtal osclloscope are connected to the PXI bus controller. The test sgnal s transmtted down the target cable va one port of the T connector. If there s a fault n the cable, because of the change of mpedance, the test sgnal wll be reflected to the dgtal osclloscope va the other port of T connector. The computer controls and synchronzes the arbtrary waveform generator and dgtal osclloscope, executes the cross-correlaton of the In order to verfy the SSTDR smulaton characterstcs, the present experments were conducted to detect short-crcut and open-crcut faults for dfferent lengths of arcraft cable. Also note that the samplng rates for reflected sgnal acquston of the SSTDR s GHz. Let the ncdent sgnal chp rate s T c =/3 MHz, and the ampltude of the ncdent sgnal s set to V. The length of MLS s N=63. Accordng to the expresson (6), t can be nown the range of the cable length whch can be tested: 3.33 m L m. Because of the low ampltude, the maxmum measurable length may not reach the theoretcal results. The reflected sgnal waveform for open-crcut fault of 35 m cable s shown n Fg. 9 (a). It can be nown the reflecton sgnal has some certan attenuaton n the ampltude, wth the tme delay compared wth the reference sgnal. Through 39

6 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp correlaton algorthm, the normalzed correlaton waveform s obtaned n Fg. 9 (b). The estmated tme of the frst pea, whch s the reference sgnal tself, s ns, whle the estmated tme of the second pea (cross-correlaton of the reference sgnal and the reflected sgnal) s about ns. Wth the sgnal velocty of propagaton on the cable (V=. 8 m/sec), the locaton of the fault can be estmated to be m. Whle the second pea s rght-sde-up, so t s dagnosed wth an open-crcut fault type. (a) The reflected sgnal for short-crcut (a) The reflected sgnal for open-crcut (b) The Normalzed Correlaton for short-crcut Fg.. The Reflected sgnal and Normalzed Correlaton waveform for short-crcut. Table. Relatve Errors. Opencrcut Shortcrcut m 5 m 35 m 43 m. %.6 %.46 %.44 %.7 %.64 %.54 %.49 % (b) The Normalzed Correlaton for open-crcut Fg. 9. The Reflected sgnal and Normalzed Correlaton waveform for open-crcut. As shown n Fg. (a) s the reflected sgnal of short-crcut fault, nverted n phase compared wth Fg. 9(a). The second pea s also nverted n ts normalzed correlaton shown n Fg. (b). Afterwards, some dfferent lengths of arcraft cables were selected for open-crcut and short-crcut faults test, and the test results are shown n Table. For a more graphcal comparson of the results, Table lsts the relatve errors of the measurement results for dfferent lengths of arcraft cables. Opencrcut Shortcrcut Table. Results of the cable locaton. m 5 m 35 m 43 m.9 m 4.9 m m 4.85 m.88 m 4.94 m 34.9 m 4.86 m From Table and Table, t can be seen that SSTDR method has the good accuracy n locatng the arcraft cable fault, showng ts strong fault detecton capablty. 5. Concluson In ths paper, short-crcut and open-crcut faults are detected accurately, usng the method of spread spectrum tme doman reflectometry. The MLS, modulated by the sne wave, wth the characterstcs of sharp correlaton pea and hgh mmunty of nose, was selected as test sgnal. The expermental results proved the feasblty of SSTDR for arcraft cable fault detecton. Moreover, SSTDR wll play the more and more part role n detectng faults n future, especally, arcraft cable fault detecton on-lne. The low test sgnal levels and hgh nose mmunty of these test methods mae them well suted to test for ntermttent wrng falures such as open crcuts, short crcuts, and arcs on cables n arcraft n flght. 4

7 Sensors & Transducers, Vol. 67, Issue 3, March 4, pp Acnowledgment Ths wor s supported by Jont Funds of the Natonal Natural Scence Foundaton of Chna and Cvl Avaton Admnstraton of Chna Key Project (Grant No. #U33); The Innovaton Fund project of Natonal Commercal Arcraft Manufacturng Engneerng Technology Research Center(Grant No. #SAMC3-JS-5-7); Key Project of Tanjn Key Technology R&D Program (Grant No. #ZCKFGX4); The Fundamental Research Funds for The Central Unverstes (Grant No. #ZXHB & No. #33P5). [4]. [5]. [6]. [7]. [8]. References [9]. []. C. Furse and R. Haupt, Down to the wre: The hdden hazard of agng arcraft wrng, IEEE Spectrum, Feb., No., pp []. Smth P. S., Sequence and spread spectrum tme doman Reflectometry, Utah State Unversty, 3. [3]. Nsh E. Y., Nshjma C., Kuwanam K., A method for fault detectng on twsted par cable networ by use of M sequence correlaton, The 3th Annual []. Conference of the IEEE Industral Electroncs Socety, Busan, Korea, 4,, pp C. Furse, P. Smth, M. Safav, and C. Lo, Feasblty of spread spectrum sensors for locaton of arcs on lve wre, IEEE Sensors, Dec. 5, 5, 6, pp V. Taylor and M. Faulner, Lne montorng and fault locaton usng spread spectrum on power lne carrer, Proc. Inst. Elect. Eng., Vol. 43, Sep. 996, pp P. Smth, C. Furse, J., Gunther. Analyss of spread spectrum tme doman reflectometry for wre fault locaton, IEEE Sensors, 5, 6, 5, pp M. Schmdt, Use of TDR for Cable Testng, M.S. Thess, Dept. Elect. Comput. Eng., Utah State Unv., Logan, UT,. P. Smth, Spread spectrum tme doman reflectometry, PhD Dssertaton, Dept. Elect. Comput. Eng., Utah State Unv., Logan, 3. F. J. Mendeta and A. Trevno and C. A. Martnez, Complementary Sequence Correlatons wth Applcatons to Reflectometry Studes, Journal of The Mexcan Socety of Instrumentaton 3, Aprl 996, pp Chen We, Wang L, The study of spread spectrum tme doman reflectometry for cable fault detecton and locaton on-lne, n Proceedngs of the Internatonal Conference on Electrc Informaton and Control Engneerng, ICEICE,, pp Copyrght, Internatonal Frequency Sensor Assocaton (IFSA). All rghts reserved. ( 4