1、93FTM8Single Flank Testing andStructure-Born Noise Analysisby: Hermann J. StadtfeldRochester Institute of TechnologyAmerican Gear Manufacturers AssociationTECHNICAL PAPERSingle Flank Testing and Structure-Born Noise AnalysisHermann J. StadtfeidRochester Institute of TechnologyThestatementsand opinio
2、nscontainedhereinarethoseof the authorand should notbe construedas anofficial action oropinion of the American Gear Manufacturers Association.ABSTRACT:Testing therunning behavior of a gearsetthatis ready for installation must takeplace on a bevel andhypoid gear tester.This paper describes singleflan
3、k generation testing and structure-bornenoise analysis of gear pairs based on a highlymodem real-time analysis device for which softwarewas specially developed for the transmission testing of gearsets.The single flank generation test which has been know for decades has experienced a virtual breakthr
4、ough into massproduction through the application of the fast measurement method. The new possibilities and a trouble shootingexample are explained in the paper.Copyright 1993American Gear Manufacturers Association1500 King Street, Suite 201Alexandria, Virginia, 22314October, 1993ISBN: 1-55589-620-0S
5、ingle Flank Testing and Structure-BornNoise AnalysisDr. Hermann J. Stadtfeld, ProfessorRochester Institute of TechnologyRochester NewYork, 146231 Introduction With the pinion driving and the gearbraked,the motion differential (separateThe method of measuring the “quasi- for drive and coast) is formu
6、lated fromstatic“transmissionbehaviorofanygiven the difference between the signals fromhypoid gearset which has been well the pinion and the gears.known for many years is now experienc-ing a renaissance. In addition to the The derivation of the velocities of themeasurement of coordinates for put- pi
7、nion and gear spindles with the result-poses of comparing nominal and actual ing differential development is asimpor-toothsurfaces,thesingleflanktestshould tant asthe relative acceleration inducedenable direct conclusions to be drawn bytooth meshing and isa measure oftheconcerning noise development
8、under ensuing disruptive forces. A Fouriercritical operating conditions. Transformation with the representationof referenced amplitudes shows accel-An appropriate testing arrangement is eration levels overoctave bands, respec-adapted to perform these single flank tively, alineallydepictedfrequencysp
9、ec-tests by fitting high resolution angle in- trum.crement encoder heads to the pinion andgearspindles (e.g.,with 18,000tickmarks A series of gearsets must be tested inon circumference), vehicles in order to determine the rela-tion-ship between noise behavior in theThe time constant T is determined
10、from vehicle and the results of single flanksinusoidal signals bydata measurement testing. Using reference gearsets withand evaluation electronics, known characteristics, i.e., “very good“,1“good“, “acceptable“ and “no longeracceptable“, an interpretation can beachieved which, however, will most cer
11、-tainly vary from one design to another.The recording of vibration noise, respec- _tively, acceleration is an important sup- Jple-ment to rotary transmission meas-urements. It best establishes the rela-tionship between real noise in the vehi-cle and the analysis on the test machine.The relationship
12、between structure-bornenoise, orvibration pattern, and transmis-sion error graphs can only be be estab-lished using the Fourier analysis andfrequency spectrums.Figure 1“ Test machine as basic forFinally, in orderto demonstrate a correc- structure born noise analysisand singletive influence on gear g
13、eometry, the re- flank testing (Source, The Gleasonsuits of a mathematical contact analysis Works)should be available since theoreticalgraphs for rotaryvariations, path ofcon- marks) the exact angle of rotation can betact, and Ease-Off topography are usu- recorded in the immediate vicinity of theall
14、y key elements in the noise optimiza- pinion and gear spindle (Figure 1).tion process.Recording and processing of measure-ment values is performed by compact2 Test machine and test configuration electronicsfromthefirm, CMS, Ettlingen.Both the electronics and the software forThe single flank test is
15、a quasi static measurement signal processing weretesting method. The pinion, therefore, developedespeciallyforoperationwithindrives at a low speed of between 30 thecharacteristicenvironmentofallkindsand 60 rpm. The gear is braked just of gears. This so-called “MessTop“sufficiently to assure constant
16、 tooth package includes a display for graphicsurface contact on the coast and drive outputofthe diagrams generated duringsides, the analysis. With a laser printer con-nected, all graphics can be printed out onThe test machine used might be a paper forarchiving purposes.high precision bevel and hypoi
17、d geartester, or a special single flank testing Using newly developed ring-shaped an-machine. The test machine should have gle increment encoders (Figure 3), thea repeating accuracy of 0.004 ram. by angle measurement takes place at thefitting Heidenhain encoders (18,000 tick leading end of the work
18、spindles. The2advantage ofthis is,aswithAbbescoef- 3 Evaluation of measurement dataficient, the proximity of the object to bemeasured. From the purely mechanical In order to carry out single flank genera-point of view, because the disks (each tion testing a high resolution angle en-with 18,000 tick
19、marks engraved on its coder is adapted to each of the drivingcircumference) are affixed to the spin- and driven gears. The shaft encoderdies and the encoder heads are fitted to signalsofthe drive and driven shafts arethe machine frame, these angle incre- transmitted to the DSPB-2CO counterment encod
20、ers can be operated at any card(onesignaleachperangleencoder).desired speed within the normal opera- Theanglevelocities ofboth spindles(seetional range ofthe test machine (1-3000 Rgure 3) can be determined from therpm). Measurement data input can be time constants of the shaft encoder ira-processed
21、up to a frequency of 500 kHZ. pulses (Figure 2). The velocity functionIt is therefore possible to work with a of the gear spindle is converted to themeasurement speed of up to approxi- pinion spindle (right) by multiplyingwithmately 1400 rpm with the encoders in the inverse ofthe transmission ratio.
22、use. The basis is therefore given whichmakes it possible in principle to perform _;_,_._=_tall_-,w_+-“fast“ single flank testing using the _i+ +_lIF4111_ Idigitalized measurement data process- I _ _ing as summarized in the following _00K.z_. ,00K.zsection. Figure 2 shows the priciple I _.:b_ _._. _-
23、.-1_function of an optical increment encoder. IC_+,.:_-+1%_1 L:_+,._ I20mHzt - _ /_ _ t2oM,zangle increment encoder = t - T, , -T,sec,- T;t T=zeII , ,derdisk il +.,-,o;, +,-i;,o,_ . FF+_l I Zi_Ii=_*lltl_p2tt) I+t_1 tooth I I I feedmarkdigital Signal _ _ +harmonic sensor signal low frequency middle f
24、requency high frequencyUe _ _ uppervoltagelimitup / . mi_,avo,tega,m,tFigure 3: Measurement data processingUa _ lowervoltagelimit0digitalizedrectangularsignal t The shaft encoderimpulses are recordedL_I = T -I with a time resolution of l/20 MHz = 50. _ J_ ns. Bya speedof 1200 rpm atime intervalof 50
25、 ns corresponds to an angle ofFigure 2: Principle function of an enco-der and sensor signal US with time rotation of 6.2 10-6rad.constantT Angle velocity and angle acceleration3are derived from the measured timeval- speed measurement devices, wherebyues. The acceleration difference (in the required
26、encoders and their suppor-Figure 3) is a magnitude for the non- ting electronics are highly expensiveconstant form of the transmission and systems.demonstrates, for example, impulses _ _senresultingfrommeshingimpactasnon- Sensor _constants. The Fourier transformation _ sotdelivers amplitude spectra
27、and makes itpossible to create the relationship withthe structure borne noise analysis, Fig. 5.To record single flank generation errors,DSPB2100 DSPB-4 COthe angle velocity curve must be inte- (option) (option)grated. The difference of the angles of a_w_,el-t ,osstTopI IJportableAT compatiblerotatio
28、n, taking into consideration the ftwara I I 80386computer -transmission, first yields a composite _ 1Integrated Statistical EvaluationDatabankfor IProcessControl I Displaysignal which contains all position fre- ,oasuramantV.u.l (Opten)I Promp,s_quencycomponents. Through appropri-ate filtering it is
29、possible to separate thelong wave (pinion, respectively, gear ro-tation), the medium or short wave portion,shown in Figure 3 bottom. Figure 4: Hard- and software structureDuring measurement, the DSPB-2COcounter card translates the data to the The first, and today, the only digital sys-evaluating com
30、puter. The maximal speed tern ofthistype for industrial series appli-may therefore be: cation istheconfiguration described hereR (l/s) = max. data throughput of corn- with Heidenhain encoders and CMS-purer/number of tick marks MessTop. Resultsfrom high-tech labo-on the shaft encoder ratory systems a
31、re now available fromthe VVZLin Aachen/2/.Given a shaft encoder with 18,000 tickmarks, dynamic measurements are It was expected - especiallyin the lowertherefore possiblewith total time resolu- partial-loadingrange, which is a decisivetion at speeds up to 1400 rpm. The elementinvehiculargearsetnoise
32、-thatthevalidity of this fast single flank genera- credibilityofthe conclusionswould be sig-tion test, asa purely dynamic measuring niflcantlyhigher than previouslybecausemethod, is unquestioned (FFT analysis, both test conditions and real operationalpresentation), conditionswere very similar.A gene
33、ralized process flow diagramm is However, today experience with fast sin-shown in Figure 4. Until now, measuring gle flank generation testing has alreadyin the immediate operational rpm vicinity indicated that an absolute correlationwas reserved strictly for analog angle exists between the two metho
34、ds.4surface structures. Shown in the bottom/IF= ,_,_,_=_ section of the figure, is the theoretical/,Fz graph of the transmission as it wouldrelate to a gearset finished exactly to,_ specification._ Singleflankdeviation(referencedtogear)/2 FZ “_ 2 FZ“_ _ I,. I il, / 4F. longwavepo.ion:- J,lll.,tll.,.
35、,/7 14 2128 16 24 32referencedfrequencyFZ referencedfrequencyFzFigure 5: Amplitude spectrums, indexing _oth11 21 31 4_ _ 6J 711 pinion revolutionerror, left- and tooth mesh error right side.Thehopedforand, interestinglyenough, middle wave portion S y _I tooth,=_ k,/actually occurring phenomenon was
36、themesham plificationof meshingirregularitiesat _“_“_“high speeds. A number oftests showed Jtooth 11 21 31 41 5i 6J 71that the fast singleflank testcausesan an 1 feed mark 1 contactlineamplification of noise determining crite- _ria as the result of the meshing dynam- _/=.-_ics. This leadsto a signif
37、icantincrease in shortwave portion_ ( i ,jthe correlation between the testing ma- _J I theoretical motion graphchine and, for example, the gear box of a Ivehicle. Figure 6: Portions of a single flankToday, it is also possible to say that variationbecauseitcombines ashort testinginter-val of only a f
38、ew seconds with better The evaluation with the equippmentanalysis detail (due to higher rpms), the determine from the measurement valuesfast single flank test is ideally suited for the criteria of the gear quality such asproduction application, indexing jump, cumulative indexing er-ror as well as co
39、ncentric running andForsingle flank error determination, the pitch errors. These magnitudes can beangle speed graph is integrated with the output after every measurement or al-lapsedtime. The difference inthe angles ready queried for “good“ or “reject“ inof rotation yields (considering the ratio the
40、 computer.used as an example), for the time being,the combined signal (Figure 6, middle). 4 Structure-borne noise analysisThe individual tooth meshes can be using seismicsensorrecorded in this graphic. By using appro- This purelydynamicmeasurementprin-priate filters, the low frequency portion(Figure
41、 6, top), e.g., as pinion or gear cipleforthedeterminationoftoothmeshrevolution, or the high frequency portion variations is also used on the test ma-(Figure bottom left), can be identifiedand chine. For this purpose, the housing ofseparated due to clearly recognizable the gear spindle is equipped w
42、ith a seis-5mic acceleration encoder. Here, the ,o,.,.,oo,o._. ,.o.,.,.,o.,.,.,_evaluation electronics is also the CMS- , conventionaldes,gn, lappedMessTop, as with the single flank test, :_however, a few additional hardware op- 0,- , , -_.tions are installed. A fully equipped _1 ,o,_,gog,oun,I “=lt
43、t_,L,t_,ar,.m_HL,hJlel 1 Fz 2 FZ 3 Fz 4 FZ /realtime computer can be used, there- ,0.,o.fore, not only for single flank testing but .,o.bo.also for structure-borne noise analysis. , The required software was specially de- -=_w I klnematlCallyoptlmtZeddOslgn,groundveloped for the evaluation and compa
44、ri- ,0 ,Fz _z _F_ ,_z :son of acceleration signals. _._,_.,.L,.,_._, JL._ .LL_,.JIn order to explain the measurement Figure 7: Results of the structure-bornprinciple and to discuss the output noise analysis of the examined variantsgraphics, the following comparison ofthree significantly different ge
45、ars is pre- tooth indexing errors in the pinion andsented/4/, gear.Aconventional, heat treated and lapped The center section of Figure 7 shows agearset variant is first presented, which ground gearset with conventional Ease-was subsequently ground with identical Off topography. Compared with theEase
46、-Off topography. Additionally, a kine- lapped variant, the side bands of thematically optimized gearset was exam- harmonics have sharply reduced noiseined only in the ground variant. The amplitudes, whilethelevelsexactlyoveraccelerations (respectively, the vibra- the tooth mesh frequency and its mul
47、ti-tions) as well as the Fourier Transforma- pies have actually increased somewhattion from time into the frequency range over those of the lapped version. Sinceare shown as the results, noise appea-rances are almost entirelymissing in the interim frequency bandsIn the right-hand sequence in Figure
48、7, because ofthe high tooth indexing preci-the acceleration amplitudes over time sion, this gearset, subjectively speaking,are displayed. The measurements took sounds louder or less pleasing than theplace atatooth mesh frequencyof200 Hz lapped variant.in the lower partial loading range. Theright-han
49、d sequence in.Figure 7 shows The effects of kinematic optimization arethe noise level in the frequency range. In to be seen in the bottom section of Figurethe upper section of the figure, the re- 7. Because the run-in is corrected, thesuits of the conventional design which noise amplitudes over the tooth meshwasfinishedthrough lapping areshown, frequency and its higher harmonics areThe noise level ofthe firstth reeharmonic substantially reduced. Simultaneously,frequencies is clearly recog
