Noise-producing mechanisms in fans

There are three principal noise generating agencies at work in the production of a fan’s total acoustic output. These may be summarised as follows:

• Aerodynamic

• Electromagnetic

• Mechanical

In most industrial fans, the order given is indicative of their rela­tive importance, although for units at the extremities of the size range, mechanical noise becomes an increasing hazard. Elec­tromagnetic noise, as would emanate from an electric motor, is often masked by the aerodynamic noise, especially where, as with a direct driven axial flow fan, this driving unit is contained within the casing and, therefore, the moving airstream. It can, however, be of great importance in slow speed machines driven, for example, by 6 to 12 pole motors which are inherently more noisy. In these cases, the electromagnetic contribution may be of a higher magnitude than the aerodynamic signature, especially in the lower frequency domain.

For centrifugal fans, where the motor is usually outside the airstream, electromagnetic noise will not contribute to the in­duct sound power level. It may, however, mask the breakout noise from the fan casing and ducting system. Many electric motors used with such fans are of the totally enclosed fan venti­lated type, and in these the cooling fan may itself be the domi­nant noise source in the free field around the unit.

Aerodynamic

There are three recognised ways in which acoustic energy may be derived from the kinetic energy produced by a fan impeller in its action on the airstream (Figure 14.11). They are, in de­scending order of radiation efficiency:

Monopole source: The most efficient generating mechanism in which the conversion from kinetic to acoustic energy is achieved by forcing the gas within a fixed region of space to fluctuate. This may be visualized as a uniformly radially pulsat­ing sphere surrounded by a perfectly homogeneous material of infinite extent, such that no end reflections occur.

Dipole source: This is thought to be the predominant sound generating mechanism in low speed turbo machinery such as

MONOPOLE

DIPOLE

A dipole oc — 8x

, F(v, At)

A quadrupole <

-.y. Vj. p.D

ФM (t) 5t

ФXj 8x,

R

Equ 14. 11 Equ 14.12

OcP

Now we know that the air power P, i. e. the power absorbed by the fan impeller PocPD5N3 xfn (ReF) We may therefore state that: Sound Power Wn .DN

P— ocPMaF xfn (ReF) for a Monopole 0

: PMaF3 xfn (ReF) for a Dipole

DN

D = characteristic dimension, is recognised as the Impeller tip diameter (m) N = impeller rotational speed (rev/sec) Thus: For a fluctuating mass or monopole the generated sound power PD2 v4 pD6 N4 Oc ——- oc ———- C c For a fluctuating force or dipole the generated sound power PD8 N6

V OC 7lDN where:

PD2 v6

1 5M (t)

A monopole ce

St

Fans. Energy conversion requires the momentum within a fixed region of space to fluctuate, the process being equivalent to a uniformly pulsating sphere oscillating in the x-direction as a rigid body. Alternatively, it may be thought of as two adjacent monopoles where one is at its maximum dimension, when the other is at a minimum. Thus the dipole is vibrating along one axis. This accounts for the directional nature of the sound gen­erated, the normal particle velocity on the sphere surface being a function of its polar location. Quadrupole source: This is the least efficient energy conver­sion mechanism in which sound is generated aerodynamically, with no motion of solid boundaries, as in the mixing region of a jet exhaust. Within a fixed region of space, there is no change of either mass or momentum. Energy conversion is achieved by forcing the rates of momen­tum flux across fixed surfaces to vary. Momentum flux is the rate at which momentum in the Xi direction is being transported in the Xj direction, with corresponding velocities Vi, Vj. A quadrupole source may be modelled as a double dipole, both oscillating along the same axis. It exhibits complex directionality. The acoustic pressure generated by these different sources may be deduced as follows:
C3 C3 Forturbulent mixing orquadrupole the generated sound power
PD2 vB pD10 N8

OcPMaF5 xfn (ReF)for a Quadrupole Equ 14.13

Is the

Where:

M(t) = rate of addition of mass from the neighbour­

Hood of the source to its surroundings

R = polar distance to the observer

Rsp = radius of the sphere

X = direction of oscillation

V = momentum flux velocity

D = characteristic dimension

P = ambient density of the air or gas

V = air or gas viscosity

At = temperature change across the region

Generally the dissipation of acoustic energy into heat by viscos­ity and heat conduction, is negligible over distances of less than say 100m, in which case the viscosity and temperature defect terms in the quadrupole equation may be neglected.

The equations detailed above may be applied to single sources, but within the acoustic field of a fan, the degree of radi­ation will depend also on the level of phase cancellation be­tween adjacent sources. Indeed, this whole question of phase difference is seen as the way forward in the reduction of fan noise. It is leading to the introduction of scimitar-shaped blades, angular cut-off pieces and other devices.

It was Lighthill who first applied dimensional analysis to the acoustic power radiated by the different sources of sound pres­sure and derived the proportionality relationships with respect to velocity. The writer has, however, extended these identities in the final column by recognising that, in a homologous series of fans, all velocities will be proportional to the impeller tip ve­locity, i. e.

DN

"c”

JtDN

As, by the re-introduction of n, we can recognise that

Mach number related to the impeller tip speed, i. e. MaF.

In high speed fans this can approach 0.3. The Reynolds num­ber function has the effect of reducing these indices.

We can also see that in a homologous series of fans, the gener­ated sound power Wn oc DKNl where k must lie between 6 and 10 whilst i is some number between 4 and 8.

Overall sound power radiation for any homologous series of fans will have a sound power/rotational velocity relationship, which depends on the relative contributions of the three sources. However, it is not simply a matter of how an acoustic mechanism varies with a typical speed, but rather how the flow conditions related to that acoustic mechanism vary with speed.

Whilst a considerable amount of work has been done in at­tempting to define a consistent relationship between fan rota­tional speed and the generated sound power, unless strict simi­larity is ensured, or design variations accounted for, the empirically derived equations may give rise to considerable er­ror. Consequently, results from various researchers differ and the exponents have been variously quoted between 6 to 8 for k and 4 to 6 for i. It should be noted here that the Beranek formula and its extrapolations assume i = 5 as power absorbed oc Qp and Q oc v, p oc v2 and the pressure term has a coefficient of 20.

The first theoretical study of noise from rotating machinery was probably that of Gutin in 1936. His basic equation assumed a steady state where the blade loading distribution was inde­pendent of time. Here an element of gas within the area swept by the rotor was considered to receive an impulse periodically with the passing of a blade. The impulses were treated as a se-

MONOPOLE blade thickness noise discrete

DIPOLE blade forces discrete + broadband

QUADRUPOLE turbulance noise broadband

FAN NOISE discrete — broadband

Ries of dipole sources distributed throughout the swept area, and of constant strength at any radius. The dipole source am­plitudes were obtained from the thrust and torque loading con­ditions, the fundamental frequency of the noise generated be­ing zN, where z is the blade number and N is the rotational frequency (revs/sec). The resultant sound field can be ana­lysed into a series containing the fundamental frequency and its integer harmonics. It is assumed that the acoustic pressure satisfies the homogeneous wave equation:

^-C2^=0 Equ 14.14

St 5x

The fluid surrounding the blade surfaces must, therefore, have velocities which are low compared to the speed of sound, such that acoustic waves can travel radially from their source, this may not be the case and it is then necessary to consider the fluid as a perfect acoustic medium containing quadrupole sound sources of T, j = pvh Vj + py — C2 5^.

As previously stated, the last two terms in this stress tensor may usually be ignored as the quadrupole strength density be­comes equal to the “fluctuating Reynolds Stress” of the gas around the blades.

It is, therefore, possible to itemise the source components of the whole radiation field such that sound produced by a fan may be regarded as generated by monopole sources related to vol­ume displacement, dipoles distributed over the machine sur­faces and quadrupoles of strength density T^ distributed throughout the surrounding gas.

Lighthill’s acoustic analogy was to regard density variations within the gas as being driven by a source distribution

For the general case of an unbounded fluid, but in the real world, solid boundaries are present. Modifications to the theory are, therefore, necessary to take account of reflections at these sur­faces and also for an uneven quadrupole distribution as these may only exist external to the blades. These have been consid­ered by Curie and John E Ffowcs Williams who have taken into account surface force distributions and moving boundaries.

Practically, sources of aerodynamic noise within a fan may be grouped under the following headings:

• thickness noise due to the passage of blades through the air

-a quadrupole source

• torque and thrust noise — quadrupole sources

• rotation noise due to the blades passing a fixed point e. g. cut-off — a dipole source

• vortex shedding due to flow separation from the blades — a dipole source with some Reynolds number dependence

• air turbulence noise due to shear forces when the blades are stalled — a quadrupole source

• interference noise due to contact between turbulent wakes and obstructions

• pulsation noise — where at high system pressures the flowrate regularly varies and a pitched tone is produced a the frequency of the pulses — a monopole source.

An overall assessment of the aerodynamic generating mecha­nisms has been made by Neise and these are shown in Figure

14.12.

It will be noted that both pure tones (discrete frequencies) and broadband (random) noise is produced. Rotating blades dis­place a mass of gas periodically and generate sinusoidal pres­sure fluctuations in the adjacent field so that thickness noise is found in all but the very highest pressure fans, the acoustic radi­ation efficiency is low and thickness noise is not, therefore, of great importance.

Often a fan will operate in a duct system where the approaching airstream is not fully developed. The velocity profile may be “peaky”, contain swirl, or indeed be axially distorted. Thus its impeller will be subjected to unsteady fluid forces, since both the magnitude of these velocities and their angle of attach will change with angular position.

Tyler and Sofrim have shown that the phase velocity of these unsteady blade forces may be much higher than the relevant impeller peripheral speed, and even be greater than the speed of sound. Their acoustic radiation efficiency will thus be very high and tonal noise will be produced at blade passing fre­quency and its harmonics. The usual cause of such noise will be the presence of bends or transformation pieces adjacent to the fan inlet. Even sagging flexible connections can be a prob­lem. In the fan design itself, upstream guide vanes or motor supports can cause wakes before the impeller and again result in unsteady blade forces.

The most important source of noise in a well-designed fan and duct system is due to vortex shedding from the backs of the im­peller blades. This is a dipole source and is usually broadband, although instances of discrete frequency have also been noted. Thus the noise generated in such fans Wn oc v6 oc D8N6.

The spectral shape of the noise from a fan varies according to its design. In very general terms, an axial flow fan may have

High noise in the octave band containing the blade passing fre­quency, zN (Blade number x rev/sec) with a declination of around 2 dB per octave on either side. The peak at blade pass­ing frequency can exceed the general spectral level by 4 to 10 dB, being especially severe where the impeller is eccentric in its casing. There may also be additional tones generated at inter­active frequencies determined by (blades + vanes), (blades- vanes) etc., the strength of these being dependent on the gap

Between them, and the ratio No.

Vane No.

Furthermore, much recent testing of axial flow fans has shown high noise levels in the 31.5 Hz and 63 Hz bands.

Perhaps there has been too much extrapolation of idealised spectra in the past. It should be remembered that in the 1950s and 1960s, measurement of noise below the 125 Hz octave was next to impossible with the state of instrumentation and knowledge at that time.

A centrifugal fan will have a spectrum with its peak towards the lower frequencies. The declination is of the order of 3 to 7dB per octave band dependent on blade shape, but this general statement requires a host of provisos. In backward-bladed fans, the blade passing tone and its harmonics may be of espe­cial importance. With the flat inclined type, they are easily iden­tified above the general broadband background. With back­ward-curved blades, they are not so pronounced, and are lowest with backward aerofoil designs.

Sound waves produced by a source within a duct will also un­dergo reflection, interference and decay according to the fre­quency of the emitted wave. Centrifugal fans usually run at lower Mach numbers than axial fans and the predominant tones have wavelengths larger than characteristic impeller or duct dimensions. The overall radiated sound power may be greatly affected by reflection properties of the casing and ductwork. This can lead to some distortion of the sound power and directivity pattern, especially at low frequencies.

Whilst an uncased centrifugal impeller usually gives a flat fre­quency spectrum, the addition of a case leads to enhancement of the noise at well defined frequencies, related to the casing geometry. Flowrate variations do not significantly affect the overall shape of the cased spectra, although the magnitude, in particular frequency bands, can vary.

It is clear, therefore, that the overall radiated sound power can be quite different from the generated power. The casing may act as a Helmholtz resonator and a major casing dimension may relate to the wavelength of some important frequency. Overall, this can mean a reduction in the speed and size indices

SOOfflm FAN -!N OUCTLwцBfe 10 ^ WATTS ROTATIONAL SPEED — rpm Figure 14.13 Sound power levels for a mixed flow fan at a range of rotational speed

Over most of the fan performance envelope with sudden in­creases at identifiable speeds (Figure 14.13).

In the example shown, the first peak was seen to be where the blade passing frequency coincided with the duct cut off fre­quency (change from plane wave propagation to more complex modes). The second peak occurred where impeller resonance coincided with the second harmonic of blade passing frequency.

Electromagnetic

Whilst a very small number of fans may be driven by prime mov­ers such as steam turbines or petrol engines, the vast majority

— in excess of 98% — are driven by electric motors. With axial flow fans, it is common for the fan impeller to be mounted di­rectly on the motor shaft extension. Centrifugal fans, may, of course, be vee belt drive or directly driven either through a flexi­ble coupling with or without an intermediate gearbox (this is common in the UK on large mine ventilation fans).

Again with the majority of fans, electric motors are of the totally enclosed squirrel cage induction type suitable for a three phase supply. Single phase motors are usually limited to fractional horsepower outputs.

The induction motor is extremely reliable and robust. In nearly all cases it may be considered symmetrical both mechanically and electrically. The windings are balanced between phases and slots. Care is taken to ensure that the rotor runs in the cor­rect position axially within the stator field, and that the airgap between the rotor and stator is the same at all axial and radial positions. However, especially with direct driven fans, there will be an end thrust due to the impeller action and this will “try’’ to take the rotor out of the magnetic field, being resisted by the magnetic forces and also such devices as wave washers in the bearing housings.

Skewing of rotor slots is often resorted to, to improve starting performance, and has also been considered as a means of re­ducing magnetic noise. This, however, has been the subject of much debate. Certainly an axial thrust is generated which may lead to increased noise emission.

Many fans are driven by 2 pole motors running at approximately

49 rev/sec on a 50 Hz or 59 rev/sec on a 60 Hz AC supply. If the rotor does not run in the centre of the stator, or if the stator core presents an unequal reluctance path, then a homopolarflux is generated which tries to circulate through the core, along the shaft returning via the end cover plates and frame. This causes noise and vibration at twice line frequency.

The heart of an induction motor is its laminated iron core and the stator and rotor windings. As the core is in no way con­nected to the power supply nor is power directly removed from it, it can be considered as passive. It is, however, the path of minimum resistance for the flux generated by the magneto motive force (mmf) set up by the stator winding, which itself is the path of least resistance for the input current.

Magnetic noise is produced by vibration of the laminations, its form being complex and taking place about all axes.

The problems of producing a low noise electric motor are se­vere. Yang has “de-mystified” the subject to a very large extent and shown that the noise emitted by a motor depends not only on the electromagnetic forces but also on the response to those forces by the motor carcase, and end — shields and to their radi­ating characteristics. He has also shown the value of parallel path winding.

The rotor must be concentric with the stator bore, and this re­quires that the bearing and end-shield location and stator pack tolerances all be closely controlled during manufacture. Bear­ing housings and end-shields need to be sufficiently rigid to avoid distortion during assembly. If the motor casing is of fabri­cated construction, stress-relieving is desirable before the final machining operation. In general terms, the greater the size of iron core per kilowatt of output at a given speed, the lower will be the level of magnetic vibration and noise.

Other features that have an effect are:

• core material, size and geometry,

• natural frequency of the core, core-to-frame fit and core-pack axial pressure,

• lamination insulation and burr height, number of stator slots, type and fit of stator coils,

• type and fit of slot wedges, pitch of coils, connection of coils and coil groups,

• impregnation, number of rotor slots, air-gap length and frame stiffness.

In summary, the power supplied to a three-phase stator winding sets up a rotating magnetic field. This induces an opposing cur­rent in the rotor winding and thus another magnetic field. Inter­action of these two fields produces a tangential force. As the ro­tor shaft is only restrained by its bearings, it has to rotate.

Viewed from a fixed point on the rotor, the air-gap performance around a rotor with R slots will have R cycles of variation. Simi­larly, a stator with S slots will produce S cycles of variation. As the powerto the stator has a frequency f, Hz, and as the winding is distributed around the stator in slots, the stator will produce vibrations, and therefore noise, proportional to field strength squared, related to the supply frequency, winding pitch and number of slots per pole-pitch.

Harmonics will also be present and, together with all the inter­active frequencies, a very complex situation results. The rotat­ing magnetic field of the stator produces low frequency vibra­tion and noise, whereas rotor slot performance variation and its reactions with supply frequency lead to higher frequencies.

These may be calculated from:

(R xf1)-2fL, Hz	Equ 14.15
Rxf„ Hz	Equ 14.16
(SxfO + 2fL	Equ 14.17

Where

FL = line frequency

F-i = rotational frequency

When R > S, equation 14.15 is usually of more importance. If

S > R, equation 14.17 predominates. Again, many harmonics will be present.

At the design stage, the stator-rotor slot combination can be chosen to minimise vibration. To achieve this, the number of vi­bration nodes should be as high as possible:

Number of nodes =(2R-2S)+ 2P Equ 14.18

Where:

P = number of poles

Nevertheless, the “magic” combinations of stator/rotor slot numbers should be viewed with suspicion at the very least.

Forces in the air-gap between rotor and stator tend to pull these together and produce vibration at double the line frequency. Normally, this vibration is small, except in 2-pole motors, and if the air-gap varies, or if the tightness of stator laminations or winding in the stator varies. The second and third harmonics may also be important.

In general, slip frequency (= fL — fi Hz) will not in itself be impor­tant, as it will be of very low frequency. Its interaction with higher frequencies can, however, produce pulsations.

If the rotor is severely unbalanced, the high spot will come closer to the stator than other points. As it passes the stator poles, a greater pull is exerted and the vibration occurs at dou­ble the slip frequency on a 2-pole motor. The magnitude of the readings in this frequency can indicate whether the problem is simply due to the lack of balance, a change in the air-gap, worn journals, broken rotor bars, etc.

If a resonance condition exists within the motor at the line fre­quency, large vibrations can be produced. More often this is a result of an unbalanced magnetic pull and can be overcome by changing stator connections.

With suspected electrical sources of noise and/or vibration, a simple check is to switch off the motor, when they should “die”, This is the opposite to mechanical sources, which will gradually decay with decreasing fan speed. The translation of vibration into noise will depend on the constructional stability of the motor and, therefore, the “radiation efficiency” of vibrating surfaces.

From all the above, it will be appreciated that the prediction of motor noise at the design stage is nearly impossible and that similarity rules to interpolate/extrapolate the measured noise from one frame-size to another, do not exist. It is fortunate for the motor designer (and unfortunate for the fan engineer) that except in the case of low synchronous speed motors (6 or more poles) the fan noise often masks the motor noise.

Care must, however, be exercised with all motors subject to variable speed control through inverters. The electrical wave­form may be distorted sufficiently from the ideal sinusoidal shape, that the motor noise may increase with reduced speed such that it dominates the fan noise.

Mechanical

Sources of noise under this heading are legion. Those of most importance to the fan designer are, however, restricted to a small number and may be categorised as follows:

• Bearings

• Couplings

• Gearboxes

• Vee belt drives

• Component vibration Bearings

Bearings used in fans are of two main types:

• plain

• rolling element.

Plain bearings, whilst used to a great degree in the past on slow speed centrifugal fans, are not now nearly so popular in ventila­tion applications. Of recent years, therefore, their use has been confined to the larger, special purpose fans where their ability to handle high journal and thrust loads is desirable. This may re­quire tilting load pads and/or forced lubrication.

Except for the very lightest loads when porous lead impreg­nated or PTFE bushes may be used, plain bearings are oil lubri­cated to minimise sliding friction. The performance of the bear­ing, in fact, depends on maintaining an oil film between the shaft and journal under the load and temperature conditions im­posed. Where the fan is handling hot gases, a water jacket may be included within the housing to take away the heat transmit­ted along the shaft and in turn, to the oil (which would otherwise lose its lubricant properties).

,Hz

1 — cos A D

D 2 . -^rcos A

,Hz

1+ —cos A D

,Hz

N = number of balls or rollers D = diameter of balls or rollers D = pitch circle diameter of race A = angle of contact of ball/rollers Fi = fundamental frequency (equivalent to N Rev/sec) It should be noted that such vibrations are attenuated before being transmitted to the rest of the fan and emitted as noise. They are therefore best recognised by vibrational velocity read­ings on the bearing housing. Severe misalignment of a race will sometimes result in vibration at a frequency of n x fi Hz, even when the bearing itself is satisfactory. In summary, modern ball and roller bearings are manufactured to a high standard and with correct installation/lubrication they are unlikely to increase the fan noise. Where the noise does in­crease, it is more often the fault of vibration due to imbalance, misalignment or use at speeds/loads/temperatures in excess of those recommended by the manufacturers. When faults are present, noise levels at the relevant frequencies may be as much as 7 dB greater than the readings of a good bearing. Great care should be taken in the selection of shaft and housing limits. An interference fit of the bearing to the shaft and a small clearance between the outer raceway and the bearing housing are preferable. Bearing end caps should be of a substantial de­sign, incorporating a sufficient number of setscrews or bolts, but differing from the number of balls or rollers. The demand for high quality and low price necessitates quan­tity production of all anti-friction bearings. Machine designers are required to select from a standard range, the items that most closely meet their requirements covering: dimensional and speed properties, frictional drag and heat generated, noise output, deflection under load, rate of wear and lubrication and life in relation to load. Of these, the life is probably of most importance, especially at the normal speeds and loads of these fans. Correct selection for life usually ensures that performance under the other headings is also acceptable.

Where

,Hz

1—- cos A D

Variations in the surface finish of shaft and journal and the means of circulating the lubricant are, therefore, the only cause of any noise emitted and these bearings do not usually contrib­ute to the fan noise signature, being effectively masked by other components. The more popular bearings in fan use are rolling element, or “antifriction” types, as they require considerably less mainte­nance, have reduced “stiction” at start-up, and are less re­stricted in the attitude at which they can operate. Grease lubri­cation is particularly favoured and in many cases, the race can be sealed-for — life. A rolling element bearing consists of four sets of working com­ponents as compared with the one in a plain bearing, these be­ing: • outer race • inner race • elements (balls or rollers — cylindrical, taper or spherical) • cage for maintaining the relative positions of the elements. The operation is a combination of rolling and sliding contact. Rolling element bearings are considered to have point (ball) or line (roller) contact between the raceways and the elements. In reality, these conditions cannot exist where a load is applied, since the smallest force would induce an infinite stress. Defor­mation, therefore, takes place and this leads to the emission of noise. The contact is over an area sufficiently large to result in a stress value that can be accepted by the bearing materials. To ensure that the stress is within the elastic limit, and to keep the contact area to a minimum, the steels used are hardened. High stresses, nevertheless, result so that under normal use, the major cause of failure is fatigue, which leads to flaking of the raceway and elements and a marked increase in noise. It has been shown by Glew that the noise emitted by a rolling el­ement bearing is a direct function of its internal clearances. Un­fortunately, many users are now requesting C3 increased clear­ance bearings as these are less susceptible to misalignment and, therefore, require lower skill levels by maintenance staff during replacement. Where loading and application permit, ball bearings should be preferred to roller. An initial preload on the outer race of the bearing by a spring waved washer will also control bearing clearances (Figure 14.14). Tapered roller thrust bearings in vertical motors have been shown to increase the noise of a 450 kW 2 pole machine by 10dB in the 2 kHz octave band, compared to the same machine	Running horizontally and fitted with ball/cylindrical roller bear­ings. Bearings are often incorrectly installed which can lead to an in­crease in their noise emission. Even the very smallest misalign­ment (well within acceptable manufacturing tolerances) can be detected. Less frequently, flaws may be present on the ele­ments and these usually result in an increase in the high fre­quency noise. These faults may be detected from a vibrational frequency analysis: Flaw in outer raceway or variation in stiffness around housing
Equ 14.19
F2=fi
Flaw in inner raceway
Equ 14.20
F3=v
Flaw in ball or roller
Equ 14.21
Irregularity in cage or rough spot on ball/roller
Equ 14.22
F5=fi
Spring waved washer Spring material thickness plus anticipated rotor shaft expansion Figure 14.14 Control of bearing clearances

Couplings are not a dominant source of noise. Where misalign­ment is severe, they can lead to the vibration of adjacent parts and this, in turn, leads to an emission of noise dependent on the radiation efficiency of the material and its geometry.

Where torsional oscillation is present, the interaction of the cou­pling elements may also lead to noise dependent on the materi­als involved and the amount of deformation which takes place.

Gearboxes

Gearboxes are only incorporated in special purpose units such as the fans for the main ventilation of coalmines. By this means, cheaper, higher speed motors can be used to direct drive the fan at the relatively low speed required. Pinion changes can also be made where development of the mine tunnels dictates an increase in duty. Vee belt drives are usually impractical due to the very high powers involved up to about 2500 kW.

Even gears with a perfect involute form emit noise due to de­flection of the teeth under load, and more importantly, the sud­den changes in deflection as the load is shared and changed between differing numbers of teeth. Noise is, therefore, emitted at the meshing frequency and its harmonics. Where the gear­box contains more than two pinions to give the necessary speed reduction, side band frequency noise will be generated at the sums, differences and products of the fundamental frequencies.

Vee belt drives

These are not a source of noise except in so far as windage may be a problem with the larger spoked pulleys. Unless there are faults such as unbalance or misalignment, they can, there­fore, be ignored in this analysis.

They are, of course, an extremely popular form of power trans­mission with centrifugal fans up to about 300kW as they enable the fan speed to be matched to the duty requirements and also have a good resistance to shock and vibration. Sometimes, be­cause they may be seen to whip and flutter, especially when the belts are unmatched for length, they are incorrectly identified as a source of noise.

Vibration from faults in pulleys and belts may be transmitted by adjacent flat metal surfaces where these are of sufficient size. Belt faults are identified at multiples of belt speed. The relevant frequencies are:

TOC o "1-5" h z. „ „ . pulley diameter, ,, r

1,2, 3 or 4 x——————— ————————— x71 xf Hz Equ 14.23

Belt length p

where:

Fp = pulley rev/sec

Likely faults are pieces broken off, hard or soft spots etc.

Faults in pulleys, such as chipped grooves etc., will be identified at the speed of the relevant pulley fpHz.

Component vibration

Vibration can be a source of noise subject to certain conditions. Usually such vibration is itself due to some fault within the fan such as imbalance, misalignment, looseness, increased clear­ances, etc. Every component will have a natural frequency at which it likes to resonate. This will be “resisted” by the effects of inertia, stiffness and damping. Aerodynamic forces can excite casing panels.

Basically, any semi-rigid flat sheet surface in the fan such as the casing side plates or bearing pedestals, can act as a noise radi­ator where its size is equal to or greater than the wavelength of the vibration frequency transmitted to it. Its efficiency as a noise producer will be inversely proportional to the self-damping properties of the material used.

As wavelength X = — where:

C = the speed of sound (m/s)

F = frequency ( Hz)

It follows that sound at 100 Hz could be transmitted by an un­supported panel of 3.4 m width, this reducing to 340 mm at 1 kHz. The need for stiffening and adequate metal thicknesses is, therefore, apparent.

Airborne noise will be emitted from any resonant point, whether an efficient radiator or not, where the excitation frequency coin­cides with the natural frequency of the element at that point, or with one of its modes as defined by its resonant frequencies.

It will be seen that component noise should not be a problem in a well-manufactured and designed fan. Where a fan has to op­erate at a range of speeds, however, it may be subject to reso­nance in some component. Often the energy in this resonance will be insufficient to cause failure, but may lead to an unfore­seen increase in noise.

It might be thought that mechanical and electrical sources of noise would be masked in all cases by those of aerodynamic or­igin. There are, however, a number of examples where this is not the case.

To isolate non-aerodynamic sources is difficult. The usual method is to replace the fan impeller by a solid disc of the same weight, so that bearing loads and drive losses are the same. This method does not, however, reproduce any end thrust ef­fects, nor is the electric motor under load. End thrust may be re-introduced by tilting the assembly as shown in Figure 14.15. With the addition of a belt dynamometer, the motor will be loaded, when its noise level will increase by up to 5 dB. Inciden­tally, we have found with electric motors that a change in core length has reduced overall fan noise by 3 dB linear.

Figure 14.15 Assembly for measuring mechanical and electrical noise

Posted in Fans Ventilation A Practical Guide