Aero Acoustic Noise Analysis of a Locomotive Cooling System Ducts and Structure Optimization

Transcription

1 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp Aero Acoustc Nose Analyss of a Locomotve Coolng System Ducts and Structure Optmzaton Shan-Shan L 1,*, Mng L 1,, Fan Yang 3, Jun-Fang L 1, Kan Wang 1 1 College of Automotve Engneerng, Jln Unversty, Changchun 13005, Chna. State Key Lab. of Automotve Smulaton and Control, Jln Unversty, Changchun 13005, Chna. 3 Dalan Locomotve and rollng stoc co., Ltd. CNR group, Dalan 1160, Chna. Receved 31 March 015; receved n revsed form 03 August 015; accepted 06 August 015 Abstract Aero acoustc nose of a locomotve cab coolng system ducts was analyzed by method of Computatonal Flud Dynamcs (CFD) and Computatonal Aero acoustc (CAA) approach. Flow characterstc of the ducts was analyzed by CFD software, then near-feld and far-feld aero acoustc nose was forecasted wth BNS model and FW-H model respectvely. Duct structure was optmzed accordng to the analyss of flow feld and sound feld. Results ndcated that nose characterstc of senstve frequency band at the poston of human ear wth the optmzed duct has a sgnfcant mprovement. Keyword: Flow feld analyss, Aero acoustc analyss, BNS model, FW-H model 1. Introducton Recently, locomotves tae an mportant role n modern transportaton. Le automotve, aero acoustc nose outsde hgh-speed tran s sgnfcant, aroused the attenton of scholars at home and abroad. Sha Huang [1] found that the curvature of tran head surface has sgnfcant nfluence on surface fluctuatng pressure, aero acoustc nose can be reduced by changng the head desgn of streamlne shape. J. Munoz- Panagua [] mproved the tran aerodynamc by changng tran head parameters, thereby reducng the aero acoustc nose. These researches are focused on tran head shape parameters, they only have connectons wth tran external aeroacoustcs. Due to the poor sealng of the locomotve cab operatons console, hgh-speed ar flows nto cab through the gap at the operatons console, combned wth aero acoustc nose from coolng system, worsened ndoor ar nose envronment of locomotve cab. Poor nose envronments not only mae drvers feel uncomfortable, but also dstract them, or even worse, reduce drvng safety and endanger the lves of drvers and crews. It s very mportant to research and control the locomotve cab nose envronment. Predctve tools for the aero acoustc behavor of components, subsystems and whole systems early on n the desgn phase are therefore desrable. Wth the development of computatonal flud dynamcs and aero acoustcs, the study on flud aerodynamc nose wth the numercal method has become possble. Currently, research method of flud aerodynamc nose manly ncludes drect nose smulaton, sound analogy, broadband nose source model method, CFD software and professonal acoustc code couplng method. The ablty of an analyss tool to characterze a system s level of nose and ts spectral dstrbuton at a recever locaton s a powerful desgn tool. It could be used as a vrtual test lab to measure whether a gven * Correspondng author. E-mal: lss13@mals.lu.edu.cn

2 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp system meets ts desgn requrements. Currently, such capablty s not avalable wthn a development cycle for complex ndustral applcatons. Ths s due to the fact that nose nformaton at recever locatons requres the soluton of both the flow feld and the acoustc feld. However, durng early stages of the product development, nose nformaton at recever locatons s not crtcal. Rather, relable system nose nformaton at the source may be suffcent to provde desgn drecton durng development. Nose nformaton at the source requres only a CFD soluton that captures unsteady flow structure. At early stages of the development cycle, a CFD tool that provdes a measure of nose levels at the source for a gven system n order to provde a desgn drecton would be very useful. In addton, f that tool s able to dentfy components or surfaces that generate most of the nose t would provde, to the development team, a valuable means to derve desgn optmzaton. In ths paper, a numercal smulaton on a locomotve cab was performed to analyze the dstrbuton of the flow feld. Then the nose predcton was performed on the bass of the flow feld results to fnd out the dstrbuton of the nose source wth broadband nose source model and the radaton characterstcs of the nose source wth FW-H sound analogy method. At last, optmzaton of the duct structure was performed, whch provded a theoretcal bass for the cab's acoustc envronment mprovement.. Model and Boundary Condton Here, a case wll be studed where the ntal level desgn of a locomotve cab coolng system duct dd not meet ts desgn requrements due to hgh levels of NVH, poor overall sound qualty, and poor arflow dstrbuton of outlet. The three-dmensonal model of the coolng system duct was created wth CATIA software, as shown n Fg.1 (a). The duct model grds were meshed by polyhedral mesh for ts geometrc adaptablty, and the ar nlet and ar outlet were refned accordngly. Extruder mesh was selected at duct nlet and outlets to ncrease flud stablty and avod unnecessary reverse flow, adverse to flow feld analyss results, n turn affectng the sound feld analyss result correctness. The fnal mesh model was shown n Fg.1 (b). (a) Geometrc model (b) Model grds Fg. 1 Duct sold model The duct nlet was set as mass flow nlet, and the value of mass flow was 0.3 Kg/s, accordng to the desgn specfcatons. Because the duct outlet s vented to the atmosphere, the duct outlet was set as pressure outlet. 3. Flow Feld Calculaton 3.1 Flud feld analyss model Flow feld analyss was based on the Reynolds-averaged Naver-Stoes model (RANS). The Realzable K-E model was to solve the N-S equatons, to obtan the average flow feld nformaton throughout the process. A formula for turbulent vscosty and a transport equaton for turbulent dsspaton rate are added to the Realzable K-E Copyrght TAETI

3 180 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp model, mang ths model has a good performance n analyss on flow separaton and complex secondary flow. Based on the theory of computatonal flud dynamcs analyss [3], the transport equatons for the realzable K-Epslon model are shown by Eqs. (1) and (): d dt V t dv ( v vg ) da ( ) da G Gb (( 0 ) M ) S A A V dv (1) d dt V dv A t ( v vg ) da ( ) da C1S ( C1C 3Gb ) C ( 0 ) S dv v A V () where S κ and S ε are the user-specfed source terms, W/m 3 ; Ɛ 0 s the ambent turbulence value n the source terms that counteracts turbulence decay, J/g; S s the modulus of the mean stran rate tensor, ɤ M s the dlataton dsspaton, m /s 3 ; μ s the dynamc vscosty, m /s; μ t s The turbulent vscosty, m /s; C ε1, C ε and σ κ are model coeffcents, G b and G κ are turbulent producton, J/g. 3. Flow feld results analyss Fg. Duct baffle area velocty cloud Fg. 3 Duct statc pressure contours and streamlne Fg. 4 Duct turbulent netc energy contours It can be seen from Fgs. - 4 that the maxmum velocty place located at baffle expanson area, and the maxmum velocty value was m/s, the statc varaton ampltude was Pa, the maxmum turbulent netc energy was J/g. The Mach number of the duct s less than 0.3, for the maxmum velocty s less than 10 m/s, t can be nferred that the dpole nose was the man nose source. The duct streamlne ndcated that arflow dstrbuton of duct outlets s poor, t s also a factor affectng the duct aeroacoustcs. Fgs. - 4 show that the hgh-speed flud flows nto duct vertcal secton through the trapezod body, and hts the surface of the duct at the corner, whch results n the flud velocty changes, and dsturbances get severe, and turbulent netc energy ncreases, and the statc pressure ncreases. When flud reaches the baffle area, the velocty of flud around the corner ncreases, Copyrght TAETI

4 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp negatve pressure regon emerges, and vortex forms, so that the unevenness of the dstrbuton of the nternal flud duct phenomenon s more obvous, and the flow dsturbance and turbulent energy dsspaton are more serous. Severe flud dsturbance ncreases flud flow resstance, whch causes energy loss, wheren part of the energy wll propagates outward n the form of acoustc energy, Jan-Peng Yao sad n hs paper [4], and generates nose. 3.3 Near-feld nose predcton model The nose source predcton was made wth broadband nose source models (BNS) whch has been recognzed by domestc and foregn scholars le Zheng-Yu Zheng [5], A. Zanon [6] and Fred Mendonca [7]. Broadband nose source model s manly dvded nto four models: (1) Curle nose source model The Curle Nose source model evaluates the nose from a turbulent boundary layer flow over a sold body at low Mach number. The model represents dpole sources of nose comng from the fluctuatng surface pressure from sold boundares actng on the flud. Specfcally, ths model computes the surface acoustc power to evaluate the local contrbuton to the total acoustc power per unt area of the body surface. C. Schram [8] showed the Curle surface ntegral by eq. (3): 1 px, t 3 4a 0 S x y r p y t t, r a 0 nds y (3) r where t s the emsson tme(s), p s the surface pressure, (Pa), ṕ s the acoustc pressure, (Pa), a 0 s the far-feld sound a 0 speed, (m/s). () Llley Nose source model The Llley equaton computes the quadrupole nose from the fluctuatng velocty flow feld. Applcatons nclude problems where local shear effects domnate, such as flow around sold bodes and shear flows n channels. Davd W. Wundrow [9] showed the Llley equaton for problems wthout heat transfer as eq. (4): d dt d a dt x u x x x a u x x u x u x (4) where a s the speed of sound, (m/s), s the stress tensor. (3) Lnearzed Euler Equaton nose source model The Lnearzed Euler Equatons (LEE) account for refracton and convecton effects n any sheared mean flows. It can be used to compute the quadrupole nose for a wder range of condtons than the Llley equaton, n whch the assocated source term s a nonlnear functon of the fluctuatng velocty flow feld. Tauya Oshma Imano [10] showed the LEE equaton as Eqs. (5)-(7): p ( x, P ( x) p( x, p ( x, a (5) Copyrght TAETI

5 18 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp u ( x, U ( x) u( x, u ( x, (6) a ( x, ( x) ( x, ( x, a (7) where the acoustc components p a, u a and ρ a are small compared to the mean components P, U and and the turbulent components p, u and. (4) Proudman nose source model The Proudman nose source model evaluates acoustc power per unt volume, and the sound s from quadrupoles. Specfcally, ths model computes the acoustc power to evaluate the local contrbuton to the total acoustc power per unt volume from the turbulent flow. Ths model has been used for rotatng parts, heat exchangers and dstrbuton ducts. S. Sarar [11] showed the local acoustc power generated by unt volume of sotropc turbulence by Eq. (8): u AP 0 l 3 u a (8) where α s a constant related to the shape of the longtudnal velocty correlaton, u s the root mean square of one of the velocty components, (m/s) l s the longtudnal ntegral length scale of the velocty, (m), ρ 0 s the far-feld densty (g/m 3 ), and a 0 s the far-feld sound speed, (m/s). 3.4 Near-feld nose predcton results Curle nose source model was chosen to analyze duct surface dpole nose, Proudman nose source model was chosen to analyze duct volume quadrupole nose for the reason that t has been well used n dstrbuton duct applcatons and t can be easly recognzed by usng unt db, not /s 3 for Llley and m/s for LEE. These two models based on Lghthll s analogy, requre lttle computatonal effort n extractng, from a steady state CFD soluton, the Acoustc Power (AP) and the Surface Acoustc Power (SAP). Both aeroacoustc varables are accessed wthn the Fluent Solver as an aeroacoustc post-processng tool. AP provdes Proudman s quadrupole nose generaton, whereas SAP provdes Curle s surface dpole nose generaton. These two aeroacoustc varables are well suted to Ar Handlng Subsystem (AHS) applcatons whch represent a complex nternal flow that ncludes rotatng parts (blowers), heat exchangers, and dstrbuton ducts. In these low Mach number flows, surface dpole nose radaton generally contrbutes most of the system nose. The locaton and strength of nose sources are shown as Fgs Fg. 5 Duct surface dpole nose source dstrbuton Fg. 6 Duct baffle area quadrupole nose source dstrbuton Copyrght TAETI

6 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp Fg. 7 Duct vertcal secton quadrupole nose source dstrbuton Fgs. 5-7 show that duct vertcal secton and baffle areas were man dpole nose sources and quadrupole nose sources, the dpole nose at these two places were db and db, the quadrupole nose at these two place were about 70 db and 65 db. The quadrupole nose was largely less than dpole nose, t can be concluded that the dpole nose s the man nose source. Meanwhle, these places are also the turbulence areas from results of flow feld. Ths ndcates that dsturbance areas propagate acoustc energy. BNS models do not provde any tonal nose nformaton or nose spectra at recever locatons. Instead they provde only the locaton of nose source and an approxmate measure of the radated nose at the source. For quantfyng the result, further analyss should be performed to obtan nose spectra at recever locatons. All prevous approaches requred a transent CFD soluton that captures the flow structure and presents some challenge n terms of resource requrements and soluton tme. Although the coupled approach s wthn reach for automotve applcatons, t cannot be ncorporated, at the present, nto a fast paced development process. 4. Unsteady state calculaton 4.1 Tme step Unsteady state calculaton was carred out after steady state calculaton gets stable, and the large eddy smulaton model (LES) was used to solve the flud nstantaneous nformaton. The tme step can be calculated as equaton (9): f 1 1 t (9) where f s the frequency expected, Hz; Δt s the tme step, s. Tae 5000 Hz as the frequency n ths paper, then the tme step s s. Based on the theory of computatonal flud dynamcs analyss [3], the transport equatons for the LES model are shown by Eqs. (10) and (11): u t x 0 (10) Copyrght TAETI

7 184 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp p ( u ) ( uu ) ( ) t x x x x x (11) where u and u are the fltered velocty component, (m/s), μ s the turbulent vscosty, (m /s), p s the pressure, (Pa), σ s the vscous stress tensor (Pa), τ s the sub grd stress, (Pa). 4. Far-feld nose predcton model Sound analoges are based on a reformulaton of the basc governng flow equatons, the compressble Naver-Stoes equatons. Rearranged, these equatons yeld an nhomogeneous wave equaton wth source terms on the rght hand sde. Wth some assumptons, the equaton expresses a lnear wave problem n a medum at rest wth the equvalent acoustc sources le monopoles, dpoles, and quadrupoles derved from the transent flow feld predcted n CFD. Due to the transformaton of the aeroacoustc problem nto a problem of classcal acoustcs, t s called sound analogy. FW-H sound analogy model was added durng the unsteady state calculaton for ts wdely used. Duct surface was selected as FW-H surfaces, derved ponts around drver ears, call them left pont and rght pont, was created as FW-H recevers. Samplng tme was started from 0.03s to 0.1s. M. Gennaro [1] showed the control equaton for FW-H acoustc model as equaton (1): 1 c0 t p x t v ( f ) f n p ( f ) f T H( f ) n x x x (1) where p' s the sound pressure, (Pa), n s the surface normal vector, v n s the surface velocty component normal to the surface, (m/s), the rght three respectvely represent the monopole, dpole and quadrupole source. 4.3 Montorng ponts In order to further understand the characterstcs of the nose generaton and radaton, montorng ponts were set at duct nternal turbulent flow regon and drvers ear poston respectvely. The flud nstantaneous nformaton of these ponts was montored usng LES solver durng the analyss, and the FW-H solver transfer the pressure fluctuaton nformaton usng Fast Fourer Transform (FFT) method. The poston of the montorng ponts s shown n Fg. 8. Fg. 8 The poston of the montorng ponts 4.4 Far-feld nose predcton results Fg. 9 ndcates that the nose duct surface produced has no obvous pea n the whole frequency band, whch means as the nose that the duct surface generated s broadband nose, and t s accurate to use the broadband nose source model (BNS) for nose predcton. In addton, duct nose can be reduced by changng duct geometrcal structure. Copyrght TAETI

8 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp The sound pressure level ranges from 40 db to 10 db near 70 Hz. In the whole frequency band, sound pressure level at vertcal secton montorng pont 1 s maxmum, ts sound pressure level ranges from 60 db to 1 db, sound pressure level at baffle area montorng pont 3 s moderate, ts sound pressure level ranges from 50 db to 118 db, and sound pressure level at baffle area montorng pont s mnmum, ts sound pressure level ranges from 40 db to 110 db. Ths ndcates that duct vertcal secton s pulsaton s most ntense and t s the man nose source, whch relatve to the flow feld results and BNS predcton results, and nose reducton n ths regon s the focus of the wor. Fg. 10 ndcates the spectral dstrbuton around the left and rght drvers ear poston are almost n same. The entre sound pressure band value ranges from 15 db near 1800 Hz to 57 db near 80 Hz. The sound pressure n the low frequency regon s normally above 40 db, and the sound pressure n medum hgh frequency regon s between 0-40 db. Qang Gu [13] sad n hs boo that humans senstve frequency band s Hz, duct sound pressure level n ths band s about 3 db. Fg. 9 Sound spectrum for orgnal duct nternal montor ponts Fg. 10 Orgnal duct sound spectrum for drver ears 5. Controllng duct aero acoustc nose 5.1 Optmzng duct geometry structure The analyss results of the flow feld and the sound feld of the orgnal duct ndcate that the vertcal secton s the man source generatng aero acoustc nose, and the sound feld s related to flow feld, so general flow feld mproved measurement can be taen to mprove sound feld. Because the connecton structure of trapezod body and vertcal secton changes suddenly, hgh speed arflow flows through here and the speed drecton and magntude changes dramatcally, whch produces a severe dsturbance that forms the aerodynamc nose. Duct connecton structure was optmzed wth a round corner to mprove the aero acoustc characterstcs. Orgnal and Improved structures are shown n Fg. 11 (a) and (b) respectvely. (a) Orgnal (b) optmzed Fg. 11 Trapezod structure schematc Copyrght TAETI

9 186 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp Flow feld results analyss The Realzable K-E model was used for the optmzed duct flow feld analyss, and the average flow feld nformaton throughout the process was obtaned. The flow feld dstrbuton was shown as Fgs Fg. 1 Duct baffle area velocty cloud Fg. 13 Duct statc pressure contours and streamlne Fg.14 Duct turbulent netc energy contours It can be seen from Fgs that the optmzed duct maxmum velocty of duct baffle area was m/s, the statc varaton ampltude was 98.5 Pa, and the maxmum turbulent netc energy was J/g. Compared wth the ntal duct, the man flow feld characters of the optmzed duct were all reduced, ndcated that the optmzed duct flow feld was mproved. 5.3 Nose predcton results Curle and Proudman sound models were used to analyze the approxmate sound feld for the optmzed duct. Same montorng ponts were arranged. Optmzed duct near-feld nose predcton results are shown wth Fg Fg. 17. Sound spectral curves are shown as Fgs Fg. 15 Duct surface dpole nose source dstrbuton Fg. 16 Duct baffle area quadrupole nose source dstrbuton Copyrght TAETI

10 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp Fg. 17 Duct vertcal secton quadrupole nose source dstrbuton It can be seen that the dpole nose at vertcal secton area and duct baffle area of the optmzed duct were 84 db and db, the quadrupole nose at these two places were about 70 db and 56 db. Compared Fgs. 5-7 wth Fgs that optmzed duct dpole nose decreased about 7 db than orgnal duct, and the quadrupole nose decreased about 5 db. The sze of nose source s smaller than orgnal duct. The ntensty of the nose source s smaller than orgnal duct. These ndcated that the sound feld of the duct was mproved from changng the duct structure. Fg. 18 Sound spectrum for optmzed duct nternal montor ponts Fg. 19 Optmzed duct sound spectrum for drver ears It can be seen from Fg. 18 that sound pressure level at vertcal secton montorng pont 1 ranges from 18 db to 50 db, sound pressure level at baffle area montorng pont 3 ranges from 10 db to 40 db, and sound pressure level at baffle area montorng pont ranges from 110 db to 0 db. Comparng Fg. 9 wth Fg. 15, the sound pressure level of each montorng pont has decreased after the modfcaton, n whch the montor 1 decreased most obvous, and the sound pressure level decreased n the hgh frequency band sgnfcantly. Although ther maxmum sound pressure level ncrease, at such low frequency band about 70 Hz, t can be neglect. It can be seen from Fg. 19 that sound pressure level near drves ears ranges from 57 db to 10 db. Comparng Fg. 10 wth Fg. 16, the sound pressure level of the optmzed duct s about 17 db at humans senstve frequency band, lower than the sound pressure level of the orgnal duct 3 db for 15 db, the nose reduce effect s obvous. 6. Concluson In ths paper, a combned CFD and CA approach for the numercal predcton of flow-nduced nose n a locomotve cab coolng system ducts s nvestgated. The so-called CAA approach conssts of two man parts: a CFD calculaton of the flow feld for the determnaton of sound sources, and the CA calculaton of the sound predcton and sound propagaton to obtan the resultng sound feld. Copyrght TAETI

11 188 Internatonal Journal of Engneerng and Technology Innovaton, vol. 5, no. 3, 015, pp Analyss results show that Broadband nose source models (BNS) can effectvely predct the locaton of duct nose source, and provde a bass for optmzaton the nose characterstcs of the duct. For the optmzed duct structure, hgh frequency sound pressure level decreases 15dB, and the nose reduce effect s obvous. In order to reduce the aero acoustc nose, t s necessary to reduce structural breas, and the connecton between varous structures should be optmzed to reduce the degree of arflow dsturbance. Although the presented results show good performance, further analyses are needed to nvestgate more closely the several boundary condtons used for the numercal smulaton wth CFD and CA. References [1] S. Huang, Research on numercal smulaton of aerodynamc nose outsde the hgh-speed tran, Master Thess, Graduate School of Vehcle Movement Engneerng, Central South Unversty, 009. (In Chnese) [] J. Munoz-Panagua, J. Garca and A. Crespo, "Genetcally aerodynamc optmzaton of the nose shape of a hgh-speed tran enterng A tunnel," Journal of wnd engneerng and ndustral aerodynamcs, vol. 130, pp , May 014. [3] F. J. Wang, Computatonal flud dynamcs analyss, 1st ed. Beng: Tsnghua Unversty Press, 004. (In Chnese) [4] J. P. Yao, "Calculaton analyss and control for ar condtonng system aerodynamc nose of a medum bus," Jln Unv., Graduate School of Automotve Eng., Master Thess, Chna, 01. (In Chnese) [5] Z. Y. Zheng, "A study on the numercal smulaton of hgh-speed vehcle s external aerodynamc acoustc feld," Doctoral Thess, Graduate School of Vehcle Eng., Southwest Jaotong Unv., Chna, 01. (In Chnese) [6] A. Zanon, M. De Gennaro, H. Kuehnelt, D. Langmayr, and D. Card, Investgaton on the nfluence of mesh topology and free stream turbulence ntensty n broadband nose predcton of axal fans, Proceedngs of ASME Turbo Expo 014: Turbne Techncal Conference and Exposton, Jun. 014, pp. V01AT10A07. [7] F. Mendonca, A. Read, V. G. Slva, and F. H. J. Imada, Effcent CFD smulaton process for aero acoustc drven desgn, SAE Techncal Paper , October 010. [8] C. Schram, "Aboundary element extenson of Curle's analogy for non-compact geometres at low-march number," Journal of Sound and Vbraton, vol. 3, pp , 009. [9] D. W. Wundrow and A. Khavaran, "On the applcablty of hgh-frequency approxmatons to Llley's equaton," Journal of Sound and Vbraton, vol. 7, pp , 004. [10] T. O. Imano, Y. Hragur, and Y. Kamoshda, "Lnearzed Eurle smulatons of sound propagaton wth wnd effects over a reconstructed urban terran usng dgtal geographc nformaton," Appled Acoustcs, vol. 74, pp , 013. [11] S. Sarar and M.Y. Hussan, "Computaton of the sound generated by sotropc turbulence," NASA Contact Report , ICASE no , pp. 1-6, Oct [1] M. Gennaro, and D. Card, Ffowcs Wllams- Hawngs acoustc analogy for smulaton of NASA SR propeller nose n transonc cruse condton, ECCOMAS CFD, 010. [13] Q. Gu and C. T. Wang, Nose controllng engneerng, 1st ed. Beng: Coal Industry, 008. (In Chnese) Copyrght TAETI